Systems for protein expression in eukaryotic cells

Abstract

Nucleic acid molecules configured to constitutively express a sequence of interest in a eukaryotic cell comprising a transcription factor binding region, a core promoter and a sequence of interest operatively linked to the transcription factor binding region and core promoter are provided. Methods of constitutively expressing a sequence of interest in a eukaryotic cell are also provided.

Description

SYSTEMS FOR PROTEIN EXPRESSION IN EUKARYOTIC CELLSCROSS REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of priority of U.S. Provisional Patent Application No. 63 / 531,576, filed August 9, 2023, the contents of which are all incorporated herein by reference in their entirety.REFERENCE TO AN ELECTRONIC SEQUENCE LISTING

[0002] The contents of the electronic sequence listing (TECH-P-0285-PCT.xml; Size: 85,010 bytes; and Date of Creation: August 6, 2024) are herein incorporated by reference in its entirety.FIELD OF INVENTION

[0003] The present invention is in the field of protein expression.BACKGROUND OF THE INVENTION

[0004] A Major challenge in alternative protein gene expression is to increase protein titers in eukaryotic cell-lines. This is due to the fact that many proteins cannot be expressed in bacteria due to a myriad of post-translation modifications (PTM), which alter their function. Historically, over-expression of proteins had been achieved only in bacteria via the T7- promoter system, which generated a huge boost in levels of mRNA and in subsequent protein levels. To date, neither T7 nor a similar high-level protein expression system has been shown to work in a eukaryotic cell line, creating the need for a new approach to generating protein over-expression in those cells.

[0005] In eukaryotic cell-lines, gene expression is controlled by transcription factors (TFs) that bind to DNA regulatory motifs, together with chemical, structural, and spatial changes in the chromatin state, and binding of the transcriptional machinery to the core promoter. DNA regulatory motifs serve as cis-acting transcription factor binding sites (TFBSs) for activators and / or repressors. The combination of different repressing and activating DNA regulatory motifs enables tuning promoter activity and, as a result, gene expression levels.The sequence space of combined DNA regulatory motifs is often referred to in the literature as the regulatory code or grammar.

[0006] Over the past two decades, many systems and synthetic -biology studies attempted to decipher the regulatory code, using two approaches. In the first approach, large regulatory regions were dissected using a traditional knock-down and rescue approach until the regulatory effect of every active TFBS was characterized. In the second approach, a multitude of synthetic cis-regulatory regions (e.g., synthetic enhancers) composed of a small number of TFBSs arranged in various configurations were encoded in an oligo library (OL) and characterized using massively parallel reporter assays (MPRA) such as SORT-seq, with OL quality confirmations using advanced analysis tools, such as SOLQC. While both these reductionist and constructionist approaches have yielded a plethora of data, the regulatory code remains largely undeciphered. This is due to the fact that in every eukaryotic genome there may be hundreds of thousands of complex cis-regulatory regions, which renders the reductionist approach impractical for high-throughput analysis. On the other hand, the constructionist approach, while promising, currently provides an insufficient coverage of the regulatory space; only a few tens of thousands of variants can be adequately characterized in one study, and the amount of regulatory space that has been studied in these high- throughput approaches is still a tiny fraction of the possible regulatory grammar. Consequently, the dataset that is needed for reliable deciphering of the regulatory code using modem computational methods (e.g., deep neural networks) needs to be expanded. A new method for understanding the regulatory grammar rules of eukaryotic cells and resulting from this improved expression constructs is greatly needed.SUMMARY OF THE INVENTION

[0007] The present invention provides nucleic acid molecules configured to constitutively express a sequence of interest in a eukaryotic cell, comprising a transcription factor binding region, a core promoter and a sequence of interest operatively linked to the transcription factor binding region and core promoter. Methods of constitutively expressing a sequence of interest in a eukaryotic cell are also provided.

[0008] According to a first aspect, there is provided a nucleic acid molecule configured to constitutively express a sequence of interest in a eukaryotic cell, the molecule comprising: a transcription factor binding region comprising at least two sequences selected from thoseprovided in SEQ ID NO: 1-41, a core promoter, and a sequence of interest operably linked to the transcription factor binding region and core promoter.

[0009] According to another aspect, there is provided a nucleic acid molecule configured to constitutively express a sequence of interest in a eukaryotic cell, the molecule comprising: a transcription factor binding region comprising at least two sequences selected from those provided in SEQ ID NO: 45, 42, 3, 30, 32, 36, 39, 43, 44, 46, 47, 48, 54, 55, 56, 57, and 58, a core promoter, and a sequence of interest operably linked to the transcription factor binding region and core promoter.

[0010] According to some embodiments, the transcription factor binding region comprises at least three different sequences selected from those provided in SEQ ID NO: 45, 42, 3, 30, 32, 36, 39, 43, 44, 46, 47, 48, 54, 55, 56, 57, and 58.[Oi l] According to some embodiments, the transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 42, 45, 55, and 57.

[0012] According to some embodiments, the transcription factor binding region comprises repeats of at least one sequence selected from those provided in SEQ ID NO: 45, 42, 3, 30, 32, 36, 39, 43, 44, 46, 47, 48, 54, 55, 56, 57, and 58.

[0013] According to some embodiments, the transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 9-22, 42, 45, 55 and 57.

[0014] According to some embodiments, the nucleic acid molecule comprises repeats of at least one sequence selected from SEQ ID NO: 9-22, 42, 45, 55 and 572.

[0015] According to some embodiments, the at least two sequences are two different sequences.

[0016] According to some embodiments, the transcription factor binding region comprises at least three different sequences selected from those provided in SEQ ID NO: 1-41.

[0017] According to some embodiments, the at least two sequences are separated by a nucleic acid spacer sequence.

[0018] According to some embodiments, the at least two sequences are selected from SEQ ID NO: 2-3, 7, 13, 17-18, 24, 30, 32-33, 36, 39-40 and 42-65.

[0019] According to some embodiments, the at least two different sequences are selected from SEQ ID NO: 3, 30, 32, 36, 39, 42-49 and 55-59.

[0020] According to some embodiments, the transcription factor binding region comprises or consists of between 100-200 nucleotides.

[0021] According to some embodiments, the eukaryotic cells are yeast cells.

[0022] According to some embodiments, the eukaryotic cells are mammalian cells.

[0023] According to some embodiments, the mammal is a human.

[0024] According to some embodiments, the core promoter is selected from an mCorel promoter comprising or consisting of SEQ ID NO: 67, an FEC-mCorel promoter comprising of consisting of SEQ ID NO: 68 and a minimal CMV promoter comprising or consisting of SEQ ID NO: 65.

[0025] According to some embodiments, the constitutive expression comprises expression without the exogenous addition of a transcription factor or transcription activator.

[0026] According to some embodiments, the nucleic acid molecule further comprises a 5’ untranslated region (UTR) between the core promoter and the sequence of interest.

[0027] According to some embodiments, the molecule is an expression vector.

[0028] According to some embodiments, the expression vector is a eukaryotic expression vector.

[0029] According to another aspect, there is provided a method of constitutively expressing a sequence of interest in a eukaryotic cell, the method comprising contacting the eukaryotic cell with a nucleic acid molecule of the invention, thereby expressing a sequence of interest in a eukaryotic cell.

[0030] According to some embodiments, the contacting comprises expressing the nucleic acid molecule of the invention in the eukaryotic cell.

[0031] According to some embodiments, the method an in vitro method and the eukaryotic cell is in culture.

[0032] Further embodiments and the full scope of applicability of the present invention will become apparent from the detailed description given hereinafter. However, it should be understood that the detailed description and specific examples, while indicating preferred embodiments of the invention, are given by way of illustration only, since various changes and modifications within the spirit and scope of the invention will become apparent to those skilled in the art from this detailed description.BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figures 1A-1F: Motif-based sURS OL - schematic and expression data. (1A) A list of all DNA motifs used in the sURSs OL. For each of the 41 motifs, the source organism, sequence logo, sequence used in the OL and the regulatory function are included. (IB) Conversion of a PPM to the mixed-base motif used in the OL. The Pou51 motif from mouse ES cells is provided as an example. (1C) Promoter design consisting of a synthetic URS encoding three motifs within a desert sequence devoid of binding sites. The sURS is positioned upstream from the mCorel promoter that can drive expression of yeCitrine at a low level. (ID) Schematic illustration of OL assembly and experimental characterization. (IE) 3D-heatmap distribution for the number of variants obtained for an integer number of reads with a particular mean FL (calculation of mean FL is described in detail in the methods). Of the 195,000 variants encoded, 147,700 were detected and analyzed. (IF) A 2D-frequency distribution showing the number of variants obtained for a particular number of reads. In our analysis we used only variants with a statistically significant number of reads (i.e., >40) that are located in region 3.

[0034] Figures 2A-2D: Empirical analysis suggests that there are functional up- and down-regulating motifs. (2A) Box-plot distributions of mean FL of all variants containing at least one occurrence of a specific motif. Box-plots indicate a statistically significant difference between the motifs positioned at the left and right of the plot, supporting a regulatory function. Black dashed line represents the mean fluorescence of the 2 desert variants in the OL, while the green dashed line represents the mean fluorescence level of the desert motif. Desert-motif variants contain 2 out of the 41 motifs, with a desert region in the middle of the variant, as a third “motif’. (2B) Median of mean FL as a function of the number of variants, for all variants containing each of the 41 motifs. Positive correlation indicates that up-regulating motifs are more likely to be detected in our experiment. The median value was calculated as the median of mean FLs for each group of variants, containing one of the 41 motifs. (2C) Pearson correlations of pairs of barcodes used to encode the same sURS. (2D) Averaging the Pearson correlation over a range of fluorescence values shows that a high correlation appears either in high- or low-expression variants, supporting a likely regulatory function.

[0035] Figures 3A-3E: Convolutional neural network identifies a set of 2435 variants with robust expression. (3A) Schematic of our convolutional neural network architecture. (3B) Pearson correlation results for various models as a function of the training set size, sorted by variant total read counts. (3C) Comparison of Pearson correlation results on avalidation set for OL-trained models with that of the de Boer model. (3D) Prediction performance comparison of the dBR, ADM, AMM, and MBO models on the 11 sURS variants validation set and the 300 variants with highest read count among the 2435 variants with 22 barcodes. Error bars were computed by bootstrapping which was carried out by subsampling 100 times 9 of the 11 points in each dataset randomly and computing the Pearson correlation for each subset. (3E) Pearson correlations between the predictions of the dBR model and the MBO and AMM models on the 11 sURS variants validation set. Source data are provided as a Source Data file.

[0036] Figures 4A-4F : mHG analysis on the 2435-variant set reveals functional up- and down-regulating motifs. (4A) p-values resulting from mHG analysis for each motif, for variant lists arranged in order of decreasing expression levels (up-regulating) and increasing expression levels (down-regulating). A cutoff of p-value <10-4 was chosen for determining up-(blue) and down-(red) regulating motifs. The p-values for the desert motif are labelled in black, and for undetermined motifs in green. (Inset) box plots for a select set of variants containing a particular motif, compared with the rest of the variant population not containing the motif. (4B-4E) Sub-motif analysis for a select set of motifs. (Right) Box-plot comparison for the inset for the full-mixed-based motif. (Left) Violin plots for all variants containing the various sub-motifs, compared with all sub-variants that do not contain the sub-motif. (4F) Sub-motif analysis for the remaining motifs. Violin plots for all variants containing the various sub-motifs (from 20 motifs containing K and / or M bases) as compared with all subvariants that do not contain the sub-motif. Blue shaded violin plots are sub-motifs that were determined to be significantly activating according to mHG analysis. Red shaded violin plots are sub-motifs that were determined to be significantly repressing, based on the analysis.

[0037] Figures 5A-5G: Single-clone experiments in yeast validate mHG predictions for motifs. (5A) A legend depicting all the different parts in the validation variants: promoters, desert sequences, and various up- and down-regulating sub-motifs. (5B-C) Fold regulation, (5B) activation and (5C) repression, relative to weak (upper plot) and strong (lower plot) promoters, showed as bar plots of various single clones as depicted by the promoter and URS schema next to each plot. In blue- activating variants, in red- repressing variants. (5D) Fold regulation of activating motifs (in blue) and repressing motifs (in red) as a function of the log of p-value computed by the mHG analysis. (5E) The average of fold regulation values for different regulatory architectures, depending on the type of regulatory motifs and the number of motifs in the cassettes. (5F) The average of the median FL values for the weak and strong promoters devoid of any added URSs. (5G) A proposed expression model for thesynthetic yeast promoters, depicting the total expression level of a gene, given its regulatory architecture. This is based on: the number of activating motifs, the presence / absence of repressing motifs, and the core promoter.

[0038] Figures 6A-6H: Synthetic TFs regulate expression in accordance with model predictions. (6A) Schema for parts used in constructing synthetic TFs and LexA-based URSs. (6B-F) Normalized fluorescence results obtained for (6B) LexA-Gal4, (6C) LexA- Mxil, (6D) LexA-Hatl, (6E) LexA-Migl, (6F) LexA-VPR and (6G) LexA alone using 16 different regulatory architectures. Read count for each architecture is noted in parenthesis, in bold- read counts over 100 reads, which corresponding expression levels are less likely to be noisy. (6H) Regulatory effect encoded within the effector domain plotted for each of the 5 synthetic TFs, showing the dependence on the number of motifs for activators, and constant reduction effect that is independent of the number of motifs for repressors.

[0039] Figures 7A-7E: sURS variants function similarly in different cell types and growth conditions. (7A) Comparison of sURS variants expression levels in yeast. Growth conditions are as follows: glucose 2% refers to SD-Ura+2% glucose, 30°C, weak core promoter (mCore), no additives, 43 variants. Glycerol 2% refers to SD-Ura+2% glycerol, 30°C, weak core promoter (mCore), no additives, 43 variants. IM NaCl refers to SD- Ura+2% glucose, 30°C, weak core promoter (mCore), IM NaCl added, 43 variants. 39 DC refers to SD-Ura+2% glucose, 39°C, weak core promoter (mCore), no additives, 43 variants. Strong refers to SD-Ura+2% glucose, 30°C, strong core promoter (FEC-mCore), no additives, 20 variants. Error-bars were computed using standard-error analysis carried out on mean flow cytometry fluorescence measurements obtained from n=3 biological repeats. (7B-C) Comparison of the regulatory output for 32 sURS variants in (7B) CHO and (7C) HeLa cells to the regulatory output obtained for the 2% glucose growth condition in yeast. Red circles correspond to strongly repressed yeast variants that express strongly in mammalian cells. (7D) Pearson correlation coefficients obtained for each pair of conditions. (7E) Pearson correlation p-values obtained for each pair of conditions.

