Gene biomarkers for mammalian cell pool productivity prediction and screening

Abstract

Methods for selecting cell pools for production of recombinant proteins are disclosed. In particular, disclosed is the use of one or more genes whose expression level can be used as a biomarker to predict which cell pool, among a plurality of cell pools, produces a recombinant protein with higher titre and / or higher quality.

Description

[0001] Gene Biomarkers for Mammalian Cell Pool Productivity Prediction and Screening

[0002] Field of the disclosure

[0003] The present invention relates to methods for selecting cell pools for production of recombinant proteins. In particular, the present invention relates to the use of one or more genes whose expression level can be used as a biomarker to predict which cell pool, among a plurality of cell pools, produces a recombinant protein with higher titre and / or higher quality.

[0004] Background

[0005] Recombinant proteins, e.g. monoclonal antibodies (mAbs) are considered one of the most gamechanging products of the biopharmaceutical industry. The selection of robust monoclonal cell lines able to deliver particularly high rates of recombinant proteins remains a challenge for biomanufacturing, especially in the case of complex format antibodies.

[0006] A successful biomanufacturing process requires stable, robust and highly productive cell lines able to satisfy the current multiple grams per litre standards [1], Chinese hamster ovary (CHO) cells are currently representing the most predominant host cell lines for mAbs production, offering the advantage of delivery close to human-like product quality at meaningful product titre. Even though production of standard monoclonal antibodies is no longer considered to be problematic for the biopharmaceutical industry, the current entry of the “difficult to express” mAbs formats demand a continuous improvement of the industrial production pipelines [2], A typical cell line development (CLD) process of biomanufacturing, starts with transfection of host cell lines, e.g. CHO cells, with the plasmid vectors containing the multiple candidate sequences [3,4], After transfection, the cells are pooled into several master pools, from which the most stable are chosen for single cell cloning [5], The pools are transfected with vectors containing sequences that translate into different mAb formats of a molecule for a specific therapeutic target. Within a format, vectors also vary in chain ratios and configurations. The final pools of a particular gene configuration are selected based on stability and mAb production titre.

[0007] Cell line selection is a crucial part in the early research and development of therapeutic monoclonal antibodies [6], The whole process is set up to identify the most promising candidate cell line that can produce the desired molecule at an appropriate product titre with the highest possible product quality attributes. To ensure such a cell line can be found, various screening steps are performed on the initial cohort of monoclonal cell lines, all of which originate from a single cell deposition out of the final transfection pools [7], The screening steps can include 1) large scale screenings of thousands of cell lines in 384 and 96 well plates, 2) mid-scale arrangements like fed-batch fermentations in 24 deep-well and 3) screens targeted more towards the suitability of cell lines in a production-like process environment. The latter are typically performed in small scale, fully automated robotic fermentation systems like Ambr15 or Ambr250 [8,9], For “difficult to express” molecule formats, a multi stage screening pipeline is crucial to rule out all cell lines producing suboptimal mixtures of undesirable side-products, while at the same time filtering for highest productivity

[0010] , However, this process can only screen out cell lines based on the initial variability that was present when the transfection pool was generated

[0011] , In this context, the monoclonal cell line generation process does not offer any novel diversity. Hence the initial diversity is key. When transfection pools are cultivated in a bioreactor environment, large variability can be observed between the pools, especially when the configuration of the transfection vectors has been modified

[0012] , Screening on this variability, at the pool level, is essential in order to give the whole cell line screening process an optimal starting point, the best pool material to screen on. Evaluating transfection pools in a bioreactor system is a time and resource consuming effort

[0013] , It requires growing out pools over several weeks until they reach cell densities that can be seeded into a bioreactor system. Also, due to limited availability of reactors, one can only look at so many candidate pools for evaluation.

[0008] In the past, multiple studies using comparative transcriptomic, proteomic and metabolomics approaches were performed aiming to identify predictive biomarkers for CHO cell productivity and cell growth performance. Yet, the identified parameters often lack transferability to other studies and / or CHO host cell lineages and / or varying product formats. An overview on genes and gene products that correlate positively and negatively with recombinant CHO productivity in more than one study are summarized by the review of Kyriakopoulos & Kontoravdi, 2014

[0014] ,

[0009] Here we disclose a number of genes, whose expression levels in a very early cell cultivation stage in the clonal cell line development process have shown to surprisingly highly correlate with the overall product titre and quality variables of a pool in a later performed production process in fed-batch bioreactors (Figure 1). We assess the overall productivity by averaging the performance of around nine hundred cell lines expressing eleven different molecules. Said genes can be used as a biomarker, to identify highly-promising pools to start a subsequent resource-intensive screening campaign in standard manufacturing cell line development pipeline. In addition, the gene activity could be used to assess on a pool level the expressibility of different molecule formats and pick the most promising candidate molecule for a certain therapeutic target.

[0010] Summary of the disclosure

[0011] Broadly, the present invention provides methods for facilitating selection of a mammalian cell pool, among several candidate cell pools, which is predicted to express a recombinant protein of interest in high titres and / or high quality compared to the other cell pools.

[0012] Accordingly, a first aspect provides a method for facilitating selection of a mammalian cell pool, from among a plurality of candidate cell pools that produce a recombinant protein, the method comprising:

[0013] (a1) measuring expression level of a gene selected from a group consisting of Cnpy3, Lambl , Ptgfrn, Slc38a10, Ppfibpl , Apmap, Znf768, Trim46, C1qtnf1 , Nfe2l1 , Syne3, Chidl , Erp29, Tedd , LOC113831255, Plaur and LOC113833088 in samples from each of said cell pools; and / or (a2) measuring expression level of a gene selected from a group consisting of Ppp6c, Cops2, LOC113835967, Alg6, Capzal , Mirlet7a, Tbc1d19, LOC100773044, Phyh, Znf639, LOC103158671 , LOC113833053, Arl8b, Sh3bgrl, Chptl and Tlr4 in samples from each of said cell pools; and

[0014] (b) comparing expression levels measured in (a1) and / or (a2); wherein the cell pool(s) with lower expression levels as measured in (a1) and / or higher expression levels as measured in (a2) are selected for recombinant protein production.

[0015] Also described herein according to a second aspect is a computer program product, comprising computer readable instructions which, when executed by one or more processors, cause the one or more processors to carry out the method of the first aspect or any embodiments thereof.

[0016] Also described herein according to a third aspect is a non-transitory computer-readable medium having stored thereon computer readable instructions which, when executed by one or more processors, cause the one or more processors to carry out the method of the first aspect or any embodiments thereof.