[0040] Figures 8A-C: A hybrid machine-learning and additive model (MLAM) predicts regulatory output in different cell types and growth conditions. (8A) Bar plot comparing MAM (blue) and MLAM (red) model performances on all 7 experimental datasets. Error bars were computed by bootstrapping which was carried out by sub-sampling 100 times 30 of the points in each dataset (16 points for the mammalian data sets) randomly and computing the Pearson correlation for each subset. (8B-C) Plots depicting the MLAM scores vs (8B) the 2% glucose and (8C) HeLa fluorescence measurements. Both data setsshow a linear relationship between the fluorescence expression data and the MLAM scores consistent with the high Pearson correlation coefficient. Error-bars were computed using standard-error analysis carried out on mean flow cytometry fluorescence measurements obtained from n=3 biological repeats.DETAILED DESCRIPTION OF THE INVENTION

[0041] The present invention, in some embodiments, provides nucleic acid molecules configured to constitutively express a sequence of interest in a eukaryotic cell comprising a transcription factor binding region comprising at least two sequences selected from those provided in SEQ ID NO: 1-41, a core promoter and a sequence of interest operatively linked to the transcription factor binding region and core promoter are provided. Methods of constitutively expressing a sequence of interest in a eukaryotic cell are also provided.

[0042] The invention is based, at least in part, on the surprising finding of regulatory sequences that will boost the expression of any core-promoter in any eukaryotic cell-type. To boost expression of any core promoter, we designed a simple regulatory code architecture composed of a minimal yeast promoter and a synthetic upstream regulatory region (sURS) that is populated with up to three TFBS motifs (from a list of 41) that are conserved across the eukaryotic lineage. To encode the sURS, we employed a mixed-base OL synthesis approach (Anavy et al., 2019, “Data storage in DNA with fewer synthesis cycles using composite DNA letters”, Nature Biotechnology, 37(10), 1229-1236, herein incorporated by reference in its entirety), in which specific mixtures of 2 nucleotides are added in the same synthesis step at selected positions. The mixed -base motifs are composed of two to sixteen sub-motifs (1 to 4 composite bases), which constitute a first low -complexity attempt at studying position weighted matrices (PWMs) in a synthetic regulatory context. Using bioinformatic and deep-learning analyses, we were able to classify active sub-motif function into either up- or down-regulating motifs. The 17 total active motifs are provided in bolded boxes in Table 2. The characterized motifs allowed us to construct a modular model for gene expression in eukaryotes, which we successfully used to predict the expression level boost of several tens of previously unseen sURSs, in yeast and mammalian cells. Finally, we showed that the grammar rules deduced from the motifs are encoded in the non-binding effector domains of several known TFs, thus providing a mechanism for the simplification of the regulatory code. Consequently, our motif-encoded sURS library yielded a panorganism regulatory architecture for boosting gene expression, which on the one handprovides further insight into the regulatory grammar, while on the other hand provides a validated protein over-expression design algorithm for yeast and mammalian cells.

[0043] By a first aspect, there is provided a nucleic acid molecule comprising a transcription factor binding region, a promoter and a sequence of interest.

[0044] The term "nucleic acid" is well known in the art. A "nucleic acid" as used herein will generally refer to a molecule (i.e., a strand) of DNA, RNA or a derivative or analog thereof, comprising a nucleobase. A nucleobase includes, for example, a naturally occurring purine or pyrimidine base found in DNA (e.g., an adenine "A," a guanine "G," a thymine "T" or a cytosine "C") or RNA (e.g., an A, a G, an uracil "U" or a C).

[0045] The terms “nucleic acid molecule” include but not limited to single-stranded RNA (ssRNA), double- stranded RNA (dsRNA), single- stranded DNA (ssDNA), double- stranded DNA (dsDNA), small RNA such as miRNA, siRNA and other short interfering nucleic acids, snoRNAs, snRNAs, tRNA, piRNA, tnRNA, small rRNA, hnRNA, circulating nucleic acids, fragments of genomic DNA or RNA, degraded nucleic acids, ribozymes, viral RNA or DNA, nucleic acids of infectious origin, amplification products, modified nucleic acids, plasmidical or organellar nucleic acids and artificial nucleic acids such as oligonucleotides. In some embodiments, the molecule is a DNA molecule. In some embodiments, the molecule is an isolated molecule. In some embodiments, the molecule is a purified molecule. In some embodiments, the molecule is a chimeric molecule. As used herein, the term “chimeric” refers to a molecule with parts from two different molecules. In some embodiments, the molecule is a not a naturally occurring molecule. In some embodiments, the molecule is an artificial molecule. In some embodiments, the transcription factor binding region is a synthetic region. In some embodiments, the transcription factor binding region is a not a naturally occurring sequence.

[0046] In some embodiments, the nucleic acid molecule is a vector. In some embodiments, the vector is an expression vector. In some embodiments, the expression vector is a eukaryotic expression vector. In some embodiments, the expression vector is a mammalian expression vector. In some embodiments, the nucleic acid molecule is for use in a method of the invention. In some embodiments, expression is constitutive expression.

[0047] As used herein, the term "expression" refers to the biosynthesis of a gene product, including the transcription and / or translation of said gene product. Thus, expression of a nucleic acid molecule may refer to transcription of the nucleic acid fragment (e.g., transcription resulting in mRNA or other functional RNA) and / or translation of RNA into aprecursor or mature protein (polypeptide). Expressing a gene / open reading frame within a cell is well known to one skilled in the art. It can be carried out by, among many methods, transfection, viral infection, or direct alteration of the cell’s genome. In some embodiments, the gene is in an expression vector such as plasmid or viral vector.

[0048] A vector nucleic acid sequence generally contains at least an origin of replication for propagation in a cell and optionally additional elements, such as a heterologous polynucleotide sequence, expression control element (e.g., a promoter, enhancer), selectable marker (e.g., antibiotic resistance), poly-Adenine sequence.

[0049] The vector may be a DNA plasmid delivered via non-viral methods or via viral methods. The viral vector may be a retroviral vector, a herpesviral vector, an adenoviral vector, an adeno-associated viral vector or a poxviral vector. The promoters may be active in mammalian cells. The promoter may be a viral promoter.

[0050] In some embodiments, the vector is introduced into the cell by standard methods including electroporation (e.g., as described in From et al., Proc. Natl. Acad. Sci. USA 82, 5824 (1985)), Heat shock, infection by viral vectors, high velocity ballistic penetration by small particles with the nucleic acid either within the matrix of small beads or particles, or on the surface (Klein et al., Nature 327. 70-73 (1987)), and / or the like.

[0051] The term "promoter" as used herein refers to a group of transcriptional control modules that are clustered around the initiation site for an RNA polymerase i.e., RNA polymerase II. Promoters are composed of discrete functional modules, each consisting of approximately 7-20 bp of DNA, and containing one or more recognition sites for transcriptional activator or repressor proteins.

[0052] As used herein, the term “core promoter” refers to a minimal DNA sequence that directs accurate initiation of transcription. In some embodiments, a core promoter is the minimal DNA sequence needed to recruit RNA polymerase to the molecule. In some embodiments, recruiting to the molecule is recruiting to the transcriptional start site (TSS). In some embodiments, the TSS is the TSS of the sequence of interest. In some embodiments, the RNA polymerase is RNA polymerase II. In some embodiments, a core promoter is the minimal DNA sequence needed to recruit the basal transcription machinery to the molecule. In some embodiments, the core promoter is constitutively active. In some embodiments, the core promoter does not comprise transcription factor binding sites. In some embodiments, a core promoter is devoid of transcription factor binding sites. In some embodiments, a core promoter is a minimal promoter. In some embodiments, a core promoter comprises orconsists of at most 125, 150, 200, 250, 300, 350, 400, 450, or 500 nucleotides. Each possibility represents a separate embodiment of the invention. In some embodiments, a core promoter comprises or consists of at most 125 nucleotides. In some embodiments, a core promoter comprises or consists of at most 150 nucleotides.

[0053] In some embodiments, nucleic acid sequences are transcribed by RNA polymerase II (RNAP II and Pol II). RNAP II is an enzyme found in eukaryotic cells. It catalyzes the transcription of DNA to synthesize precursors of mRNA and most snRNA and microRNA.

[0054] In some embodiments, mammalian expression vectors include, but are not limited to, pcDNA3, pcDNA3.1 (±), pGL3, pZeoSV2(±), pSecTag2, pDisplay, pEF / myc / cyto, pCMV / myc / cyto, pCR3.1, pSinRep5, DH26S, DHBB, pNMTl, pNMT41, pNMT81, which are available from Invitrogen, pCI which is available from Promega, pMbac, pPbac, pBK- RSV and pBK-CMV which are available from Strategene, pTRES which is available from Clontech, and their derivatives.

[0055] In some embodiments, expression vectors containing regulatory elements from eukaryotic viruses such as retroviruses are used by the present invention. SV40 vectors include pSVT7 and pMT2. In some embodiments, vectors derived from bovine papilloma virus include pBV-lMTHA, and vectors derived from Epstein Bar virus include pHEBO, and p2O5. Other exemplary vectors include pMSG, pAV009 / A+, pMTO10 / A+, pMAMneo- 5, baculovirus pDSVE, and any other vector allowing expression of proteins under the direction of the SV-40 early promoter, SV-40 later promoter, metallo thionein promoter, murine mammary tumor virus promoter, Rous sarcoma virus promoter, polyhedrin promoter, or other promoters shown effective for expression in eukaryotic cells.

[0056] In some embodiments, recombinant viral vectors, which offer advantages such as lateral infection and targeting specificity, are used for in vivo expression. In one embodiment, lateral infection is inherent in the life cycle of, for example, retrovirus and is the process by which a single infected cell produces many progeny virions that bud off and infect neighboring cells. In one embodiment, the result is that a large area becomes rapidly infected, most of which was not initially infected by the original viral particles. In one embodiment, viral vectors are produced that are unable to spread laterally. In one embodiment, this characteristic can be useful if the desired purpose is to introduce a specified gene into only a localized number of targeted cells.

[0057] Various methods can be used to introduce the expression vector of the present invention into cells. Such methods are generally described in Sambrook et al., MolecularCloning: A Laboratory Manual, Cold Springs Harbor Laboratory, New York (1989, 1992), in Ausubel et al., Current Protocols in Molecular Biology, John Wiley and Sons, Baltimore, Md. (1989), Chang et al., Somatic Gene Therapy, CRC Press, Ann Arbor, Mich. (1995), Vega et al., Gene Targeting, CRC Press, Ann Arbor Mich. (1995), Vectors: A Survey of Molecular Cloning Vectors and Their Uses, Butterworths, Boston Mass. (1988) and Gilboa et at. [Biotechniques 4 (6): 504-512, 1986] and include, for example, stable or transient transfection, lipofection, electroporation and infection with recombinant viral vectors. In addition, see U.S. Pat. Nos. 5,464,764 and 5,487,992 for positive-negative selection methods.

[0058] It will be appreciated that other than containing the necessary elements for the transcription and translation of the inserted coding sequence (encoding the polypeptide), the expression construct of the present invention can also include sequences engineered to optimize stability, production, purification, yield or activity of the expressed polypeptide.

[0059] In some embodiments, the transcription factor binding region comprises at least one sequence selected from those provided in SEQ ID NO: 1-41. In some embodiments, the transcription factor binding region comprises at least one sequence selected from those provided in SEQ ID NO: 1-64. In some embodiments, SEQ ID NO: 1-41 are SEQ ID NO: 1-64. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 1. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 2. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 3. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 4. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 5. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 6. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 7. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 8. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 9. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 10. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 11. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 12. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 13. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 14. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 15. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 16. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 17. Insome embodiments, the transcription factor binding region comprises SEQ ID NO: 18. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 19. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 20. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 21. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 22. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 23. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 24. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 25. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 26. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 27. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 28. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 29. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 30. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 31. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 32. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 33. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 34. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 35. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 36. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 37. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 38. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 39. In some embodiments, the transcription factor binding region comprises SEQ ID NO:40. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 41. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 42. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 43. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 44. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 45. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 46. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 47. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 48. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 49. In some embodiments, the transcription factor binding region comprises SEQ ID NO:50. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 51. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 52. Insome embodiments, the transcription factor binding region comprises SEQ ID NO: 53. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 54. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 55. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 56. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 57. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 58. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 59. In some embodiments, the transcription factor binding region comprises SEQ ID NO:60. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 61. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 62. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 63. In some embodiments, the transcription factor binding region comprises SEQ ID NO: 64.

[0060] In some embodiments, the transcription factor binding region comprises at least two sequences selected from those provided in SEQ ID NO: 1-41. In some embodiments, the transcription factor binding region comprises at least two sequences selected from those provided in SEQ ID NO: 1-64. In some embodiments, at least two is at least three. In some embodiments, at least two is at least four. In some embodiments, at least two is at least 5. In some embodiments, the at least two sequences are different sequences. In some embodiments, the transcription factor binding region comprises at least 2, 3, 4, 5, 6, 7, 8, 9 or 10 sequences selected from those provided in SEQ ID NO: 1-41. Each possibility represents a separate embodiment of the invention. In some embodiments, the transcription factor binding region comprises at least 2, 3, 4, 5, 6, 7, 8, 9 or 10 sequences selected from those provided in SEQ ID NO: 1-64. Each possibility represents a separate embodiment of the invention. In some embodiments, the transcription factor binding region comprises at least 2, 3, 4, 5, 6, 7, 8, 9 or 10 different sequences selected from those provided in SEQ ID NO: 1-41. Each possibility represents a separate embodiment of the invention. In some embodiments, the transcription factor binding region comprises at least 2, 3, 4, 5, 6, 7, 8, 9 or 10 different sequences selected from those provided in SEQ ID NO: 1-64. Each possibility represents a separate embodiment of the invention. In some embodiments, the sequences are different sequences. In some embodiments, the transcription factor binding region comprises a sequence selected from SEQ ID NO: 9-22. In some embodiments, the at least two sequences are selected from SEQ ID NO: 9-22. In some embodiments, at least one of the at least two sequences are selected from SEQ ID NO: 9-22. It will be understood by a skilledartisan that all possible combinations of motifs from SEQ ID NO: 1-64 are contemplated as separate embodiments of the invention.

[0061] In some embodiments, the transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 2-3, 7, 13, 17-18, 24, 30, 32-33, 36, 39-40 and 42-64. In some embodiments, the at least two sequences are selected from SEQ ID NO: 2-3, 7, 13, 17-18, 24, 30, 32-33, 36, 39-40 and 42-64. In some embodiments, the at least one of the at least two sequences are selected from SEQ ID NO: 2-3, 7, 13, 17-18, 24, 30, 32-33, 36, 39- 40 and 42-64.

[0062] In some embodiments, the transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 5, 6, 15, 20, 32, 36, 41. In some embodiments, at least 1 is at least two from SEQ ID NO: 5, 6, 15, 20, 32, 36, 41. In some embodiments, the transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 3, 9, 22, 28, 29, 30. In some embodiments, at least 1 is at least two from SEQ ID NO: 3, 9, 22, 28, 29, 30. In some embodiments, the transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 5, 6, 15, 20, 32, 36, 41, 3, 9, 22, 28, 29, 30. In some embodiments, at least 1 is at least two from SEQ ID NO: 5, 6, 15, 20, 32, 36, 41, 3, 9, 22, 28, 29, 30.

[0063] In some embodiments, the transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 32, 36, 42, 43, 44, 45, 46, 47, 48. In some embodiments, at least 1 is at least two from SEQ ID NO: 32, 36, 42, 43, 44, 45, 46, 47, 48. In some embodiments, the transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 3, 30, 39, 54, 55, 56, 57, 58. In some embodiments, at least 1 is at least two from SEQ ID NO: 3, 30, 39, 54, 55, 56, 57, 58. In some embodiments, the transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 32, 36, 42, 43, 44, 45, 46, 47, 48, 3, 30, 39, 54, 55, 56, 57, 58. In some embodiments, at least 1 is at least two from SEQ ID NO: 32, 36, 42, 43, 44, 45, 46, 47, 48, 3, 30, 39, 54, 55, 56, 57, 58. In some embodiments, the transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 42, 45, 55 and 57. In some embodiments, at least 1 is at least two from SEQ ID NO: 42, 45, 55 and 57.

[0064] In some embodiments, the transcription factor binding region comprises at least one activating sequence. In some embodiments, a sequence is a motif. In some embodiments, the transcription factor binding region comprises at least one activator motif. In some embodiments, an activating sequence is selected from SEQ ID NO: 2, 5, 6, 14, 15, 17, 20,24, 32, 36, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53. In some embodiments, an activating sequence is selected from SEQ ID NO: 2, 5, 6, 14, 15, 17, 20, 24, 32, 36, 41. In some embodiments, an activating sequence is selected from SEQ ID NO: 5, 6, 15, 20, 32, 36, 41. In some embodiments, an activating sequence is selected from SEQ ID NO: 5, 6, 15, 20, 32, 36, 41, 42, 43, 44, 45, 46, 47, 48. In some embodiments, an activating sequence is selected from SEQ ID NO: 32, 36, 42, 43, 44, 45, 46, 47, 48. It will be understood that binding regions comprising any combination of 1, 2, 3, 4, and so on of the recited sequences is also contemplated.

[0065] In some embodiments, the transcription factor binding region comprises at least one repressing sequence. In some embodiments, the transcription factor binding region comprises at least one repressor motif. In some embodiments, a repressing sequence is selected from SEQ ID NO: 3, 7, 9, 10, 13, 18, 19, 22, 26, 28, 29, 30, 31, 33, 38, 39, 40, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64. In some embodiments, a repressing sequence is selected from SEQ ID NO: 3, 7, 9, 10, 13, 18, 19, 22, 26, 28, 29, 30, 31, 33, 38, 39, 40. In some embodiments, a repressing sequence is selected from SEQ ID NO: 3, 7, 9, 13, 18, 22, 26, 28, 29, 30, 33, 39, 40, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64. In some embodiments, a repressing sequence is selected from SEQ ID NO: 3, 7, 9, 13, 18, 22, 26, 28, 29, 30, 33, 39, 40. In some embodiments, a repressing sequence is selected from SEQ ID NO: 3, 9, 13, 22,28, 29, 30, 33, 7, 40, 18, 39, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64. In some embodiments, a repressing sequence is selected from SEQ ID NO: 3, 9, 22, 28, 29, 30. In some embodiments, a repressing sequence is selected from SEQ ID NO: 3, 30, 39, 54, 55, 56, 57, 58. In some embodiments, a repressing sequence is selected from SEQ ID NO: 3, 9, 22, 28,29, 30, 39, 54, 55, 56, 57, 58. In some embodiments, the transcription factor binding region comprises at least one activating sequence and at least one repressing sequence. It will be understood that binding regions comprising any combination of 1, 2, 3, 4, and so on of the recited sequences is also contemplated.

[0066] In some embodiments, the transcription factor binding region comprises a repeat of at least one of the sequences. In some embodiments, repeats are identical. In some embodiments, each of the sequences is repeated. In some embodiments, a repeat is at least 2, 3, 4, 5, 6,7, 8 or 9 copies of the sequence. Each possibility represents a separate embodiment of the invention. In some embodiments, a repeat is at least two copies of the sequence. In some embodiments, the transcription factor binding region comprises at least 3 copies of the sequence. In some embodiments, the transcription factor binding regioncomprises at least 4 copies of the sequence. In some embodiments, the transcription factor binding region comprises at least 5 copies of the sequence.

[0067] In some embodiments, the sequences are separated by a spacer. In some embodiments, at least two sequences are separated by a spacer. In some embodiments, the at least two sequences are separated by a spacer. In some embodiments, all sequences are separated from another sequence by a spacer. In some embodiments, the region comprises a spacer at the 5’ end. In some embodiments, the region comprises a spacer at the 3’ end. In some embodiments, the spacer is a nucleic acid spacer. In some embodiments, the spacer comprises at least one nucleotide base. In some embodiments, the spacer comprises at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 bases. Each possibility represents a separate embodiment of the invention. In some embodiments, the spacer comprises about 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, or 20 bases. Each possibility represents a separate embodiment of the invention. In some embodiments, the spacer comprises at least 17 bases. In some embodiments, the spacer comprises about 17 bases.

[0068] In some embodiments, the core promoter is an mCorel promoter. In some embodiments, the mCorel promoter comprises ccttaattaacggaagcttataaaagagcactgttgggcgtgagtggaggcgccggaaaaaagcatcgaaaaaa (SEQ ID NO: 67). In some embodiments, the mCorel promoter consists of SEQ ID NO: 67. In some embodiments, the core promoter is an FEC-mCorel promoter. In some embodiments, the FEC-mCorel promoter comprises cctccttgaaactgaaattttagcatgtgattaattaacttgtaatattctaatcaagcttataaaagagcactgttgggcgtgagtggag gcgccggaaaaaagcatcgaaaaaa (SEQ ID NO: 68). In some embodiments, the FEC-mCorel promoter consists of SEQ ID NO: 68. In some embodiments, the core promoter is a CMV promoter. In some embodiments, the CMV promoter is the CMV minimal promoter. In some embodiments, the CMV minimal promoter comprises ctacttactgtggtaccggcgtgtacggtgggaggcctatataagcagagctcgtttagtgaaccgtcagatcgcctggagatttcg agctccccggggatccaccat (SEQ ID NO: 67). In some embodiments, the CMV minimal promoter consists of SEQ ID NO: 67.

[0069] In some embodiments, the transcription factor binding region is 5’ to the core promoter. In some embodiments, the transcription factor binding region is adjacent to the core promoter. In some embodiments, the transcription factor binding region is 3 ’ to the core promoter. In some embodiments, adjacent is directly adjacent. In some embodiments, the transcription factor binding region and core promoter are separated by a spacer. In someembodiments, the transcription factor binding region and core promoter are contiguous. In some embodiments, the transcription factor binding region, spacer and core promoter are contiguous.

[0070] In some embodiments, the core promoter is comprised within a promoter. In some embodiments, the promoter is an endogenous promoter. In some embodiments, the transcription factor binding region is added to the promoter. It will be understood that the transcription factor binding region of the invention can boost the expression of any promoter and render expression constitutive in eukaryotic cells. Thus, though only a core promoter is necessary the transcription factor binding region can be added to any promoter and will still be effective.

[0071] In some embodiments, the sequence of interest is an open reading frame. In some embodiments, the sequence of interest comprises a transcriptional start site. In some embodiments, the sequence of interest transcribes to an mRNA. In some embodiments, the sequence of interest is a cDNA. In some embodiments, the sequence of interest is a sequence of an mRNA. In some embodiments, the sequence of interest is a sequence of a regulatory RNA. In some embodiments, a regulatory RNA is a non-coding RNA. Examples of regulatory RNAs include, but are not limited to microRNAs (miRNAs), small interfering RNAs (siRNAs), short hairpin RNAs (shRNAs), and long non-coding RNAs (IncRNAs).

[0072] In some embodiments, the transcription factor binding region is an artificial sequence. In some embodiments, the transcription factor binding region is heterologous to the core promoter. In some embodiments, a sequence in the transcription factor binding region is heterologous to the core promoter. In some embodiments, the sequence is selected from SEQ ID NO: 1-41. In some embodiments, the sequence is selected from SEQ ID NO: 1-64. In some embodiments, the transcription factor binding region is from a different protein as the core promoter. In some embodiments, a sequence in the transcription factor binding region is from a different protein as the core promoter. In some embodiments, the transcription factor binding region is heterologous to the sequence of interest. In some embodiments, a sequence in the transcription factor binding region is heterologous to the sequence of interest. In some embodiments, the sequence is selected from SEQ ID NO: 1- 41. In some embodiments, the sequence is selected from SEQ ID NO: 1-64. In some embodiments, the transcription factor binding region is from a different protein as the sequence of interest. In some embodiments, a sequence in the transcription factor binding region is from a different protein as the sequence of interest. In some embodiments, thesequence of interest is heterologous to the core promoter. In some embodiments, the sequence of interest is from a different protein as the core promoter.

[0073] In some embodiments, the molecule is configured to express the sequence of interest. As used herein, the term “operably linked” is intended to mean that the nucleotide sequence of interest is linked to the regulatory element or elements in a manner that allows for expression of the nucleotide sequence (e.g., in an in vitro transcription / translation system or in a host cell when the vector is introduced into the host cell). In some embodiments, the molecule is configured to express the sequence of interest in a target cell. In some embodiments, the target cell is a eukaryotic cell. In some embodiments, the eukaryotic cell is a yeast cell. In some embodiments, the eukaryotic cell is a mammalian cell. In some embodiments, the mammal is a human. In some embodiments, the mammal is a hamster.

[0074] In some embodiments, the molecule is configured to produce tuned expression. In some embodiments, tuned expression is precise expression. In some embodiments, tuned expression is enhanced expression. In some embodiments, enhanced comprises expression which is increased over the expression produced by the core promoter. In some embodiments, enhanced comprises expression which is increased over the expression produced by the promoter to which the transcription factor binding region is linked. In some embodiments, linked is operably linked. In some embodiments, linked is in the same molecule. In some embodiments, linked is upstream of. In some embodiments, tuned expression is constitutive expression. In some embodiments, the molecule is configured to produce constitutive expression. In some embodiments, expression is constitutive expression. In some embodiments, constitutive expression is continuous expression. In some embodiments, constitutive expression is not cell type specific expression. In some embodiments, constitutive expression is expression in any mammalian cell. In some embodiments, constitutive expression comprises expression without the exogenous addition of a transcription factor. In some embodiments, constitutive expression comprises expression without the exogenous addition of a transcription activator. In some embodiments, constitutive expression comprise expression in a target cell without any exogenous additions. In some embodiments, addition is addition to the cell. In some embodiments, addition is addition to the media in which the cell is growing. In some embodiments, media is a culture. In some embodiments, the molecule is for use in producing expression in a target cell.

[0075] In some embodiments, the molecule further comprises an untranslated region (UTR). In some embodiments, the UTR is a 5’ UTR. In some embodiments, the UTR is a 3’ UTR.In some embodiments, the molecule comprises a 5’ and a 3’ UTR. In some embodiments, the 5’ UTR is between the core promoter and the sequence of interest. In some embodiments, the sequence of interest comprises the UTR. In some embodiments, the UTR is a spacer between the core promoter and the sequences of interest. In some embodiments, the 5’ UTR comprises the TSS.

[0076] By another aspect, there is provided a method of producing expression of a sequence of interest, the method comprising contacting a cell with a nucleic acid molecule of the invention, thereby producing expression of the sequence of interest.

[0077] By another aspect, there is provided a method of expressing a sequence of interest in a target cell, the method comprising contacting the target cell with the nucleic acid molecule of the invention, thereby expressing a sequence of interest in a target cell.

[0078] In some embodiments, the method is an ex vivo method. In some embodiments, the method is an in vitro method. In some embodiments, the method is in culture. In some embodiments, the target cell is in culture. In some embodiments, the target cell is a eukaryotic cell. In some embodiments, the contacting comprises expressing the nucleic acid molecule of the invention in the cell. In some embodiments, expressing comprises delivering the molecule into the cytoplasm of the cell.

[0079] By another aspect, there is provided a method of producing a nucleic acid molecule of the invention, the method comprising: a. providing a sequence of interest; b. providing a core promoter; c. providing a transcription factor binding region comprising at least one sequence selected from those provided in SEQ ID NO: 1 to 41; and d. generating a nucleic acid molecule comprising the provided sequence of interest operably linked to the transcription factor binding region and the core promoter; thereby producing a nucleic acid molecule.