[0017] Also described herein according to a fourth aspect is a system, comprising: at least one processor; and at least one non-transitory computer readable medium containing instructions that, when executed by the at least one processor, cause the at least one processor to perform the method of the first aspect or any embodiments thereof.

[0018] Brief Description of the figures

[0019] Figure 1 depicts cell line development, from pool generation until monoclonal cell line evaluation in the Ambr15 fermentation system. Fig 1 a schematic and Fig 1 b more detailed.

[0020] Figure 2 shows an example of original versus interpolated measurements of one sample during fermentation process.

[0021] Figure 3 shows the Venn-diagram for comparing the significant gene generated by in each statistical approach and each cell line performance label.

[0022] Figure 4 shows the linear correlation of Titer Cedex Day 10 versus Cnpy3 gene expression. A negative correlation between the two variables is shown with Pearson r2= 0.32.

[0023] Figure 5 shows the linear correlation of Titer Cedex Day 10 versus Cnpy3 gene expression in each single project. The association of Cnpy3 expression to the Titer Cedex Day10 is reduced compared to the correlation of Titer Cedex Day 10 versus Cnpy3 gene expression in the mixed samples approach.

[0024] Figure 6 shows the linear correlation of average Titer Cedex Day 10 per project versus average Cnpy3 gene expression per project.

[0025] Figure 7 shows the linear correlation of average Titer Cedex Day 10 per project versus average Cnpy3 gene expression per project. Overlapping average values of validation projects 10 & 11 . Figure 8 shows the linear correlation of average Titer Cedex Day 10 per pool versus average Cnpy3 gene expression per pool, using data from all 11 projects.

[0026] Figure 9 shows pairwise t-test between the Cnpy3 expression levels of lower producer pools and higher producer pools per project format.

[0027] Figure 10 shows linear correlation of average Effective Titer SEC-MS per project versus average Slc38a10 gene expression per project. Overlapping average values of validation projects 10 & 11.

[0028] Figure 11 shows linear correlation of average Effective Titer SEC-MS per pool versus average Slc38a10 gene expression per pool, using data from all 11 projects.

[0029] Figure 12 shows pairwise t-test between the Slc38a10 expression levels of lower producer pools and higher producer pools per project format.

[0030] Figure 13 shows linear correlation of average Titer Cedex Day 10 per project versus average Lambl gene expression per project. Overlapping average values of validation projects 10 & 11 .

[0031] Figure 14 shows linear correlation of average Titer Cedex Day 10 per pool versus average Lambl gene expression per pool, using data from all 11 projects.

[0032] Figure 15 shows pairwise t-test between the Lambl expression levels of lower producer pools and higher producer pools per project format.

[0033] Figure 16 shows linear correlation of average Effective Titer Cedex Day 14 per project versus average Ptgfrn gene expression per project. Overlapping average values of validation projects 10 & 11.

[0034] Figure 17 shows linear correlation of average Effective Titer Cedex Day 14 per pool versus average Ptgfrn gene expression per pool, using data from all 11 projects.

[0035] Figure 18 shows pairwise t-test between the Ptgfrn expression levels of lower producer pools and higher producer pools per project format.

[0036] Figure 19 shows linear correlation of average Titer Cedex Day 14 per project versus average Ppfibpl gene expression per project. Overlapping average values of validation projects 10 & 11 .

[0037] Figure 20 shows linear correlation of average Titer Cedex Day 14 per pool versus average Ppfibpl gene expression per pool, using data from all 11 projects.

[0038] Figure 21 shows pairwise t-test between the Ppfibpl expression levels of lower producer pools and higher producer pools per project format.

[0039] Figure 22 shows linear correlation of average Titer Cedex Day 12 per project versus average LOC113834280 gene expression per project. Overlapping average values of validation projects 10 & 11. Figure 23 shows linear correlation of average Titer Cedex Day 12 per pool versus average LOC113834280 gene expression per pool, using data from all 11 projects.

[0040] Figure 24 shows pairwise t-test between the LOC113834280 expression levels of lower producer pools and higher producer pools per project format.

[0041] Figure 25 shows linear correlation of average Effective Titer Protein A per project versus average Apmap gene expression per project. Overlapping average values of validation projects 10 & 11 .

[0042] Figure 26 shows linear correlation of average Effective Titer Protein A per pool versus average Apmap gene expression per pool, using data from all 11 projects.

[0043] Figure 27 shows pairwise t-test between the Apmap expression levels of lower producer pools and higher producer pools per project format.

[0044] Figure 28 shows linear correlation of average Titer Cedex Day 12 versus average Ppp6c gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0045] Figure 29 shows linear correlation of average Titer Cedex Day 12 versus average Cops2 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0046] Figure 30 shows linear correlation of average Titer Cedex Day 10 versus average Znf768 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0047] Figure 31 shows linear correlation of average Titer Cedex Day 10 versus average Trim46 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0048] Figure 32 shows linear correlation of average Titer Cedex Day 14 versus average C1qtnf1 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0049] Figure 33 shows linear correlation of average Titer Cedex Day 12 versus average LOC113835967 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0050] Figure 34 shows linear correlation of average Titer Cedex Day 14 versus average Alg6 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 . Figure 35 shows linear correlation of average Titer Cedex Day 12 versus average Capzal gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0051] Figure 36 shows linear correlation of average Eff. Titer SECMS versus average Mirlet7a gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0052] Figure 37 shows linear correlation of average Titer Cedex Day 10 versus average Nfe2l1 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0053] Figure 38 shows linear correlation of average Titer Cedex Day 10 versus average Tbc1d19 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0054] Figure 39 shows linear correlation of average Eff. Titer SECMS versus average LOC100773044 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0055] Figure 40 shows linear correlation of average Titer Cedex Day 14 versus average Phyh gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0056] Figure 41 shows linear correlation of average Titer Cedex Day 14 versus average Syne3 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0057] Figure 42 shows linear correlation of average Titer Cedex Day 10 versus average Znf639 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0058] Figure 43 shows linear correlation of average Titer Cedex Day 10 versus average Chidl gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0059] Figure 44 shows linear correlation of average Eff. Titer Cedex Day 14 versus average Erp29 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0060] Figure 45 shows linear correlation of average Titer Cedex Day 14 versus average LOC103158671 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 . Figure 46 shows linear correlation of average Titer Cedex Day 10 versus average LOC113833053 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0061] Figure 47 shows linear correlation of average Eff. Titer SECMS versus average Tedd gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0062] Figure 48 shows linear correlation of average Titer Cedex Day 10 versus average Arl8b gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0063] Figure 49 shows linear correlation of average Eff. Titer SECMS versus average Sh3bgrl gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0064] Figure 50 shows linear correlation of average Titer Cedex Day 10 versus average Chptl gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0065] Figure 51 shows linear correlation of average Titer Cedex Day 14 versus average Tlr4 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0066] Figure 52 shows linear correlation of average Titer Cedex Day 10 versus average LOC113831255 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0067] Figure 53 shows Linear correlation of average Eff. Titer Cedex Day 14 versus average Plaur gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0068] Figure 54 shows Linear correlation of average Titer Cedex Day 14 versus average LOC113833088 gene expression per project dataset. Project 1-9 used for biomarker identification and correlation coefficient. Overlapping average values of validation Projects 10 & 11 .