[0080] By another aspect, there is provided a method of producing a nucleic acid molecule of the invention, the method comprising: a. providing a sequence of interest; b. providing a core promoter;c. providing a transcription factor binding region comprising at least one sequence selected from those provided in SEQ ID NO: 3, 30, 32, 36, 39, 42-48 and 54-58; and d. generating a nucleic acid molecule comprising the provided sequence of interest operably linked to the transcription factor binding region and the core promoter; thereby producing a nucleic acid molecule.

[0081] In some embodiments, the operably linking is such that the transcription factor binding region and the core promoter drive expression of the sequence of interest. In some embodiments, expression is in a target cell. In some embodiments, expression is tuned expression. In some embodiments, expression is constitutive expression. In some embodiments, the method further comprises testing expression of the generated nucleic acid molecule. In some embodiments, the method further comprises testing expression of the sequence of interest. In some embodiments, the testing is upon contacting a target cell with the generated molecule. In some embodiments, the method further comprises selecting a nucleic acid molecule that expresses in the target cell. In some embodiments, the method comprises testing that the expression is tuned. In some embodiments, the method comprises testing that the expression is constitutive. In some embodiments, the method comprises testing that the expression is increased as compared to the expression with the core promoter and without the transcription factor binding region. In some embodiments, a molecule with increased is selected. In some embodiments, the method further comprises selecting a molecule with increased expression.

[0082] In some embodiments, a transcription factor binding region such as is described hereinabove is provided. In some embodiments, a transcription factor binding region of the invention is provided. It will be understood that the transcription factor binding region can comprise any combination of the recited sequences and that each combination is considered a separate embodiment of the invention.

[0083] In some embodiments, increased comprises at least a 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 125, 150, 175, 200, 250, 300, 350, 400, 450 or 500% increase. Each possibility represents a separate embodiment of the invention. In some embodiments, the increase is a significant increase. In some embodiments, significant is statistically significant.

[0084] As used herein, the term "about" when combined with a value refers to plus and minus 10% of the reference value. For example, a length of about 1000 nanometers (nm) refers to a length of 1000 nm+- 100 nm.

[0085] It is noted that as used herein and in the appended claims, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Thus, for example, reference to "a polynucleotide" includes a plurality of such polynucleotides and reference to "the polypeptide" includes reference to one or more polypeptides and equivalents thereof known to those skilled in the art, and so forth. It is further noted that the claims may be drafted to exclude any optional element. As such, this statement is intended to serve as antecedent basis for use of such exclusive terminology as "solely," "only" and the like in connection with the recitation of claim elements, or use of a "negative" limitation.

[0086] In those instances where a convention analogous to "at least one of A, B, and C, etc." is used, in general such a construction is intended in the sense one having skill in the art would understand the convention (e.g., "a system having at least one of A, B, and C" would include but not be limited to systems that have A alone, B alone, C alone, A and B together, A and C together, B and C together, and / or A, B, and C together, etc.). It will be further understood by those within the art that virtually any disjunctive word and / or phrase presenting two or more alternative terms, whether in the description, claims, or drawings, should be understood to contemplate the possibilities of including one of the terms, either of the terms, or both terms. For example, the phrase "A or B" will be understood to include the possibilities of "A" or "B" or "A and B."

[0087] It is appreciated that certain features of the invention, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the invention, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable sub-combination. All combinations of the embodiments pertaining to the invention are specifically embraced by the present invention and are disclosed herein just as if each and every combination was individually and explicitly disclosed. In addition, all subcombinations of the various embodiments and elements thereof are also specifically embraced by the present invention and are disclosed herein just as if each and every such sub-combination was individually and explicitly disclosed herein.

[0088] As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” include plural referents, unless the context clearly dictates otherwise. The terms“a” (or “an”) as well as the terms “one or more” and “at least one” can be used interchangeably.

[0089] Furthermore, “and / or” is to be taken as specific disclosure of each of the two specified features or components with or without the other. Thus, the term “and / or” as used in a phrase such as “A and / or B” is intended to include A and B, A or B, A (alone), and B (alone). Likewise, the term “and / or” as used in a phrase such as “A, B, and / or C” is intended to include A, B, and C; A, B, or C; A or B; A or C; B or C; A and B; A and C; B and C; A (alone); B (alone); and C (alone).

[0090] Wherever embodiments are described with the language “comprising,” otherwise analogous embodiments described in terms of “consisting of’ and / or “consisting essentially of’ are included.

[0091] Additional objects, advantages, and novel features of the present invention will become apparent to one ordinarily skilled in the art upon examination of the following examples, which are not intended to be limiting. Additionally, each of the various embodiments and aspects of the present invention as delineated hereinabove and as claimed in the claims section below finds experimental support in the following examples.

[0092] Various embodiments and aspects of the present invention as delineated hereinabove and as claimed in the claims section below find experimental support in the following examples.EXAMPLES

[0093] Generally, the nomenclature used herein and the laboratory procedures utilized in the present invention include molecular, biochemical, microbiological and recombinant DNA techniques. Such techniques are thoroughly explained in the literature. See, for example, "Molecular Cloning: A laboratory Manual" Sambrook et al., (1989); "Current Protocols in Molecular Biology" Volumes I-III Ausubel, R. M., ed. (1994); Ausubel et al., "Current Protocols in Molecular Biology", John Wiley and Sons, Baltimore, Maryland (1989); Perbal, "A Practical Guide to Molecular Cloning", John Wiley & Sons, New York (1988); Watson et al., "Recombinant DNA", Scientific American Books, New York; Birren et al. (eds) "Genome Analysis: A Laboratory Manual Series", Vols. 1-4, Cold Spring Harbor Laboratory Press, New York (1998); methodologies as set forth in U.S. Pat. Nos. 4,666,828; 4,683,202; 4,801,531; 5,192,659 and 5,272,057; "Cell Biology: A Laboratory Handbook", Volumes I-III Cellis, J. E., ed. (1994); "Culture of Animal Cells - A Manual of Basic Technique" by Freshney, Wiley-Liss, N. Y. (1994), Third Edition; "Current Protocols in Immunology" Volumes I-III Coligan J. E., ed. (1994); Stites et al. (eds), "Basic and Clinical Immunology" (8th Edition), Appleton & Lange, Norwalk, CT (1994); Mishell and Shiigi (eds), "Strategies for Protein Purification and Characterization - A Laboratory Course Manual" CSHL Press (1996); all of which are incorporated by reference. Other general references are provided throughout this document.Materials and methods

[0094] Desert upstream regulating sequence design: We designed a lOlbp-long synthetic sequence, which lacks any known binding sites for yeast TFs, in silico in MATLAB (GTGCAATGGTCGCAATACCAGTCGACATACCGGGATAGAGCGCTTCAGTCTAC TTACCATCTAACTTACTGCCTCAGTATGGCTAACAGAGCGACTTGAGT; SEQ ID NO: 66). The aim was to design a sequence with a characterized basal level of expression, that maximizes the probability that only the designed TFs bind (given that their TFBS motifs are incorporated into the desert sequence). This so-called “desert” sURS was used as either the spacer sequence between TFBS motifs, or as a no-TFBS control. To generate this sequence, we used the IUPAC motifs deposited in "The Yeast Transcription Factor Specificity Compendium" (YetFaSCo) database (version 1.02) (yetfasco.ccbr.utoronto.ca / tour.php).

[0095] The desert sequence was generated by avoiding YetFaSCo TFBSs, with the following realistic limitations: 1. Short motifs (fewer than 5 nucleotides) were eliminated from the screen due to their high probability to appear in the designed sequence. 2. All lowercase motifs in the database were eliminated. According to the IUPAC code, lower-case nucleotides indicate lower frequency. 3. Motifs containing 5 or less upper-case characters were removed from the database, regardless of their length.

[0096] Motif selection and encoding into the OL: We selected motifs for the OL from a list created by an empirical assay designed to test protein activity in a broad swath of organisms (Wei, B. et al. “A protein activity assay to measure global transcription factor activity reveals determinants of chromatin accessibility”. Nat Biotechnol 36, 521-529 (2018), the contents of which are hereby incorporated by reference in their entirety). That study focused on developing a protein activity assay specifically for TFs in cell and tissue extracts. Based on the study’s data, we selected 41 enriched motifs from various organisms and tissue cells (e.g., yeast, fly, and mouse tissue cells), according to the following criteria: (i) different motiftypes: 8 organism- shared motifs, 5 mouse tissue-shared motifs, 14 unshared motifs (unique to an organism), and 14 unknown motifs (with unknown regulatory function); (ii) known / unknown regulatory function: 27 of the selected motifs have known regulatory function and thus were anticipated to be the control motifs for the 14 unknown motifs that we wanted to characterize.

[0097] Motifs were encoded via a mixed based approach using Twist Bioscience novel DNA synthesis technology. Additional complexity in the designed library, apart from the motif- shuffled variants, were the K (G / T) and M (A / C) mixed bases, incorporated in 20 motifs at pre-determined positions in the motifs according to their PWMs. K and M substitutions were based on the percentage calculation of the respective G / T and A / C occurrences in each position of the motif. Positions within the motif with dominant percentages were replaced by either K or M in the final design. 70% threshold was set to determine the K / M substitution, but the actual calculated threshold was higher at 73% (see Results section).

[0098] Design of motif-based sURS OL: The designed motif-based sURS OL includes the following DNA parts: a unique 15nt barcode sequence for each oligo in the OL and a lOlnt variable region, which is based on the designed desert sURS chassis. Specifically, barcodes were designed with a minimal 3nt edit distance, and restricted to maximum 3nt of sequentially repeated bases, with GC content between 35% to 65%. Both ends of each oligo were flanked with restriction sites (Spel and EagI) for library cloning purposes and forward / reverse primer sequences for the library’s PCR amplifications. Motifs from the set of 41 TFBS motifs (see Table 1) were placed at fixed positions along the desert sequence, depending on the number of motifs placed in the sequence (single / double / triple-motif oligos). The motif positions were at the 46th base for single-motif variants, at the 28th and 66th bases for double-motif variants, and at the 18th, 46th and 74th bases for triple-motif variants. Additionally, motif sequences were “shuffled” in the desert chassis (in accordance with the motif placement guidelines mentioned above), resulting in 70,644 different variants, including the desert variant lacking the 41 motifs. Additional complexity in the library, apart from the shuffled motifs, were the K (G / T) and M (A / C) mixed-bases, incorporated in 20 motifs, at pre-determined positions in the motifs, according to their position weight matrices (see Figure IB for further details). Each variant was represented by 2 barcodes, but 2435 variants, picked randomly from the pool, had an additional 20 barcodes (22 barcodes in total). In total, 189,990 barcoded oligos were designed.

[0099] Cloning the QL plasmid: A p416-based plasmid with a synthetic truncated core promoter, developed by Redden and Alper (Redden and Alper, 2015, “The development andcharacterization of synthetic minimal yeast promoters”, Nature Communications, 6(1), p. 7810, herein incorporated by reference in its entirety), was used to clone the OL upstream to the minimal corel (mCorel) promoter (SEQ ID NO: 67). The mCorel promoter has a weak activity and regulates the fluorescent yeast-enhanced Citrine (yeCitrine) gene. The plasmid has URA3 as a yeast selection marker and lacks the cen / ars sequence, enabling plasmid integration into the yeast genome at the URA3 locus. To prepare the plasmid for the OL cloning, a short, duplexed DNA (made by standard oligo hybridization), containing Spel and EagI restriction sites, was inserted into the plasmid upstream of the mCorel region. Initially, the plasmid was digested with Asci (NEB #R0558) and treated with CIP (NEB #M0290) to prevent self-ligation. Then, the mCorel plasmid and the short-duplexed DNA were ligated at 1:3 ratio by T4 DNA ligase (NEB #M0202). Ligated plasmid was transformed to E. coli TOPIO chemically competent bacterial cells (Thermo Fisher Scientific) and a positive clone was verified by Sanger sequencing. Next, double sequential digestion was performed on the OL plasmid: first 4pg of plasmid was digested with SpeLHF (NEB #R3133) for 1 hour at 37°C, heat inactivated at 80°C for 20 minutes. Then, EagLHF (NEB #R3505) was added to the reaction and the mixture was incubated again at 37°C for 1 hour, and heat inactivated at 65 °C for 20 minutes.

[0100] The library was PCR amplified using Q5 High-Fidelity DNA polymerase (NEB #M0491), lOng OL template in each reaction, 0.5pl of each primer (in concentration of lOOpM), Ipl dNTPs (lOmM). In total, we amplified in 40 50pl reactions, to conserve the pool’s complexity, which resulted in DNA concentration of 60.8ng / pl (overall 60pl), measured by Qubit. The PCR conditions were set using the manufacturer’s protocol for 14 cycles, with annealing temperature of 65°C, and 30 seconds for elongation time. Next, two sequential digestions were performed on the OL: first, 150ng of OL were digested with 0.5pl SpeLHF for 1 hour at 37°C, heat inactivated at 80°C for 20 minutes. Second, 0.5pl EagLHF was added to the reaction and the mixture was incubated again at 37°C for 1 hour, and heat inactivated at 65 °C for 20 minutes.

[0101] To clone the OL into the plasmid, both the plasmid and the OL amplicon pool were digested with SpeLHF and EagLHF, the reactions and ligated together with T4 DNA ligase at 22°C for 2 hours, then heat-inactivated at 65°C for 20 minutes. To conserve the OL complexity, 40 identical ligation reactions were performed, with a plasmid:insert molar ratio of 1:1. Generally, 150ng plasmid (5442bp) and 3.528ng OL (126bp) were ligated in a 20pl ligation reaction. All OL amplifications and other enzymatic reactions were purified usingPromega Wizard DNA Clean-Up System, according to manufacturer’s protocol (Promega #A7280).