[0069] Figure 55 shows the linear correlation of Eff. Titer Cedex Day 14 versus average (n=8) normalized Cnpy3 gene count [VST] per pool in project 12. A negative correlation between the two variables is shown with Pearson r2= 0.986.

[0070] Figure 56 shows the linear correlation of Eff. Titer Cedex Day 14 versus average (n=8) normalized Cnpy3 gene count [logcpm] per pool in project 12. A negative correlation between the two variables is shown with Pearson r2= 0.995. Detailed description

[0071] In describing the present invention, the following terms will be employed, and are intended to be defined as indicated below.

[0072] “and / or” where used herein is to be taken as specific disclosure of each of the two specified features or components with or without the other. For example “A and / or B” is to be taken as specific disclosure of each of (i) A, (ii) B and (iii) A and B, just as if each is set out individually herein.

[0073] Further, as used in the following, the terms "particularly", "more particularly", "specifically", "more specifically" or similar terms are used in conjunction with optional features, without restricting alternative possibilities. Thus, features introduced by these terms are optional features and are not intended to restrict the scope of the claims in any way. The invention can, as the skilled person will recognize, be performed by using alternative features. Similarly, features introduced by "in an embodiment of the invention" or similar expressions are intended to be optional features, without any restriction regarding alternative embodiments of the invention, without any restrictions regarding the scope of the invention and without any restriction regarding the possibility of combining the features introduced in such way with other optional or non-optional features of the invention.

[0074] A composition as described herein may be a pharmaceutical composition which additionally comprises a pharmaceutically acceptable carrier, diluent or excipient. The pharmaceutical composition may optionally comprise one or more further pharmaceutically active polypeptides and / or compounds. Such a formulation may, for example, be in a form suitable for intravenous infusion.

[0075] A compound as described herein may be a small molecule (e.g. a small molecule inhibitor, activator, cofactor, etc.) or a large molecule (e.g. a biologic, therapeutic protein or peptide such as an antibody or compound derived therefrom, a nucleic acid, etc.). A compound may be an organic compound. A compound may be a pharmaceutically active agent (also referred to as a drug or therapeutic agent), or a degradation product thereof.

[0076] The systems and methods described herein may be implemented in a computer system, including or in addition to the structural components and user interactions described. As used herein, the term “computer system” includes the hardware, software and data storage devices for embodying a system or carrying out a method according to the above described embodiments. For example, a computer system may comprise a processing unit such as a central processing unit (CPU) and / or graphics processing unit (GPU), input means, output means and data storage, which may be embodied as one or more connected computing devices. Preferably the computer system has a display or comprises a computing device that has a display to provide a visual output display. The data storage may comprise RAM, disk drives or other computer readable media. The computer system may include a plurality of computing devices connected by a network and able to communicate with each other over that network. It is explicitly envisaged that computer system may consist of or comprise a cloud computer. The methods described herein may be provided as computer programs or as computer program products or computer readable media carrying a computer program which is arranged, when run on a computer, to perform the method(s) described herein. As used herein, the term “computer readable media” includes, without limitation, any non-transitory medium or media which can be read and accessed directly by a computer or computer system. The media can include, but are not limited to, magnetic storage media such as floppy discs, hard disc storage media and magnetic tape; optical storage media such as optical discs or CD-ROMs; electrical storage media such as memory, including RAM, ROM and flash memory; and hybrids and combinations of the above such as magnetic / optical storage media.

[0077] Figure 1 shows the schematic representation of a typical cell line development (OLD) in biomanufacturing. It starts with transfection of host cell lines, e.g. CHO cells, with the DNA constructs (e.g. plasmid vectors) encoding multiple candidate sequences translating into the recombinant protein (e.g. chains of the antibody) of interest. At this stage, the transfected cell lines are called pools. For the development of one recombinant protein, e.g. an antibody, several plasmid vectors are constructed. The vectors are varying on the final molecule format translated out of the sequences. For example, in case of an antibody, within the vectors that are translating into the same format, there is additional variability in the chain ratios and the chain configuration. This means that for a particular antibody to be developed, several formats and several chain ratios and configurations within each format are being examined at the pool level. To identify highly productive and stable pools to be used for the rest of the cell line development process, all the different cell culture pools need to be expanded and further evaluated on stability and production rate. Thus, the pool assessment process is a time-consuming (~ 8 weeks) and resource expensive activity. Out of this process, the final pools of a particular format and chain ratio and configuration are selected and they are used as starting material for single cell cloning generation (single cell cloning requires another ~ 4 weeks). Monoclonal cell lines originating from a single cell (one cell separated from the pool in each well of a 96 well plate) are screened for growth, productivity and product quality and only high producer clones are being progressively scaled-up from batch to fed-batch conditions. As such, highly productive pools will result in highly productive monoclonal cell lines. While in controlled mini-bioreactors systems (e.g. Ambr15 bioreactor on the far right of the figure), they are evaluated for antibody titre and product quality, until a lead clone is selected as the “best” producer clone.

[0078] As the first aspect, the present invention provides a method for facilitating selection of a mammalian cell pool, from among a plurality of candidate cell pools that produce a recombinant protein, the method comprising:

[0079] (a1) measuring expression level of a gene selected from a group consisting of Cnpy3, Lambl , Ptgfrn, Slc38a10, Ppfibpl , Apmap, Znf768, Trim46, C1qtnf1 , Nfe2l1 , Syne3, Chidl , Erp29, Tedd , LOC113831255, Plaur and LOC113833088 in samples from each of said cell pools; and / or (a2) measuring expression level of a gene selected from a group consisting of Ppp6c, Cops2, LOC113835967, Alg6, Capzal , Mirlet7a, Tbc1d19, LOC100773044, Phyh, Znf639, LOC103158671 ,

[0080] LOC113833053, Arl8b, Sh3bgrl, Chptl and Tlr4 in samples from each of said cell pools; and (b) comparing expression levels measured in (a1) and / or (a2); wherein the cell pool(s) with lower expression levels as measured in (a1) and / or higher expression levels as measured in (a2) are selected for recombinant protein production.