[0102] Bacterial transformation and QL plasmid purification: 25pl E. cloni 10G CLASSIC Electrocompetent Cells (Lucigen #60117) were used to transform 2pl from the ligated OL- plasmid product. In total, 15 transformations were done. Cells were electroporated using constant settings (1600 Volt for 5 milliseconds). Immediately after electroporation, cells were recovered with 975pl recovery medium (supplemented by Lucigen) for 1 hour in 37°C, 250 rpm. After recovery, batches of 250pl cells were plated on LB-agar+ampicillin 140X20mm petri dishes and grown overnight at 37°C. For quality control and colony counting, recovered cells from each electroporation were plated on 2 LB+ampicillin 90X15mm petri dishes: first, 20pl recovered cells were diluted in 180pl LB. Then, 10 pl and 50 pl diluted cells were plated. The number of colonies on these plates was used to estimate the overall number of transformants in a single electroporation. The total number of transformant was estimated to be 5.6x106 cells.

[0103] Additionally, to preserve the OL as glycerol stocks, lOOpl recovered cells were cultured in 200ml LB+ ampicillin overnight at 37°C in 250rpm shaker. The next day, cells were pelleted in 4°C and re-suspended in ~6ml liquid LB, cells were divided into 1ml aliquots and frozen immediately with 600pl sterile 80% glycerol:20% water in -80°C freezer.

[0104] OL plasmids were extracted from the plated bacteria. First, cells were scraped from the plates with 10ml LB per plate, the resulting ~600ml mixture was centrifuged (10 minutes, 5000rpm at 4°C) and the plasmids were purified from the bacterial cells using Qiagen’s Plasmid Maxi kit (#12162), following the manufacturer’s manual. Each volume of 150ml mixture was cleaned separately, using one column from the kit. For the final elution, 400pl of TE buffer were used. All maxi prep eluates were combined into one tube, and concentration (~1800ng / pl) and purity were measured by Nanodrop.

[0105] Yeast integrations: Purified OL plasmid was linearized by Apal (NEB #R0114) restriction in the URA3 gene in the plasmid, for yeast genomic integration at the URA3 locus. Overall, 40 digestion reactions were performed, each had 8pg plasmid and Ipl Apal in a total of 20pl reaction. Reactions were incubated for 2 hours at 25 °C and heat- inactivated at 65 °C for 20 minutes.

[0106] Yeast W3O3-1A strain was cultured in liquid YPD medium (2% Glucose, 2% Bacto Peptone, 1% Yeast extract) at 30°C and 250 rpm overnight. The next morning, culture’sOD600 was determined by spectrophotometer, and the yeast cells were diluted to OD600=0.3 in 350 ml YPD media, and cultured in seven batches of 50 ml, each in a 500 ml flask. Cells grew until OD600 reached ~1, then 50ml cultured cells were harvested by centrifugation for 4 minutes at 4000rpm and washed with 20ml sterilized deionized water. 1ml of 0.1M lithium acetate (LiAc) was added, cells were centrifuged for 3 minutes at 3000rpm, and pellet was re-suspended in 0.1M LiAc. Subsequently, for each lOOpl cells, the cells were centrifuged again and 24pl sterilized water, 36pl IM LiAc, 5pl of boiled single stranded DNA (lOmg / ml salmon sperm ssDNA, Sigma-Aldrich #D7656), 20pl of Apa digested OL vector and 240pl 50% polyethylene glycol were added. Cells were incubated at 30°C for 30 minutes and then transferred to 42°C for 15 minutes. After the heat shock at 42°C, cells were spun down for 3 minutes at 3000rpm and resuspended in 200ml SD with 2% glucose, supplemented with tryptophan, leucine, and histidine amino acids, divided into four batches of 50ml, each in a 500ml flask. Lastly, cells were grown at 30°C for 60 hours at 30°C, 250rpm. After two days of growth, 15ml from each 50ml culture were taken for glycerol stock. Cells were spun down for 4 minutes, 4000rpm and resuspended in 1.5ml sterile 50% glycerol:50% water.

[0107] SQRT-seq and QL genomic amplification: Yeast cells with integrated OL were grown in liquid media (SD+2% glucose with supplement of amino acids, without uracil) for 60 hours, at 30°C and 250rpm. On the third day, cultures’ OD were measured, 10- 15ml batches of the grown cells were centrifuged (4 minutes, 4000rpm) and resuspended in 50ml fresh SD+2% glucose, supplemented with tryptophan, leucine, and histidine amino acids. A day later, 10ml cultured cells were spun down and washed with 20ml sterile deionized water. Cells were resuspended in fresh 50ml PBSF (50ml PBSX1 supplemented with 50pl BSA X100, 20mg / ml, NEB #B9200S). Resuspended cells were kept on ice until the FACS run. As controls, WT W3O3-1A, weak mCorel promoter (SEQ ID NO: 67), and strong FEC- mCorel promoter (SEQ ID NO: 68), were analyzed as well. Integrated yeast cells were analyzed and sorted into 4 yeCitrine fluorescence bins using a BD FACSAria-IIIu cell sorter. Bin numbers represent sequential ranges of fluorescence, with binl for cells with the lowest fluorescence and bin4 for cells with the highest fluorescence. In each bin, at least 10M cells were collected and kept on ice until the subsequent step. Next, binned cells were harvested in 4°C for 10 minutes at 5000rpm. Supernatant was discarded carefully by pipet aid, leaving ~lml of PBSF (to avoid suction of cells). Binned cells were re-grown overnight at 30°C, 250rpm in 25ml rich media (SD+2% glucose) and supplemented with tryptophan, leucine, and histidine amino acids. The following morning, OD absorbance (expected 4-10 OD) wasdetermined, binned cells were centrifuged for 4 minutes at 5000rpm, and the supernatant was discarded carefully. Next, for each of the four binned cultures, genomic DNA was isolated from the cells in the “Bust n’ Grab” method, as follows: for each 1.5ml of centrifuged medium, 200pl of lysis buffer (1% SDS, 2% Triton X-100, lOOmM NaCl, lOrnM Tris-HCl pH=8, and ImM EDTA pH=8) was added to resuspend the cells. Cells were frozen in -80°C for at least 2 minutes, to ensure complete freeze. Then, the cells were immediately heated at 95°C for 1 minute to quickly thaw, followed by vigorous vortex for 30 seconds. 200pl chloroform was added to each tube, followed by another vigorous vortex for 2 minutes. Tubes were centrifuged at 17,000g for 6 minutes at room temperature. The aqueous phase (containing the genomic DNA) was transferred to a tube with 400pl ice-cold 100% ethanol. DNA was precipitated for 5 minutes at room temperature and then centrifuged 10 minutes at 17,000g at room temperature. Supernatant was removed and DNA pellet was washed with 500pl 70% ethanol (diluted with ultra-pure water, UPW). Tubes were centrifuged at 17,000g for another 10 minutes at room temperature and supernatant was discarded. Ethanol was evaporated by heating the tubes for 5 minutes at 60°C. DNA was resuspended in 50pl TE (Biolab, lOmM Tris, ImM EDTA pH=8) or UPW and incubated at 65°C for 10 minutes. Samples’ genomic DNA concentration was read using Nanodrop and stored at -20°C. The OL genomic regions for each bin were amplified using QuantStudio Real-Time PCR System (Thermo Fisher) for 23 cycles in a 50pl reaction, with additions of: Ipg genomic DNA, 0.5pl Q5 High-Fidelity DNA polymerase, 1.5pl 100% DMSO, Ipl lOmM dNTPs, 10 pl 5X Q5 reaction buffer, 2.5pl 20X Evagreen dye (Biotium #31000), and 2.5pl lOmM from each forward and reverse primers. Overall, nine 50pl PCR reactions were run. PCR reactions were cleaned using the Wizard® SV Gel and PCR Clean-Up system kit (Promega #A7280), following the manufacturer’s manual. DNA concentrations were determined by Qubit Fluorometer, using the dsDNA HS Assay kit (Invitrogen™ #Q32851). In addition, the pool’s length and samples’ quality control were assessed by TapeStation (Agilent). Finally, the OL was sent to three NGS runs: two Nova-seq and one Next-seq run, generating a total of 1500M paired-end reads.

[0108] NGS data processing and read normalization: First, overlapped paired-end NGS reads were assembled using PEAR (PEAR - Paired-End reAd mergeR) code, version 0.9.8. Second, for ease of use and to fit into memory, the fastq files were split into IM-read files, using the SeqKit tool. All split files were subsequently analyzed in MATLAB. The reads were counted according to their specific bin barcodes and variant barcodes. Reads in each bin were normalized as follows: 1) For each bin, the total number of reads was calculated.2) For each variant in a specific bin, we divided its read count by the total number of reads of that bin. 3) We calculated the percentage area of each bin (“%Bin”) in the histogram derived from the sorting experiment. 4) We multiplied the outcome in step 2 by the corresponding “%Bin”, resulting in adjusted reads per variant per bin. 5) For each variant, we summed the adjusted reads in all four bins. 6) Finally, for each variant, we divided the adjusted reads per bin from step 4 by the sum calculated in step 5, resulting in normalized reads for each bin.

[0109] After normalizing the reads, the mean FL for each variant in the library was calculated as follows: Mean FL = MeanBinl*x+MeanBin2*y+MeanBin3*z+MeanBin4*w, where: MeanBinl = 607, MeanBin2 = 1364, MeanBin3 = 2596, and MeanBin4 = 7541. MeanBini are the mean FLs of bins i =1, 2, 3, 4 used during the cell sorting, and x, y, z, and w correspond to the normalized read counts in each bin for the specific variant.

[0110] Minimum hypergeometric analysis: Minimum hypergeometric analysis (mHG) is a statistical analysis methodology with a specific application as a computational method to discover motifs in ranked lists of DNA sequences. The mHG procedure notes whether a motif is present in each of the sequences and labels these as ‘ 1’, sequences without the motif are labeled as ‘0’ . In the case of motif enrichment at the top (or bottom) of the list, the list is divided into 2 subsets: 1) the target set: the part of the list in which the motif is highly enriched, and 2) the background set: the remainder of the list, with lower occurrences of the motif. For each motif, the mHG determines the optimal partition cutoff in the list to separate between enriched and non-enriched sequences. It calculates the mHG p-value of the motif enrichment based on hypergeometric probability of the order of T's in the ranked list. Additionally, several other outputs are available: the motif s IUPAC representation, its position- weight matrix, and the optimal partition cutoff.

[0111] Herein, we used a python code implementation of mHG to calculate the mHG p- value for each of the 41 motifs and the desert motif individually, both for the activating and repressing effects. For each motif, we sorted all 2435 variants from highest to lowest mean FL (to find activating enriched motifs), and from lowest to highest (for repressing enriched motifs).

[0112] For the sub-motif analysis, we first “expanded” the 2098 variants containing motifs with mixed bases (of the 2435 variants with 22 barcodes each) into all possible variant submotifs (containing A / T / G / C in place of mixed bases). This resulted in the original 21 nonmixed base motifs and an additional 80 sub-motifs, derived from the mixed-base motifs.Afterwards, for each sub-motif, we applied the mHG analysis. We present mean fluorescence levels for each of the sub-motifs in Figures 4B-4F.

[0113] Machine-learning model to predict mean fluorescent expression levels: We developed a CNN to predict the mean fluorescent expression level (Fig. 3A). The input to the network is a one -hot encoded DNA sequence. We tested two inputs: the 101 nt-long variant sequence, or the 116 nt-long variant and barcode sequence. The input goes through multiple ID-convolutional kernels, a global max-pooling layer, a fully connected layer, and finally an output layer. The output layer is either a single neuron with linear activation or four neurons with softmax activation, depending on whether we aimed to predict the mean fluorescent expression level or the distribution of normalized read counts across the four bins. We trained the network with MSE loss function for mean fluorescent expression level prediction and an average-weighted cross-entropy loss function to predict the normalized bins distribution, ADAM optimizer, and with or without sample weights which we defined as the total read count for each variant. The hyper-parameters were: 1,024 kernels of width 6 with ReLU activation, 16 neurons with ReLU activation, 5 training epochs, and batch size 32. We did not test other hyper-parameters as results were already good with our initial choice. We developed the network using Keras and TensorFlow python packages.

[0114] We finally selected the architecture of the ADM, AMM, and MBO models with the following attributes: predicting a mean FL value for each variant, utilizing sample weights during the training process, and omitting the 15 nt-long barcode region from the input sequence. To improve prediction performance and robustness of the AMM and MBO models, we used a random ensemble initialization technique. We trained 100 models on the datasets differing in the initial weights and training batches and used the average over their predictions as the final prediction. To further improve prediction performance and enable prediction over sequences longer than those in the training set, we added a 41 -long motif count vector as input that is concatenated following the max -pooling layer, and uniformpadded the training input to the maximum prediction length.

[0115] Validation experiments: Two sets of variants were designed and synthesized by IDT as gBlocks gene fragments. The first set of variants includes 11 variants from the sURS OL, differing by their FL MBO-predicted output (2700-3400), chosen after analysis of the NGS and deep-learning model development. The second set of variants consists of 32 variants, based on the significant motifs found via the mHG analysis. For this set, we designed a longer desert sequence in silico using MATLAB (186 bp in length) to test longer variantsthan those in the original OL (101 bp). Notably, for the longer variants, we had no ML predictions because of the change in length.

[0116] For validation in yeast cells, all variants were cloned individually upstream to the weak mCorel promoter, in the same manner that the sURS OL was cloned. Vectors were transformed into bacterial E. coli TOPIO competent cells, and positive colonies were verified after colony PCR by Sanger sequencing. Verified plasmids were purified from bacteria and digested with Apal, for subsequent integration to the yeast genome in URA3 locus, as described previously. After integration, 2-8 colonies were picked and tested as single variants using flow cytometry. Colonies were grown in liquid SD medium with 2% glucose, supplemented with leucine, histidine, and tryptophan amino acids. After overnight growth, cells were diluted and grown for 12 hours, until reaching OD600 of 0.7-2. Then, cells were washed and resuspended in PBSF. Lastly, cells were analyzed in the MACSQuant VYB flow cytometer (Miltenyi Biotec).

[0117] In addition, part of the second set of variants was cloned upstream to the strong FEC- mCorel promoter. The variants that were used for this validation contain either repressing motifs to down-regulate the promoter activity, or strong activating motifs, in an attempt to enhance the high activity of the promoter even more.

[0118] For the validation in a mammalian model (CHO cells), the first set of 11 variants were cloned upstream to a minimal CMV promoter and the mCherry target gene in pTRETightBLRY-0 vector, after removal of the TetO operator. pTRETightBLRY-0 was a gift from Phil Sharp (Addgene #31463). The TetO operator was removed by plasmid digestion with Xhol and Apal restriction enzymes, and the backbone was cleaned from gel using the Wizard DNA clean-Up System kit (Promega). Two complementary oligos with phosphorylated Xhol and Apal ends encoding the minimal CMV promoter were annealed. The minimal promoter sequence was taken from the pTRE-Tight-BLAcGFPl vector from Clonetech (#631066). The annealed product was ligated using T4 ligase to the Xhol- and ApaLrestricted backbone. Afterwards, to clone the variants upstream the pCMVmin promoter, the backbone was cut with Xhol and KpnLHF, cleaned by Promega Wizard kit and ligated with Xhol and KpnLdigested variants. The ligated plasmids were transformed to E. coli TOPIO cells, and positive clones were verified by sequencing. A positive control with a full CMV promoter was created by amplifying the promoter from the commercially available pTwist-CMV-BetaGlobin plasmid, also adding Xhol and Kpnl sites at the ends of the promoter. The amplified promoter was digested with Xhol and KpnLHF and ligated with the backbone, upstream to the mCherry gene. Additionally, the CMV enhancer region wascloned upstream to the pCMVmin promoter, as a control to restore the promoter’s activity in the CHO cells. All verification plasmids were Sanger sequenced.

[0119] For mammalian transfections, CHO cells were grown for 24h, 48h, and 72h posttransfection measurements of mCherry. 1.2x104 CHO cells were seeded in a 96-wells plate in lOOpl of F12 medium (Biological Industries #01-095-lA), supplemented with 10% FBS (Biological Industries #04-007-lA) and 5% penicillin / streptomycin solution (Biological Industries #03-031-1). After 24h, cloned plasmids were diluted to a 50ng / pl concentration, and lOOng per reaction of cloned plasmid DNA was transfected by mixing the DNA to final volume of 5pl OptiMEM buffer (Thermo Fisher #31985062). Img / ml stock of linear polyethylenimine (PEI, 25kDa, Polysciences #23966-1) was used, by mixing 0.67pl PEI per reaction to final volume of 2.5pl OptiMEM buffer. Then, PEI-OptiMEM mixture was added to DNA-OptiMEM mixture and incubated at room temperature for 15min, before transfecting 7.5pl of DNA-PEI-OptiMEM solution per reaction to the appropriate well. Each cloned plasmid was transfected in three replicates. A day after transfection, the medium was changed with a fresh medium and the DNA-PEI mixture was removed. Cells were grown for 24h, 48h, and 72h post-transfection and mCherry expression was measured in a flow cytometer (MACSQuant). Until their measurements, cells were incubated at 37°C in 5% CO2 incubator throughout the experiment, until the fluorescence measurements.

[0120] Cloning LexA chimeras: We fused the effector domains of 5 different TFs to the bacterial TF LexA: the activators Gal4, Hatl and the synthetic VPR (VP64-p65-Rta), and the repressors Migl and human Mxil. Cloning was performed in a pUG34-based backbone, with HIS 3 auxotrophic marker and CEN / ARS region for maintenance and self-replication in yeast. The fusions contained the SV40 nuclear localization signal (NLS). The fused proteins were expressed under the regulation of Gall-10 promoter, induced by galactose. The plasmids were transformed to yeast cells using the standard transformation protocol described above in the yeast integration method, with the following changes: BY4741 yeast cells were grown to mid-log in YPD medium, and circular plasmid (~2pg) was used to transform the yeast cells. In addition, yeasts were plated on SD with 2% glucose, supplemented with uracil and leucine, without histidine, and incubated for 2-3 days at 30°C. A colony from each fusion clone was isolated as a master strain for subsequent BRS OL integration.

[0121] Design and cloning of LexA bacterial regulatory sequence OL: The design of the LexA bacterial regulatory sequence (BRS) OL includes the lOlbp desert sequence as a chassis, and the placement of 1, 2, 3 or 4 LexA binding motifs throughout the desert region,at various positions. In addition, each oligo in the OL contains a unique 15nt barcode sequence. The positions of the LexA motifs changes in every variant, while keeping at least a 7nt-distance gap between the motifs from the end of one motif to the start of the next. For this BRS OL design, we used 9 LexA operators as DNA motifs, with 2nt edits in the middle of the operator (AT / CT / AG / CG / AK / CK / MT / MG / MK). The LexA motif is described as ‘TACTGTATGATCATACAGTA’ (SEQ ID NO: 69), where underlined bases are the edited bases. Each variant was represented by 2, 4, 6, 8, or 18 different barcodes, generating a total of 67,572 variants in the BRS OL. As internal controls for the library, we included 14 additional variants (represented by 50 barcodes each): desert sequence, 10 random sURSs (generated in silico by MATLAB, with 0.25 probability for each base), 1 negative control with 4 Migl binding sites (a repressing variant), and 2 positive controls with either 4 Gal4 or 4 Gcn4 binding sites (activating variants). In total, the BRS OL contains 68,273 barcoded variants.

[0122] BRS OL was inserted into 2 plasmids, which are based on the previous two p416 plasmids described above, one with a weak promoter (mCorel) and the other with a strong promoter (FEC-mCorel), to screen potential repressing and activating BRS variants. We cloned 2 homologous regions for URA3 (200bp and 221bp long) at the ends of the yeCitrine reporter cassette in the plasmids, flanking with Nhel restriction sites, for subsequent Nhel digestion and integration at the URA3 genomic locus in the yeast. In addition, we inserted 2 restriction sites (Spel and SphI), using a short-duplexed DNA fragment, in an AscLdigested plasmid (as was described above), to prepare the plasmids for the OL cloning. The construction of the plasmids was done by Gibson assembly and the final plasmids were verified by Sanger sequencing.

[0123] The amplification, cloning and NGS verifications (not shown) for the BRS OL were performed similarly to the former sURS OL, described in detail above, with minor alterations. In short, the amplified BRS OL was digested with SpeLHF and SphLHF restriction enzymes and cleaned by Promega Wizard kit. The digested OL was ligated with SpeLHF and SphLHF digested plasmids, using T4 ligase.

[0124] For the BRS integrations into yeast, each LexA-effector domain chimera strain was cultured until mid-log OD in SD with 2% glucose (supplemented with uracil and leucine). BRS genomic integrations were performed as described above, and post-integrations, integrated yeasts were grown in liquid SD with 2% glucose (with leucine addition) for 3 days at 30°C, 250 rpm. On the third day, the medium was substituted with fresh SD with 2% raffinose and 2% galactose to induce LexA-effector chimera expression. Cells were grownfor 2 days until sorting. On sorting day, OD600 values were measured, and yeast cultures were diluted to OD600 of ~1 in PBSF solution for the FACS experiment. All subsequent procedures were done as described above.

[0125] Mammalian cell experiments: CH0 / K1 cells were grown in F12 medium, supplemented with 10% FBS and 1% penicillin / streptomycin solutions, and incubated at 37C and 5% CO2. Cells were seeded into 96-well plates at a seeding density of 1.2xlOA4 cells in a lOOul of media per well. 24 hours post-seeding, media was changed into lOOul fresh media. Per reaction, 0.65ul of linear PEI (Mw 25kDA) in a final volume of 2.5ul OptiMEM buffer was added to lOOng of plasmid DNA in a final volume of 5ul OptiMEM buffer. This yielded a DNA-PEI mixture, that was left to incubate for 15minutes, before being added to each well. 24 hours post-transfection, media was changed again into lOOul fresh media, removing media containing PEI. 72 hours post-transfection, media was removed, and cells were washed with 50ul of IxPBS per well. PBS was removed and 25ul of 0.05% trypsin solution per well was added and the plate was incubated for 3min at 37C and 5% CO2. Afterwards, cells were resuspended using lOOul of fresh media and measured using flowcytometry (MacsQuant).

[0126] Statistics and Reproducibility: Oligo library was sequenced three times. First, directly by amplifying the purchased Twist DNA variants. Second, after cloning into plasmids via E.cloni. Third, after integration into yeast cells and sorting via flow cytometry. Yeast Integrated oligo library experimental sorting and extraction of promoter library was carried out once. Libraries for NGS sequencing amplified from extracted yeast integrated promoters were assembled twice on separate dates and sequenced separately.

[0127] Each validation variant characterization experiment was carried out via flow cytometry, repeated in technical duplicates and carried out in biological triplicates over three days. This was done for all 43 validation variants in yeast, and the 32 validation variants in CHO and HeLa cells.

[0128] No statistical method was used to predetermine sample size. No data were excluded from the analyses. The experiments were not randomized. The Investigators were not blinded to allocation during experiments and outcome assessment.

[0129] Further data and information can be found in Vaknin et al, “A universal system for boosting gene expression in eukaryotic cell-lines”, Nat Commun. 2024 Mar 16; 15(1):2394 as well as the SRA website under BioProject accession number PRJNA1061345 (ncbi.nlm.nih.gov / sra / ?term=PRJNA1061345), BioSample accessions: SAMN39265674,SAMN39265675, SAMN39265676, SAMN39265677, SAMN39265678, SAMN39265679, SAMN39265680, SAMN39265681, the contents of which are all hereby incorporated by reference in their entirety.Example 1: Motif-based design of a synthetic URS oligo library

[0130] We constructed a motif-based sURS OL for yeast cells by encoding 41 DNA regulatory motifs mined from the following organisms: S. cerevisiae, S. pombe, D. melanogaster S2 cells, different murine tissues, and murine ES cells (Fig. 1A, Table 1). We reasoned that conserved motifs are more likely to function similarly in most organisms and can thus form the regulatory backbone for a universal boosting system. To increase the likelihood for successfully identifying universal “boosting” motifs, we encoded the sURS library using a mixed base OL synthesis approach, in which specific mixtures of 2 nucleotides are added in the same synthesis step at selected positions. The mixed base motifs are composed of two to sixteen sub-motifs (1 to 4 composite bases), which constitute a first low-complexity attempt at studying position-weighted matrices (PWMs) in a synthetic regulatory context. 17 of the motifs were previously classified as repressing or activating motifs, and 10 motifs were previously found to have dual regulatory functionality (i.e., they can act as either repressors or activators, depending on the promoter and the other TFs with which they interact). Finally, 14 motifs were recently discovered in a large binding screen (Wei, et al., 2018, “A protein activity assay to measure global transcription factor activity reveals determinants of chromatin accessibility”, Nature Biotechnology, 36(6), 521-529, herein incorporated by reference in its entirety) and their regulatory effect in yeast is unknown.

[0131] Table 1: 41 DNA regulatory motifs

[0132] To encode a first approximation to the position weighted probability matrixes (PWMs) of our selected motifs, we incorporated K (G / T) and M (A / C) mixed bases in 20 of the motifs, at positions where the two nucleotides represented by either K or M were likely to appear (>73%), while using the most probable base for the remaining positions (Fig. IB). Overall, 189,990 barcoded oligos were designed according to the following characteristics: multiple barcodes for each variant, number of motifs in the variants (0, 1, 2, or 3 motifs), motif s K / M substitutions, and the order of motifs in each variant. For 2435 of the oligos, picked randomly, we generated 22 barcodes, while for the rest of the oligos we generated 2 barcodes. To ensure the same background expression level for all variants, we designed a “desert” chassis in silico to exclude any known yeast S. cerevisiae TFBS motifs. To generatethis desert sequence, we computationally excluded all known consensus yeast motifs as specified in the YeTFaSCo database. In all the variants, the motifs were embedded within the desert sequence, at fixed positions, with 17 bp spacing between motifs. Each variant in the OL consists of a variable sURS, which regulates the mCorel minimal promoter activity driving yeCitrine expression (Fig. 1C). The OL (Twist Bioscience) was amplified using PCR, then cloned into a plasmid and amplified in E. cloni 10G electrocompetent cells (see Materials and methods). Mid-cloning deep sequencing was performed to evaluate the OL coverage. Purified plasmids were linearized and integrated into the yeast URA3 genomic locus (Fig. ID). After integration, the cells were grown in selective media for 3 days and sorted into 4 bins by fluorescence-activated cell sorting (FACS), according to yeCitrine fluorescence (see Materials and methods). Finally, genomic DNA was extracted from each bin, the OL region was amplified and bin-barcoded by PCR and sent to next-generation sequencing (NGS) (Illumina Nova-seq). Overall, we identified -400M valid reads, which allowed us to retrieve 147,731 variants (of 189,990 designed) from the OL.

[0133] Upon initial analysis of the NGS data, we discovered three separate groups of retrieved variants, distributed across a wide range of mean fluorescent expression levels (Expression (A.U.) (-500-7500)). This indicates that the OL encodes a broad range of regulatory activity of yeCitrine expression (Fig. IE). The groups were defined as follows: Group 1 contains 107,868 variants with a wide range of mean fluorescent expression level (as shown in Fig. IE) with at least 40 total reads (i.e., variants with total number of reads greater than the minimum observed in Fig. IF). Group 2 corresponds to the 28,431 variants whose reads are above 40 and appear predominantly (i.e., >90%) in the upper bin, and thus correspond to “saturated” or under-sampled mean fluorescent expression level range. Group 3 corresponds to all variants with <40 reads, this group contains 13,108 variants in total. Group 1 demonstrates a Gaussian-like read distribution consistent with a well-sampled library, while Group 3 has an inverse exponential behavior resulting from a group of undersampled variants. Group 1 also contains single-bin variants (bins 1-3), apparent as protrusions at mean fluorescent expression level values of 600, 1300, and 2600, likely indicating fluorescence from a small number of cells (despite having >40 reads). We proceeded with only Group 1 variants in subsequent analysis.Example 2: Variants manifest a broad range of regulatory behavior

[0134] We analyzed the unsaturated variants with high read count (Group 1, 107,868 variants in total). First, for each motif (41 TFBS motifs and the desert motif) we grouped allvariants containing that motif at least once and calculated the mean fluorescent expression level distribution for each group. The results are shown as individual boxplots in Figure 2A, arranged by decreasing order of the median computed for each distribution (red lines), where the green line delineates the median computed for the desert motif’s mean fluorescent expression level distribution. Figure 2A shows a clear dependence of median values on motifs, which vary from <2000 for the lowest median to -3000 for the highest median. When comparing the group of variants containing each motif to the group of variants containing motifs that are ranked by the median at least 5 motifs away, we observed significant differences for all motif-containing groups except for motifs 16-21 (out of 42) in decreasing order (p-value < 0.05, two-sided Wilcoxon rank-sum test, Benjamini-Hochberg-corrected with FDR = 0.1). Interestingly, the median of the desert motif distribution scored at the higher end of the median scores, indicating that the majority of motif distributions generate a reduced median expression score with respect to the desert motif reference. The desert variant itself displays lower expression level, as expected (black line in Fig. 2A). Therefore, motif mean fluorescent expression level distributions positioned at the top and bottom of the list are more likely to be enriched with up-regulating and down-regulating binding sites, respectively.