[0081] The methods of the present aspect may have any of the features described in relation to any other aspect.

[0082] The terms “host cell”, “host cell line” and “host cell culture” are used interchangeably and refer to cells into which exogenous nucleic acid has been introduced, including the progeny of such cells. Host cells include “transformants” and “transformed cells”, which include the primary transformed cell and progeny derived therefrom without regard to the number of passages. Progeny may not be completely identical in nucleic acid content to a parent cell, but may contain mutations. Mutant progeny that have the same function or biological activity as screened or selected for in the originally transformed cell are included herein.

[0083] The terms “level of expression” or “expression level” in general are used interchangeably and generally referto the amount of a gene product in a biological sample. “Expression” generally refers to the process by which information (e.g., gene-encoded and / or epigenetic information) is converted into the structures present and operating in the cell. Therefore, as used herein, “expression” may refer to transcription into a polynucleotide (e.g. mRNA), translation into a polypeptide, or even polynucleotide and / or polypeptide modifications (e.g., posttranslational modification of a polypeptide). Fragments of the transcribed polynucleotide, the translated polypeptide, or polynucleotide and / or polypeptide modifications (e.g., posttranslational modification of a polypeptide) shall also be regarded as expressed whether they originate from a transcript generated by alternative splicing or a degraded transcript, or from a posttranslational processing of the polypeptide, e.g., by proteolysis. “Expressed genes” include those that are transcribed into a polynucleotide as mRNA and then translated into a polypeptide, and also those that are transcribed into RNA but not translated into a polypeptide (for example, transfer and ribosomal RNAs).

[0084] The term “measuring” the expression level of a gene, as used herein refers to the quantification of the gene product, e.g. to determining the level of the corresponding protein, or preferably RNA (e.g. mRNA), in the sample, employing appropriate methods of detection described elsewhere herein. Methods of measuring RNA and protein are known to one skilled in the art and also disclosed herein.

[0085] In an embodiment, mRNAs or cDNAs from a gene of interest is detected using Northern, dot-blot, or polymerase chain reaction (PCR) analysis, array hybridization, RNase protection assay, or using DNA microarrays, which are commercially available, including DNA microarray snapshots. For example, realtime PCR assays such as quantitative PCR assays are well known in the art. In an illustrative embodiment of the invention, a method for detecting mRNA from a gene of interest in a sample comprises producing cDNA from the sample by reverse transcription using at least one primer; amplifying the cDNA so produced; and detecting the presence of the amplified cDNA. In addition, such methods can include one or more steps that allow one to determine the levels of mRNA in a sample (e.g., by simultaneously examining the levels a comparative control mRNA sequence of a “housekeeping” gene such as an actin family member). Optionally, the sequence of the amplified cDNA can be determined. Northern blot analysis is a conventional technique well known in the art and is described, for example, in Molecular Cloning, a Laboratory Manual, second edition, 1989, Sambrook, Fritch, Maniatis, Cold Spring Harbor Press. Typical protocols for evaluating the status of genes and gene products are found, for example in Ausubel et al. eds., 1995, Current Protocols In Molecular Biology, Units 2 (Northern Blotting), 4 (Southern Blotting), 15 (Immunoblotting) and 18 (PCR Analysis). Furthermore, gene expression can be measured also by NGS (next generation sequencing), an example of which is described in Example 1 herein.

[0086] From an industrial perspective, the disclosed genes (i.e. Cnpy3, Lambl , Ptgfrn, Slc38a10, Ppfibpl , Apmap, Znf768, Trim46, C1qtnf1 , Nfe2l1 , Syne3, Chidl , Erp29, Tedd , LOC113831255, Plaur, LOC113833088, Ppp6c, Cops2, LOC113835967, Alg6, Capzal , Mirlet7a, Tbc1d19, LOC100773044, Phyh, Znf639, LOC103158671 , LOC113833053, Arl8b, Sh3bgrl, Chptl , Tlr4) can be used as biomarkers to assess the expressibility of different molecule formats during the format screening process. Complementary, we suggest using this biomarker for pool selection for an already identified lead format. Both implementations open up opportunities for optimizing the recombinant protein, e.g. antibody, development process. Both different format pools and different gene configuration pools are cultivated in fed-batch bioreactor systems, which is a time consuming step that can take several weeks. By applying our biomarker-based screening method, the expressibility and productivity of the pools could be easier evaluated. The process can be optimized by examining only pools that are showing the desired biomarker activity. In that way, one can significantly reduce the occupation time of the bioreactor system, freeing up capacity for screening more molecules or just saving time and resources.

[0087] Being able to select pools that will give a better starting material, will result in monoclonal cell lines able to produce higher rates of titre. An increase in the later stage monoclonal cell line productivity rates is an invaluable advantage. Correspondingly, the whole cell line and bioprocess development pipeline of the “difficult to express” recombinant proteins will be substantially improved. Furthermore, gene biomarkers linked to the phenotype of increased recombinant protein production are always attractive choices for host cell line engineering, an actionable approach yet to be examined.

[0088] Therefore, the use of the RNA biomarkers in cell line and molecule format screening processes will drastically increase the throughput and possibilities of representative bioreactor performance prediction compared to techniques and workflows of nowadays.

[0089] In a preferred embodiment, the gene measured in step (a1) is selected from a group consisting of Cnpy3, Lambl , Ptgfrn, Slc38a10, Ppfibpl , Apmap. In a more preferred embodiment, the gene measured in step (a1) of the first aspect is Cnpy3.

[0090] In an embodiment, at least one gene from the group recited under (a1) (e.g. Cnpy3) and at least one gene from the group recited under (a2) (e.g. Ppp6c) is measured. Information on Cnpy3 gene from human and Chinese hamster is accessible via NCBI GenBank Gene ID: 10695 and 100769298, respectively. Methods for identifying or retrieving from databases (e.g. GenBank, Ensembl etc.) of orthologues, transcripts and protein sequences of a gene, e.g. Cnpy3, are well known to the person skilled in the art.

[0091] In an embodiment, the mammalian cells are CHO cells.

[0092] In an embodiment, the recombinant protein is an antibody or an antibody fragment. In an embodiment, at least two of said plurality of mammalian cell pools express the antibody or fragments thereof in a different format. In an exemplary embodiment, one cell pool expresses a 2+1 antibody and another cell pool expresses a 3+1 antibody. Further examples of antibody formats include, but are not limited to, those presented in Table 1.

[0093] The term “antibody” herein is used in the broadest sense and encompasses various antibody structures, including but not limited to monoclonal antibodies, polyclonal antibodies, multispecific antibodies (e.g. bispecific antibodies), and antibody fragments so long as they exhibit the desired antigen-binding activity.

[0094] An “antibody fragment” refers to a molecule other than an intact antibody that comprises a portion of an intact antibody that binds the antigen to which the intact antibody binds. Examples of antibody fragments include but are not limited to Fv, Fab, Fab', Fab’-SH, F(ab')2; diabodies; linear antibodies; single-chain antibody molecules (e.g. scFv, and scFab); single domain antibodies (dAbs); and multispecific antibodies formed from antibody fragments. For a review of certain antibody fragments, see Hol liger and Hudson, Nature Biotechnology 23:1126-1136 (2005).

[0095] In an embodiment, each of said plurality of cell pools contains more than one cell line.

[0096] In an embodiment, the amino acid sequence of the recombinant protein expressed by the cells within each cell pool is the same.

[0097] In an embodiment, different cell pools have been transfected by different DNA constructs (e.g. expression vectors) encoding the same recombinant protein. Said constructs can differ in terms of gene configuration, untranslated regions (e.g. 3’-UTR and / or 5’-UTR), promoter(s), poly(A) signals, and / or codon usage, while still encoding the same amino acid sequence of the recombinant protein.

[0098] In an embodiment, different cell pools have been transfected by different DNA constructs (e.g. expression vectors) encoding different recombinant proteins.

[0099] In an embodiment, the method further comprises: based on the selection, advancing one or more cell pools of the plurality of candidate cell pools to a next step of cell pool or cell line screening. In an embodiment, the next step of cell pool or cell line screening comprises a fed batch cell culture stage. In another aspect, a computer program product is disclosed, comprising computer readable instructions which, when executed by one or more processors, cause the one or more processors to carry out the method of any of embodiments of any method described herein.

[0100] In another aspect, a non-transitory computer-readable medium is disclosed, having stored thereon computer readable instructions which, when executed by one or more processors, cause the one or more processors to carry out the method of any of embodiments of any method.

[0101] In another aspect, a system is disclosed, comprising: at least one processor; and at least one non- transitory computer readable medium containing instructions that, when executed by the at least one processor, cause the at least one processor to perform the method of any of embodiments of any method.

[0102] The various concepts introduced above and discussed in greater detail below may be implemented in any of numerous ways, and the described concepts are not limited to any particular manner of implementation. Examples of implementations are provided for illustrative purposes. The following is presented by way of example and is not to be construed as a limitation to the scope of the claims.

[0103] Examples

[0104] Example 1

[0105] Here we present the largest and most diversified CHO transcriptomics dataset consisting of data derived from 892 different monoclonal cell lines, producing 11 different “difficult to express” antibodies, derived from 23 different gene configuration pools. We introduced three robust feature selection methods, one traditional differential expression analysis method and two machine learning approaches to the transcriptomics data to discover genes correlated to high product titre and quality. Several significant genes have been identified, with Cnpy3 gene expression showing a very strong negative correlation (Pearson r2=0.94) to the overall project productivity. These results were validated by hold-out cell lines expressing 2 different unseen antibody formats. Additionally, a comparison of high and low producer pools shows that Cnpy3 is significantly upregulated in less productive pools. Thus, we conclude that Cnpy3 gene has the potential to be used as a screening biomarker for assessing format expressibility and selecting pools with a high potential to deliver subsequent higher productivity rates, resulting in a substantially improved cell line and process development.

[0106] Materials and Methods

[0107] Monoclonal CHO cell lines culture

[0108] All cell lines were generated using the CHO Host Cell Line (International patent publication number WO 2019 / 126634 A2)

[0029] , Pools of cells that stably express the therapeutic proteins were generated as described in

[0030] , To produce samples from 1009 cell lines in a standardized way, the cell lines were cultivated in batches of around 87-96 samples at the time. The same procedure was repeated for 11 batches expressing 11 different therapeutic protein formats (Table 1). Each batch contains monoclonal cell lines derived from 2 different pools expressing different gene configurations of the sample recombinant protein. In Project 06, the 3 different pools were used for the generation of the monoclonal cell lines. The cell lines from each batch were randomly selected after single cell cloning by limited dilution and cultivated in 24 deep well plates at 350 rpm, 37°C, 85% rH and 5% CO2. Cells were passaged 3 times at a seeding density of 3 x 1 OA5 cells / ml every 3-4 days. During the third passage, cell banking was performed, saving 3 frozen vials per cell line. Three days after the last passage, pellets from each cell line were harvested fortranscriptomics analysis. In total 1008 samples were collected for sequencing. The whole set of samples derived from the 11 different batches were separated in two datasets. The first dataset consists of Projects 01-09 and the second of Projects 10-11. Projects 01-09 were used for biomarker identification analysis and Projects 10-11 were used as validation datasets for the biomarkers identified from the first dataset.

[0109] Table 1. Types of therapeutic proteins collected and number of samples per batch.

[0110] RNAseq sample measurements

[0111] RNA was isolated from the pellets using the Direct-zol RNA MiniPrep Kit from Zymo Research according to the protocol. The RNA sequencing was performed on a NovaSeq 6000 System (Illumina, Inc) in S4 flowcell using read mode SE100 for the first 3 projects and PE100 for the remaining 8 projects. Due to low RNA concentration 12 samples were discarded from the sequencing measurements.

[0112] Seed-train and fed-batch production assay

[0113] One batch of cell banking vials was used for a 4 weeks manual seed-train cultivation followed by fed- batch fermentation. CHO cells were cultured in 125ml shake flasks at 150 rpm, 37°C, 70% rH and 5% CO2. Cells were passaged at a seeding density of 3-14 x 1 OA5 cells / ml every 3-4 days. Fed-batch cultures were performed in the Ambrl 5 miniature bioreactor system (Sartorius Stedim Biotech, Sartorius AG) with chemically-defined production media. Cells were seeded with 8 x 1 OA6 cells / ml on Day 0 of the production stage after adaptation to production media during two passages. Cultures received daily feed medium after Day 3 and additional feed bolus on Days 3, 7 and 10. Cells were cultivated for 14 days. Cultivation in the ambrl 5 system was operated at set points of 36°C, DO 30%, pH 7.0 and an agitation rate of 1200 rpm. After the above several cultivation steps, out of the initial 1009 cell lines, 892 survived and were used for the whole study.