[0135] We next calculated the Pearson correlation between the medians of mean fluorescence expression level distributions and the number of variants in each distribution (Fig. 2B). We found a positive correlation of -0.3 between the median values and the number of variants within a certain motif distribution. This correlation suggests that motif distributions that are characterized by low median value and low number of variants are likely indicative of a repressive or inhibitory motif. The reduced number of variants in these low median expression level distributions is, therefore, likely due to strongly repressed variants that do not pass the SORT-seq assay’s minimum fluorescence threshold. Conversely, highly expressing variants were collected by all bins, thus naturally leading to motif mean expression level distributions that are characterized by both a higher median and a larger number of variants that are, in turn, likely indicative of an upregulating motif.

[0136] We next grouped all identical variant pairs (i.e., identical oligos with two different barcodes) and computed the correlation between the read- 4-vector (corresponding to the number of reads in each of the four bins) obtained for each variant. We plot the correlation values obtained as a heatmap on an X- Y scatter plot, where the X and Y axes correspond to the mean fluorescent expression level of barcodes 1 and 2, respectively (Fig. 2C). The scatter plot shows variants exhibiting no correlation, complete anti-correlation, and also fullycorrelated mean fluorescent expression levels (see color bar for Pearson correlations from - 1 to +1). A closer examination of the plot reveals four identifiable correlation regimes: predominantly correlated pairs at low expression (Expression A. U. <2000) and high expression (Expression (A.U.) >5000), predominantly uncorrelated pairs in the intermediate expression range (2000<Expression (A.U.) <5000), and an anticorrelated set of pairs. The anticorrelated pairs are likely due to various sources of noise, which include low read count, low number of viable yeast cells that were sorted, variation in expression. The high correlation regions are consistent with up-regulating sURSs (high-fluorescence regime - Fig. 2C-top right) and down-regulating sURSs (low-fluorescence regime - Fig. 2C-bottom left), where the regulatory effect of the motifs strongly affects the expression. Finally, in the intermediate fluorescence regime, where the predominantly uncorrelated pairs are found (Fig. 2D -middle, between mean Expression (A.U.) of 1850 to 4150), the absence of strong regulatory effect may be a major cause for lack of correlation. For these weakly -regulating variants, the small variation in sURS sequence due to the barcodes (for example) may have a dominant role in the expression output.Example 3: Deep learning reveals that only variants with multiple barcodes generate statistically reliable data

[0137] To provide a more quantitative analysis of the data, we trained a convolutional neural network (CNN) to predict the gene expression level given the DNA sequence (Fig. 3A). First, we aimed to isolate the subset of variants in the OL for which robust and statistically reliable mean fluorescent expression levels were measured. Specifically, our goal was to identify the training set which yields the best regulatory predictive performance on a withheld test set of sURS variants from the OL. We tested 8 models, which differed by combinations of three characteristics: with / without a 15nt leader barcode sequence, variants characterized by a single mean fluorescent expression level or by a 4- vector corresponding to variant distribution over the 4 expression bins, and with / without considering their total read count in the training process (i.e., sample weight). Sample weights corresponded to the proportional contribution of the variant's total read count to the total read count of the library, or per bin, depending on the model. Each model was tested with multiple read cutoffs, varying from 8957 down to 4 total reads per variant. This in turn yielded an increasing number of variants by which the model was trained, ranging from 10,000 to 140,000, respectively. We tested the model on 2000 randomly held-out variants, selected from the 10,000 highest read-count variants. Altogether, 112 models were tested (for full details see Methods).

[0138] The results of the Pearson correlation between predicted and measured mean fluorescence expression level of the held-out set are shown in Figure 3B. In general, all models behave similarly: Pearson correlation are at their highest values when the models are trained on up to 20,000 sequences with a total read cut-off of 6,009, and they decrease as the training set size increases (Fig. 3B). Interestingly, the prediction performance of the different models does not seem to be affected by including or excluding the barcodes, or by whether the variants are scored by mean fluorescence expression level or by the 4-bin distribution. Overall, the highest Pearson correlation achieved was a modest -0.45 at the plateau range, indicating that various sources of noise become increasingly dominant with a reduced readcount threshold. Consequently, we chose the 20,000 variants with total read count >6009 to be the training set for our initial “all-data” model (ADM).

[0139] Given the omni-present limiting noise in all models, we hypothesized that two dominant sources of noise may be the use of only two barcodes (the case for most OL variants: 146,802 variants), and a large subset of motifs that do not encode a regulatory function. This prompted us to construct two additional models, based on the following training sets. The first model is based on the observation from the various ADM models that neglecting barcodes does not seem to affect the predictive capacity of the model and is thus trained on the subset of 2435 variants with 22 total barcodes for each variant. We call this model the “all-motif’ model (AMM), as it contains all variants with 22 designed barcodes. The second model uses the subset of 2098 variants with 22 barcodes for each variant and at least one mixed-based motif. This model is constructed based on the assumption that mixedbased motifs are more likely to be regulatorily active due to the fact that they are composed of a larger number of potentially functional sub-variants. This model is referred to as the “mixed-based only” (MBO) model.

[0140] We then compared the performance of all three models on a held-out test set comprising 20% of the variants for each model. The Pearson correlations show a modest increase (Fig. 3C-blue bars) from the ADM baseline of 0.45 to approximately 0.48 for the AMM and 0.51 for the MBO model, despite the reduction by an order of magnitude in the size of the training set. Furthermore, we used the de Boer-Regev (dBR) model, trained on a different yeast expression library, to score the variants that make up the training set of each of our models. Here too (Fig. 3C-red bars), we observed a modest increase in the Pearson correlation between dBR-predicted and experimentally measured mean fluorescent expression levels from the ADM (0.05) to the AMM (0.21) and best-performing MBO (0.23) variant sets.

[0141] To experimentally validate the predictions of the various models, we constructed 11 sURS variants containing an uncharacterized mix of three of our 41 motifs. The variants utilized consensus TFBS motif sequences without mixed bases and were chosen as follows: the desert variant and 10 additional variants with MBO-based predicted mean fluorescence expression levels spanning the fluorescence range of 2700-3400 (A.U.). These variants were cloned individually upstream of the weak mCorel promoter as before, integrated into the yeast genome, and their expression levels were measured via flow cytometry (see Methods). As an evaluation on a small set of only 11 variants is highly sensitive to small changes in prediction, we used bootstrapping to report robust correlations. We sampled all subsets of 9 variants out of the 11 and reported the average correlation over these subsets. Moreover, we held out as an additional test set of 300 variants that were supported by the most reads from the set of 2435 variants with 22 barcodes for each variant.

[0142] The prediction-performance evaluation on the 11 sURS dataset shows the dBR model outperforms both the AMM and ADM models (Fig. 3D). The dBR model achieves an average Pearson correlation of 0.65+0.066, surpassing the AMM model with an average correlation of 0.45+0.087, and significantly outperforming the ADM model, which achieved an average Pearson correlation of -0.18+0.144. However, the prediction-performance evaluation on the test set of 300 high-quality variants reveals that ADM model performs better than both the AMM and dBR models. The ADM model achieved a Pearson correlation of 0.67, surpassing the AMM model with a Pearson correlation of 0.64, and significantly exceeding the dBR model, which achieved a Pearson correlation of 0.11. The MBO model achieved the best combined prediction performance on both test sets: an average Pearson correlation of 0.60+0.075 on the 11 sURS variants, and a Pearson correlation of 0.61 on the 300 high-quality variants (Fig. 3E).Example 4: mHG analysis identifies activating and repressing motifs and sub -motifs from the 2435 variant subset

[0143] We next proceeded to identify the motifs which likely play a regulatory role in our system by employing a minimum hypergeometric (mHG) analysis on the 2435-variant set (i.e., each variant encoded with 22 barcodes see Materials and Methods). Based on this analysis, we computed the mHG p-value for enrichment of each of the 41 motifs and the desert motif (Fig. 4A). We identified 7 motifs that were enriched at the top of the list (blue bars - mHG p-values <10A-4), 6 motifs that were enriched at the bottom of the list (red bars - mHG p-values <10A-4), and the remainder of the motifs were either unenriched or weakly enriched. The 10A-4 enrichment limit was set based on the results of the desert motif (blackbar) and the apparent empirical jump in p-values at the top of the list between 10A-2 and 10A-4. In the inset of Figure 4A, we present mean fluorescence expression level boxplots of all variants containing selected top-enriched (activating) and bottom-enriched (repressing) motifs, and compared to the rest of the variants, showing a clear up-shift or down-shift or down-shift in the distributions that aligns with the mHG p-value computation. Of the 13 identified significantly enriched motifs, 8 were mixed-based (5 of 7 activating, and 3 of 6 repressing) and 5 were of the non-mixed variety, providing further validation for the success of the MBO model.

[0144] Next, we applied mHG analysis to identify enriched sub-motifs within both the enriched and unenriched mixed-base motifs. To do so, we evaluated separately the number of reads and mean fluorescence expression level for each variant containing a particular submotif and reconstructed the mHG distributions in accordance with the new sub-variants (see Materials and Methods for detail) (Fig. 4B-4E). In each figure, the motif box -plots as compared with the rest of the population are shown on the left, while on the right we show the split sub-variant distribution, which can vary from 2 (e.g. Fig. 4B) to 16 (Fig. 4F). The figures show that in some cases, only one of the sub-motifs is actually enriched (Figure 4B- 4D), while in others all sub-motifs are enriched but with different p-values, suggesting a varying degree of enrichment (Fig. 4E). Altogether we identified 5 enriched non-mixed base motifs and 13 enriched sub-motifs (two of which were redundant leaving 12 enriched submotifs) (mHG p-value < 10A-4, Table 2 and Fig. 4F), which can be considered likely candidates for functional TFBSs. Of these functional motifs, 8 motifs are associated with down-regulation, while 10 are associated with up-regulation. Interestingly, 5 out of the 18 functional motifs do not have a known associated TF in any organism.

[0145] Table 2: Likely functional TF-BSs, those below the 10A-4 cutoff are in bold boxesExample 5: Boosting expression with statistically significant motifs that are embedded within sURS.

[0146] Based on the mHG analysis, we chose 23 motifs and sub-motifs for validation experiments: 20 motifs and sub-motifs with p-value <0.001, and 3 sub-motifs with higher p- values for sub-motif comparison (up to p-value of 0.01). In addition, a longer desert regulatory sequence was designed in silico, in a similar fashion as was done previously for the OL desert (see Materials and Methods). This enabled testing of the regulatory effect of up to six DNA motifs, compared to the OL variants, which incorporated up to three motifsper variant. Validation oligos were designed to address the following: 1) Evaluate the regulatory effect of weaker, but potentially significant, motifs (10A-4<mHG, p-value<10A- 3), 2) Validate new and strongly activating / repressing motifs individually (p-value<10A-4), and 3) Examine the effect of the motifs’ positions within the regulatory sequence on expression levels. The results (Fig. 5A-5C) show that sURSs behave as predicted: cassettes containing predicted up- and down-regulatory motifs induce and inhibit expression, respectively. Up- and down-regulation were observed for both promoter backgrounds (weak promoter-mCorel ; strong promoter-FEC-mCORl), suggesting that the total expression level is a multiplicative product of the core-promoter expression level and the degree of regulation generated by the sURS. In particular, up-regulation of x3-10 and x2-x4 was observed for the weak and strong core promoters, respectively, while approximately xl.5-x4 inhibition of expression was observed for both promoter backgrounds. Also, in general, no major position-dependent effect was detected, both experimentally and based on our ML model (Fig. 5B-5C).