[0114] Cell line analytics and data pre-processing

[0115] Forthe final cell line characterization, 4 different productivity evaluation assays were performed. Product titre was measured from supernatant harvested from the fed-batch fermentation on days 10, 12 & 14 using Cedex Bio HT Analyzer (Roche CustomBiotech, F. Hoffmann-La Roche Ltd). Additionally, titre was measured by analytical protein A chromatography and ultra high performance liquid chromatography with ultraviolet (UV) light detection (Dionex Ultimate 3000 UHPLC fitted with POROS A 20 pm Column, Thermo Fisher Scientific Inc.). Product quality was assessed as percentage of correctly assembled protein (main peak on the chromatogram) and was measured by capillary electrophoresis sodium dodecyl sulphate (CE-SDS; HT Antibody Analysis 200 assay on the LabChip GXII system, PerkinElmer) under non-reducing conditions by relative quantification of the expected protein size to total protein content. Additional effective titre measurements were performed using quantitative size exclusion liquid chromatography (Dionex UltiMate 3000 UHPLC PREP System & Vanquish Flex UHPLC Systems, Thermo Fisher Scientific Inc.) coupled with mass spectrometry (Q Exactive UHMR Hybrid Quadrupole Orbitrap, Thermo Fisher Scientific Inc.), which will be subsequently denoted as qSEC-MS effective titre. The raw data was further processed with the Chromeleon™ Chromatography Data System (CDS) Software (Thermo Fisher Scientific Inc.). The last three measurements of titre with protein A chromatography, CE-SDS main peak percentage and qSEC-MS effective titre were measured only on supernatant harvested from fed-batch fermentation day 14 (hereafter referred to as Titer Cedex Day 14, Titer Protein A, Eff. Titer SECMS).

[0116] RNAseq data pre-processing

[0117] The quality of the raw fastq files was evaluated using the FastQC software (version 0.11.9). For the adapter's removal (ILLUMINACLIP:TruSeq3-PE.fa:2:30:10) Trimmomatic software (version 0.39) was used. Sequences for the transgene and the reference genome were mapped and furtherly analysed separately. For this study only the host genes were utilized for further analysis, which excludes the transfected genes coding for the antibody of interest. Reads were aligned to the reference genome file (CriGri-PICR, RefSeq Assembly accession: GCF_003668045.1)

[0031] using the HISAT2 software (version 2.2.1). The feature quantification was performed using HTseq (version 0.13.5). Both reference genome and genome annotation files were retrieved from the NCBI database (Release 252: October 15 2022). The final reads were normalized to the library size and library composition using EdgeR’s (version 3.38)

[0032] trimmed mean of M value method (TMM). Additional voom transformation and batch effect removal functions from limma package (version 3.52)

[0033] were applied to the reads to prepare the data for linear modelling and to remove the sequencing batch effects that are observed. After filtering genes with low reads, 15913 genes were used as features space for the subsequent analysis.

[0118] Statistical methods for analysing RNAseq data

[0119] Differential expression analysis

[0120] Limma package is a core component of Bioconductor, an R-based open-source software development project in statistical genomics

[0033] , It is widely applied for the identification of differentially expressed genes. In our case, the samples are not annotated with qualitative but rather continuous quantitative labels, namely titre and product quality. The limma approach can relate gene expression to such continuous labels, often referred to as quantitative traits in the field of bioinformatics. In this case, the log fold change represents the change in gene expression for each unit increase in the quantitative trait. In this study we utilized the robust linear model based method that require the normal assumptions of response variable:

[0121] Y Expressioni P() + P1 *X Titer + Pp * X Project + Si (1) where Expressioni indicates the normalized gene expression value. Xnter represents each continuous trait such as Titer Cedex, Titer Protein A, MP CESDS and the effective titres. XprOject represents the covariate effect of each project. From the model, the differentially expressed genes can be identified by t-statistic for each gene under the null hypothesis HO : p1 = 0, while correcting for multiple testing errors using Bonferroni correction.

[0122] Machine learning algorithms

[0123] In addition to the standard differential gene expression approach referred to above, we also applied a more general machine learning algorithm to the RNAseq expression data. By nature of RNAseq expression data the number of features (genes) is much higher than the number of observations (monoclonal cell lines). Therefore we selected models for predicting the productivity attributes that are naturally that are most suitable for this setup. Tree-based and regularization models are resistant to non-informative predictors

[0034] , Tree-based and regularization models do parameter estimation and feature selection at the same time. Here, in this study an elastic net regression model

[0035] and a random forest regression model

[0036] were used for uncovering genes that are highly associated with our quantitative traits. Elastic net is a regularization technique. Similar to the lasso

[0037] , the elastic net simultaneously does automatic variable selection and continuous shrinkage, and it can select groups of correlated variables

[0035] , The elastic net encourages a grouping effect, where strongly correlated predictors tend to be in or out of the model together. This is particularly useful when gene biomarkers discovery is the purpose of the study. With elastic net the genes that correlate together are all grouped together and either are reported as a group associated with the continuous trait or they are shrunk to zero and excluded from the equation. During an elastic net regression all the genes are used together in one equation similar to equation (2) to predict the quantitative trait, in contrast to the limma model, in which each gene's association to the train is being examined separately.

[0124] YTiter fi() + t *X Expression i + 2*X Expression^ + ... + fin^X Expression,, + p * X Project + e (2)

[0125] Elastic net utilizes L1 and L2 penalties on the sum of squared residuals in order to regularize the features coefficients

[0035] , As the L2 penalty is increased, the feature coefficients move closer to 0 at different rates. Adding L1 penalty has as consequence that some features coefficients are actually set to zero for some values of A.

[0126] Another regression model used for identifying significant genes in our study is a random forest model

[0036] , Tree-based models conduct intrinsically feature selection; if a feature is not used in any split during the construction of a tree, the prediction equation is functionally independent of the feature

[0034] , Decision Trees are great for obtaining non-linear relationships between input features and the target variable. Thus, we are using them in this study to identify genes that potentially have nonlinear relationship to the product titre. Random forest regression seems like a black box that cannot easily be represented as a function. It is an ensemble of decision trees, meaning many trees are constructed in a certain “random” way. Each tree is created from a different sample of observations and at each node, a different sample of features is selected for splitting. Each of the trees makes its own individual prediction and these predictions are then averaged to produce a single result.

[0127] For the implementation of the machine learning algorithms, the packages caret (version 6.0-92,

[0038] ), glmnet (version 4.1-4) and random forest (version 4.7-1) from R CRAN were utilized. For both random forest and elastic net regression the same approach was applied. The total amount of samples of the projects 01-09 we partitioned in 80% model training and 20% model testing split. During the training phase the model was set in 10 fold cross validation, for parameters tuning.

[0128] Results

[0129] Product titre variables

[0130] The performance of the clones was evaluated after the 14 days fed batch. This is a point during the development pipeline that is considered to be quite accurate with regards to the performance of the cell lines to the later production stage. In this study we have collected several different measurements for evaluating the performance of the cell lines during the Ambrl 5 fermentation (Table 2). We have collected 3 different titre analytics. On days 10, 12, and 14 of the fed batch fermentation, titre was measured using Cedex Bio HT Analyzer. The Cedex Bio HT Analyzer measurements were performed in several timestamps during the process in the unit of days. For example, measurements of day 10 were taken in a range of 9.5 to 10.5 days, similarly for day 12 in a range of 11 .5 to 12.5 days. For day 14 since it was the last fermentation day the measurement span was in the range of 13 to 13.9 days. Therefore alignment of titre measurements across a specified set of timestamps across all samples is necessary. The titre measurements were recalculated using linear interpolation, by specifying the interpolation to take place in the timestamps days 0.5, 3, 5, 7, 10, 12 and day 13.5. For simplicity, later the day 13.5 is referred to as day 14. Figure 2, demonstrates how the measurements of the different days were linearly interpolated to the timestamps 0.5, 3, 5, 7, 10, 12 and 13.5 days.

[0131] Additional titre measurement was performed using analytical protein A chromatography on day 14. On the same day the main product quality was assessed using CE-SDS. The CE-SDS measurements give as output the amount of correctly assembled main product contained in a concentration of 100 mg / L product. Multiplying the titre measured by Cedex BioHT by the main product percentage abundance (assessed by CE-SDS or SEC) we created the measure of effective titre Cedex and similarly the effective titre Protein A. Additionally effective titre was measured directly by a native quantitative SECMS (qSEC-MS) chromatography and analysis of the main product amount in harvested cell culture supernatant. All these labels were used as labels to predict in the machine learning model.

[0132] Table 2. Measurements for cell line performance evaluation.

[0133] Identification of genes correlated to high productivity and quality For identifying titre related genes, three statistical methods were employed. As a standard method to identify biomarkers from gene expression data we applied differential expression analysis using robust regression, for detecting gene expression correlated to the quantitative trait (product titre). In addition two machine learning approaches were employed here, namely elastic net regression and random forest regression. In both models 15913 genes were used as features for predicting each one of the bioreactor cell line product titre variables.

[0134] In all three methods, we included the batch name as surrogate for the specific antibody format as a covariate into the model, since it is a well-known variable that highly influences the therapeutic protein production. The amount of identified genes, per method and across each possible combination of methods are demonstrated in the venn diagram (Figure 3). Significant genes in robust regression were selected by applying the threshold of adjusted P-value > 0.05. In elastic net regression all genes with a coefficient different from zero have been taken into consideration. Lastly in random forest regression, we select genes that gather 5% of the cumulative feature importance.

[0135] In general, robust regression is identifying fewer significant genes compared to the machine learning approaches.

[0136] We double-checked each identified gene in the intersection of the three methods, by calculating the Pearson correlation coefficient between the gene levels and each cell line performance metric (Table 3). Two different approaches were employed in this calculation. The first, is a correlation of the gene expression versus the product titre variable in a complete set of mixed samples derived from the 9 different batches. The second, is the correlation between the average gene expression and the average of each product titre variable per project. Comparing the two approaches of mixed samples and average value per project, correlation coefficients are always higher in the latter case, indicating a stronger relationship between the overall project productivity rates and the genes levels.

[0137] Table 3. Intersection genes of the three different statistical approaches and Pearson correlation of each gene to each label.

[0138] Top hits (R2>0.5) from the correlation table are the genes Cnpy3, Lambl , Ptgfrn, Slc38a10, Ppfibpl , Apmap and LOC113834280. The expression of Cnpy3 gene shows the highest Pearson correlation coefficient to the titre Cedex measurements. Indicatively, Figure 4 demonstrates the linear correlation of Cnpy3 gene expression versus Titer Cedex Day 10. A negative correlation between the two variables is shown with r2= 0.32, indicating that as the expression of Cnpy3 is decreasing, the amount of recombinant protein produced is increasing. Figure 5 shows the linear correlation within the samples of each project separately. In a single project the correlation of Cnpy3 expression to the Titer Cedex Day10 is dramatically reduced. It seems like on a cell line level within a single project, other effects are overshadowing the Cnpy3 signal and hence there is no direct correlation with later bioreactor titre performance visible.

[0139] Biomarker for format screening

[0140] Evaluating the association across projects by correlating the average titre and the average gene expression of each project, the association of the Cnpy3 expression and the Titer Cedex Day10 is very high (Figure 6). A very strong negative correlation between the two variables is shown with r2= 0.94, indicating that the expression of Cnpy3 is decreasing while the mean amount of recombinant protein produced per project is increasing. Considering the recombinant protein format that each of the projects produces, we observe that batches expressing the most complicated formats like Projects 05 and 06 are having the highest Cnpy3 expression, whereas Projects 07 and 08 that are proteins of lower structural complexity are having lower expression of Cnpy3. Respectively, the rest of the projects of intermediate complexity are appearing with intermediate gene expression levels.

[0141] Validation of significant gene using hold-out datasets

[0142] From the results of the analysis, we observe that there is a very strong correlation between the levels of Cnpy3 gene expression and the overall project productivity. In order to revalidate these results, the levels of average titre and gene expression of the Projects 10 & 11 were evaluated to check if they follow the correlation trend of the other nine projects. In Figure 7, the average gene expression levels and titre measurements of projects 10 and 11 are projected on the correlation plot of Projects 01-09. Figure 7 shows that the overlapping mean values of projects 10 & 11 are following the correlation trend validating the mutual relationship discovered using only the first 9 projects. It shows a very strong negative correlation between the levels of expression of Cnpy3 gene and the titre rates measured by Cedex. The above evidence highlights the potential of Cnpy3 to be used as a biomarker to assess the expressibility of different molecule formats and pick the most promising candidate molecule.