[0147] To determine whether the effect of the motifs is additive as predicted by the billboard model, we inspected the fold-regulation measurement as a function of the mean loglO(p- value) that was previously determined by mHG analysis for all the motifs in the 6-motif cassettes (Fig. 5D). Here, we assume that the p-value is a good experimental proxy for the degree of fold regulation, as was shown in Figure 4A. The plot shows that for 6-motif cassettes whose fold-regulation was measured over the weak promoter background (blue squares), a distinct linear dependence on the mean log 10(p- value) is observed (Pearson correlation >0.85). This finding supports an additive regulatory effect, and the validated motifs have no interaction with one another. By contrast, for the downregulating 6-motif cassettes measured over the strong-promoter (red circles), no additive effect was observed. Instead, a constant ~x2 fold down-regulation is observed, independent of mean loglO(p- value) computed for each cassette. Finally, we compared the mean fold boost and attenuation effects for the 6- and 3-motif cassettes (Fig. 5E). For boost (blue bars), a clear dependence on the number of motifs is observed, while for attenuation (red bars), no such dependence is detected. Taken together, our mHG analysis and validation experiments support a regulatory model (Fig. 5G), where the expression level can be quantified as a product of the core expression level (Fig. 5F), the cumulative total boost contribution of all motifs on the sURS, and a ~x2 inhibition if one or more attenuating motifs are present. In our experiment and with these characterized motifs, this “motif additive model” (MAM) yielded a library ofverified sURSs in yeast spanning ~x50 range in expression level of yeCitrine with the same core promoter.Example 6: Boosting expression using inducible synthetic TFs with a constant sURS sequence

[0148] Given the observed differences between the rules for up- and down-regulation (Fig. 5G) and the fact that we used a similar architecture for all sURSs, we wondered whether this regulatory grammar is encoded within the sequence motifs of the TFBSs or somewhere else (e.g., within the effector domains of the putative TFs which binds these sequence motifs)? To answer this question, we designed a bacterial regulatory sequence (BRS) library composed of different combinations of the TFBSs of bacterial TF LexA, ranging from one to four TFBSs in each cassette (Fig. 6A). We constructed 15 different architectures of 1, 2, 3, and 4 binding sites, where binding sites were positioned in preset locations on the cassettes within a desert background. We encoded the library by using high-affinity (~lnM) binding site motifs picked at random (see Materials and Methods) at the LexA binding-site positions within the cassette, yielding a total of 21,502 unique variants, and 2, 4, 6, 8, or 18 barcodes per variant. To induce transcriptional regulation from the synthetic BRSs, we constructed five synthetic transcription factors using the full sequence of LexA fused to an effector domain of either Gal4, Hatl, the synthetic VPR (VP64-p65-Rta), Migl, or human Mxil (referred to as Mxil). LexA was used as a non-effector control. The LexA-Gal4, LexA-Hatl, and LexA- VPR chimeras were designed to be activators based on the effectors' native regulatory function, while LexA-Migl and LexA-Mxil were designed to be repressors. The BRS OL was cloned in yeast upstream of either the weak mCorel or strong FEC-mCorel core promoter to facilitate proper characterization of up-regulation or down-regulation, respectively (Fig. 6A), as was done for the sURS library characterized previously.

[0149] Upon expression of the chimera TFs in the presence of the BRS library and the relevant core-promoter, we observed either up-regulation or down-regulation indicating that the BRSs and synthetic TFs were functional in yeast. We next examined the normalized fluorescence of each of the architectures in the presence of the different synthetic TFs. The normalized fluorescence is defined in this case as the ratio of the mean expression level for a particular variant to that of the mean fluorescence measured for the desert sequence. In Figure 6B, we plot the expression levels for LexA-Gal4, showing a noisy but steadily increasing expression level as a function of the number of binding sites used. The plot shows the number of reads (in parenthesis) obtained for each architecture, which vary over a large range, thus leading to the noisy outcome. Normalized fluorescence expression data for theother synthetic TFs is plotted in Figures 6C-6F. Due to the inherent noise in expression, we were unable to determine whether or not there was a dependence of expression on binding site arrangement within the BRS variants. However, the total number of binding sites within the BRS cassettes does seem to play a role in the data for the up-regulators, as expected from what we observed for the sURS library. To confirm this observation, we computed the regulatory effect generated by the effectors. This was done (Fig. 6H-inset) by taking the ratio of the normalized expression level for a synthetic TF to the normalized expression level of the LexA alone (Fig. 6G), as measured for the same promoter and total number of LexA binding sites within the BRS. We plot the results for this ratio computation in Figure 6H. The results show that for all activators, a gradual increase in the ratio is observed as the number of binding sites increases (Fig. 6D, 6F), while for the LexA-Migl repressor the same level of down-regulation is observed (Fig. 6E), independent of the number of binding sites. Note, that LexA-Mxil does not seem to function well in our assay, as there is no distinct difference observed in its normalized fluorescence compared to that of LexA alone (Fig. 6C). Taken together, our analysis of a universal LexA BRS OL with a set of synthetic TFs containing known effector domains provides support for the model shown in Figure 5G, where the grammar rules seem to be encoded by the effectors rather than the DNA binding components of the TFs.Example 7: Boosted expression level correlates between yeast and mammalian cells.

[0150] The model shown in Figure 5G, and verified in yeast via the LexA BRS, indicates that for a given core promoter, gene expression is a product of independently functioning cis-regulatory elements and the core promoter strength. Since the motifs used in our experiments are predominantly conserved across organisms, the model suggests that the predictions generated thus far should also be valid for any growth condition in yeast and in mammalian cells if the biological role of the TFs that bind these DNA motifs is also conserved in higher eukaryotes across genomes.

[0151] To test this assertion, we first tested the 43 sURS variants that were synthesized to validate the MBO model (11 variants - Fig. 3D) and the mHG analysis (32 variants - Fig. 5A) in three additional growth conditions: SD-Ura / mCore promoter / 2%-glycerol / 30°C, SD- Ura / mCore-promoter / 2%-glucose / 39°C, and SD-Ura / mCore promoter / 2%- glucose / 30°C / lM NaCl. Together with the two conditions tested in Figures 5B-C (i.e. SD- Ura / mCore promoter / 2%-glucose / 30°C and SD-Ura / FEC-mCore promoter / 2%- glucose / 30°C), we were able to compare the results of four separate growth conditions and two promoter architectures. For each condition and architecture, measurements were carriedout in biological triplicates and growth curves were assessed to ensure that there were no adverse effects generated by the sURS sequences on the growth. The results show (Fig. 7A) a tightly correlated set of expression levels for all five conditions supporting regulatory roles of the sURS sequences that are independent of growth conditions and core promoter architecture in yeast.

[0152] Next, to test the broader applicability of the sURS sequences, we created for mammalian cells a regulatory architecture similar to the one synthesized for yeast that consisted of a sURS sequence positioned immediately upstream of the minimally active mammalian promoter pCMVmin (SEQ ID NO: 65) and an mCherry reporter. We created 32 mammalian sURS variants containing the 3 and 6 motifs that we used previously to validate the mHG analysis (see Fig. 5A). As a positive control for mCherry expression, we used the full pCMV promoter.

[0153] We transfected the mammalian variants (encoded on a plasmid) to CHO-K1 MI- HAC cells (hereby referred to as CHO) and assayed mCherry reporter expression. We compared the mean fluorescence expression levels in CHO cells directly to the levels obtained for the yeast variants grown in 2% glucose (Fig. 7B). The data shows a relatively correlated group of variants (blue), and an uncorrelated group of variants (red). The 24 “blue” variants that correlated well generated a statistically significant Pearson correlation of 0.75 (p-value 2.41e-5). The uncorrelated “red” variants were all various permutations of three motifs that were found to be repressive in yeast (GCTGCGCCAC (SEQ ID NO: 32), GAGGCGCAGC (SEQ ID NO: 42), AAACCCACACCCC (SEQ ID NO: 43)). This means that one or more of these motifs is likely activating in CHO cells due to the high mean expression observed from the sURSs. Next, we tested our variants in HeLa cells, and also compared the results to the 2% glucose yeast sample (Fig. 7C). The results show a similar relationship to yeast expression as was observed for CHO cells with a Pearson Correlation of 0.41 (p-value 0.054) for the same blue and red grouping of variants.

[0154] To further test for the cross-species applicability of these validation variants, we computed the correlation coefficient (Fig. 7D) and p-value (Fig. 7E) for any pair of datasets of the seven tested. The plots show that the regulatory output of the sURS variants in yeast are highly correlated across all five yeast conditions with a high degree of statistical significance (p-value <le-10). For the mammalian variants, the correlation between the HeLa and CHO data sets was similar to what was obtained by correlating the 24 blue CHO variants (i.e., blue circles in Fig. 7B) and the various yeast conditions, yielding moderately significant p-values that ranged between le-3 to le-5. The correlation for blue HeLa variantswith yeast was somewhat lower than what was obtained for CHO but was nevertheless marginally statistically significant for the 2% glucose growth condition (p-value -0.05). The similar cross-species correlations of yeast-CHO, yeast-HeLa, and CHO-HeLa with moderate statistical significance provide support to an interpretation that 24 of the 32 sURS sequences generate a regulatory response that is independent of the eukaryotic cell-type that was used.Example 8: Modelling expression for all cell-types using a combined machine-learning and mechanistic model

[0155] To provide further support to the cell-type-independent interpretation of the experimental data, we hypothesized that a ML model trained on yeast data should also provide adequate predictions for the mammalian expression data. Thus far, in this work, we have developed two models: the MBO model that was trained on the 2435- variant set, and a mechanistic motif additive model (MAM - see Fig. 5G) that was developed based on the mHG analysis. Both models are deficient for the above stated task. The MBO model cannot be applied to the 6-motif variants, as it can only provide reliable predictions for 3-motif sequences that are identical in length to what it was trained on. The mechanistic MAM model is based on a simplified first-order model of transcriptional regulation, which assumes that the regulatory output is proportional to the sum of the individual contributions of each functional motif. Consequently, we needed to construct a hybrid ML-additive model (MLAM) that on the one hand can be applied to a sURS containing any number of motifs, while on the other, properly models second-order sequence-level events that are captured by the MBO and neglected by the MAM.

[0156] For the MLAM model, we extended our CNN by additional input of a 41 -long motif count vector. This additional input was provided by concatenating it to the output of the pooling layer to enable combination of both sequence-level features and motif-occurrence counts. In addition, we expanded the input to the maximum length and trained on shorter sequences by padding them first. We tested the MLAM model by providing a prediction for every single variant of the 43 -variant validation set. First, we compared the performances of the MAM to the MLAM model for all five yeast conditions tested. The data shows (Fig. 8A) that the MLAM model improved prediction performance over all five datasets indicating that it captures not only the first-order mechanistic effects, but also second-order sequencelevel effects to improve its reliability. We then tested both models on the 24 “blue” mammalian variants that experimentally correlated well with yeast. The data shows that while the MAM model failed to generate a reliable prediction for both the CHO and HeLacells, the MLAM model was able to generate predictions that were similarly reliable for both yeast (Fig. 8B) and mammalian (Fig. 8C) cells.

[0157] Although the invention has been described in conjunction with specific embodiments thereof, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art. Accordingly, it is intended to embrace all such alternatives, modifications and variations that fall within the spirit and broad scope of the appended claims.

Claims

CLAIMS:

1. A nucleic acid molecule configured to constitutively express a sequence of interest in a eukaryotic cell, the molecule comprising: a transcription factor binding region comprising at least two sequences selected from those provided in SEQ ID NO: 45, 42, 3, 30, 32, 36, 39, 43, 44, 46, 47, 48, 54, 55, 56, 57, and 58, a core promoter, and a sequence of interest operably linked to said transcription factor binding region and core promoter.

2. The nucleic acid molecule of claim 1, wherein said transcription factor binding region comprises at least three different sequences selected from those provided in SEQ ID NO: 45, 42, 3, 30, 32, 36, 39, 43, 44, 46, 47, 48, 54, 55, 56, 57, and 58.

3. The nucleic acid molecule of claim 1 or 2, wherein said transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 42, 45, 55 and 57.

4. The nucleic acid molecule of any one of claims 1 to 3, wherein said transcription factor binding region comprises repeats of at least one sequence selected from those provided in SEQ ID NO: 45, 42, 3, 30, 32, 36, 39, 43, 44, 46, 47, 48, 54, 55, 56, 57, and 58.

5. A nucleic acid molecule configured to constitutively express a sequence of interest in a eukaryotic cell, the molecule comprising: a transcription factor binding region comprising at least two sequences selected from those provided in SEQ ID NO: 1-41, a core promoter, and a sequence of interest operably linked to said transcription factor binding region and core promoter.

6. The nucleic acid molecule of claim 5, wherein said transcription factor binding region comprises at least one sequence selected from SEQ ID NO: 9-22, 42, 45, 55 and 57.

7. The nucleic acid molecule of claim 5 or 6, wherein said transcription factor binding region comprises repeats of at least one sequence selected from SEQ ID NO: 9-22, 42, 45, 55 and 57.

8. The nucleic acid molecule of any one of claims 5 to 7, wherein said at least two sequences are two different sequences.

9. The nucleic acid molecule of any one of claims 5 to 8, wherein said transcription factor binding region comprises at least three different sequences selected from those provided in SEQ ID NO: 1-41.

10. The nucleic acid molecule of any one of claims 1 to 9, wherein said at least two sequences are separated by a nucleic acid spacer sequence.

11. The nucleic acid molecule of any one of claims 4 to 10, wherein said at least two sequences are selected from SEQ ID NO: 2-3, 7, 13, 17-18, 24, 30, 32-33, 36, 39-40 and 42-65.

12. The nucleic acid molecule of claim 11, wherein said at least two different sequences are selected from SEQ ID NO: 3, 30, 32, 36, 39, 42-48 and 54-58.

13. The nucleic acid molecule of any one of claims 1 to 12, wherein said transcription factor binding region comprises or consists of between 100-200 nucleotides.

14. The nucleic acid molecule of any one of claims 1 to 13, wherein said eukaryotic cells are yeast cells.

15. The nucleic acid molecule of any one of claims 1 to 14, wherein said eukaryotic cells are mammalian cells.

16. The nucleic acid molecule of claim 15, wherein said mammal is a human.

17. The nucleic acid molecule of any one of claims 1 to 16, wherein said core promoter is selected from an mCorel promoter comprising or consisting of SEQ ID NO: 67, an FEC-mCorel promoter comprising of consisting of SEQ ID NO: 68 and a minimal CMV promoter comprising or consisting of SEQ ID NO: 65.

18. The nucleic acid molecule of any one of claims 1 to 17, wherein said constitutive expression comprises expression without the exogenous addition of a transcription factor or transcription activator.

19. The nucleic acid molecule of any one of claims 1 to 18, further comprising a 5’ untranslated region (UTR) between said core promoter and said sequence of interest.

20. The nucleic acid molecule of any one of claims 1 to 19, wherein said molecule is an expression vector.

21. The nucleic acid molecule of claim 20, wherein said expression vector is a eukaryotic expression vector.

22. A method of constitutively expressing a sequence of interest in a eukaryotic cell, the method comprising contacting said eukaryotic cell with a nucleic acid molecule of any one of claims 1 to 21, thereby expressing a sequence of interest in a eukaryotic cell.

23. The method of claim 22, wherein said contacting comprises expressing said nucleic acid molecule of any one of claims 1 to 21 in said eukaryotic cell.

24. The method of claim 22 or 23, wherein said method an in vitro method and said eukaryotic cell is in culture.

Images