[0143] Gene configuration pools selection

[0144] In the next steps we analysed the correlation of the average gene expression values and product titre variables per pool. Each project consists of monoclonal cell lines derived from 2 different pools. Each pool within one project is producing the same format, however is transfected with a different gene configuration. Exception is Project 06 which has samples derived from 3 different pools. Subsequently, we examine the relationship between Cnpy3 levels and pool average titre levels. Figure 8 demonstrates the linear trend of the average “Titer Cedex Day 10” value versus average Cnpy3 gene expression per pool, using data from all 11 projects. The strong correlation already observed in the format level is maintained in the pool level correlation. A very strong negative correlation between the two variables is shown with r2= 0.83, indicating that the expression of Cnpy3 is decreasing while the mean amount of recombinant protein produced per pool is increasing. To furtherly evaluate the reliability of the differentiation between low producer pools and high producer pools within one molecule format, a pairwise t-test was performed (Figure 9). The total number of pools were separated to low and high producer pools. Low producer is the one out of the 2 pools per project that has the lower average “Titer Cedex Day 10” value and the high producer is the one with the higher average. For Project 06 which consists of 3 pools, the average titre of the lower producer pool was used twice for pairing twice, once per each one of the other 2 pools. Prior to the paired t-test, a normality test was performed to check whether the differences of the pairs follow a normal distribution. The Shapiro- Wilk’s test

[0039] was employed. The output of p-value=0.7, is greater than the significance level 0.05 implying we can assume the normality. For the pairwise t-test the null hypothesis (HO) was set as the mean difference (Cnpy3 expression in the higher pool - Cnpy3 expression in the lower pool) >= 0 and the alternative hypothesis (Ha) as mean difference (Cnpy3 expression in the higher pool - Cnpy3 expression in the lower pool) < 0. The outcome of adjusted p-value= 0.003 which is less than the significance level alpha = 0.05. We can then reject the null hypothesis and conclude that the expression of Cnpy3 is significantly higher in the pools that have lower titre. These results provide supporting evidence that Cnpy3 could be used as a screening biomarker, to identify pools with a high potential to deliver subsequent higher productivity rates when implemented in the standard project cell line development pipeline.

[0145] Slc38a10, Lambl , Ptgfrn, Ppfibpl , LOC113834280 and Apmap

[0146] Similar analysis to the one conducted for Cnpy3 indicates the potential of Slc38a10 to be used as a biomarker to assess the expressibility of different molecule formats and pick the most promising candidate molecule (Figure 10). Respectively, Figures 11 and 12 show that Slc38a10 could be used as a screening biomarker, to identify pools with a high potential to deliver subsequent higher productivity rates when implemented in the standard project cell line development pipeline.

[0147] Similar analysis for the genes Lambl (Figures 13, 14 and 15), Ptgfrn (Figures 16, 17 and 18), Ppfibpl (Figures 19, 20 and 21), LOC113834280 (Figures 22, 23 and 24) and Apmap (Figures 25, 26 and 27) shows the potential of these two genes to be used as biomarkers to assess the expressibility of different molecule formats and pick the most promising candidate molecule. However, there the evidence for identifying pools with a high potential to deliver subsequent higher productivity rates are weaker.

[0148] Identification of additional genes from one or two statistical methods

[0149] Apart from the genes identified using the three different statistical methods, robust regression, elastic net regression and random forest regression, there are additional genes identified by one (Random Forest, Robust Regression, Elastic Net) or two methods (Random Forest-Elastic Net, Random Forest - Robust regression, Robust Regression - Elastic Net). We selected as top genes those that are in the top 10% of the R2values and are identified in three or more of the 8 product titer labels studied (Titer Cedex Day 10, Titer Cedex Day 12, Titer.Cedex.Day.14, Titer.Protein.A, Eff..Titer.SECMS, Eff..Titer Cedex.Day.14). The repeated detection of certain genes across various product titer variables indicates a strong and reproducible signal. This adds confidence to the findings, as these genes are less likely to be false positives or artefacts of the analysis. We set as threshold the top 10 percentile of the genes R2, to reduce the number of genes to a manageable number that prioritizes the genes that show the highest correlation to the product titer variables. In total, we identified 27 genes (Ppp6c, Cops2, Znf768, Trim46, C1qtnf1 , LOC113835967, Alg6, Capzal , Mirlet7a, Nfe2l1 , Tbc1d19, LOC100773044, Phyh, Syne3, Znf639, Chidl , Erp29, LOC103158671 , LOC113833053, Tedd , Arl8b, Sh3bgrl, Chptl , Tlr4, LOC113831255, Plaur, LOC113833088) additional to those identified by all three statistical methods. Table 4, shows the 27 genes among the correlation coefficients, the product titer variables they are detected and the statistical methods that detected them. Figures 28-54 depict the correlation of the average gene expression to the average product titer variable along with the overlapping validation projects.

[0150] Table 4. Genes identified by one or two statistical methods, detected to three or more product titer variables and represent the top 10% of highest R2values. LOC1

[0151] In a subsequent study, we aimed to analyze the Cnpy3 gene transcript level directly in polyclonal CHO pools, all expressing stably a 1+1 bispecific xMAb antibody 5. For this, five (5) different CHO pools, pool 9, 60, 64, 66 and 67, with different gene configurations were cultivated in a chemically defined seed train medium. For the assessment and control of cell culture variance, eight (8) mini-pools were generated from each of the five lead pools. Each mini pool was used for the assessment of Cnpy3 gene copy number analysis by DRUG-seq sequencing. By this, a very strong negative correlation was detected between the average normalized Cnpy3 gene per pool and eff.Titer.Cedex.Day14 [mg / L] with Pearson correlation coefficient r2> 0.98 and r2> 0.99 using gene count either on VST (Figure 55) or logcpm (Figure 56) within project 12, respectively.

[0152] This data further supports that the Cnpy3 gene expression can be used for CHO cell productivity performance prediction, already directly in polyclonal CHO pools expressing a biologic products.

[0153] The FASTQ files were processed using Alithea recommendation (MERCURIUS-DRUG-seq-kit-PN- 10841-10851-11041-11051_User-Guide-v0.6.2. revA.pdf). In brief, the STAR index generation using the assembly CriGri-PICRH-1 .0 (GCF_003668045.3) concatenated with the target chain sequences in fasta format. Quality check of the FASTQ files was performed with FASTQC vO.12.1. STAR Solo mapping was conducted providing the ensembl v111 reference annotation with the target chain sequences as additional genes. Finally, aggregation of the mtx files was done in get format. Count normalization

[0154] The gene counts were normalized using the Variance Stabilizing Transformation, implemented in DESeq242and in CPM log 2 transformed, using the following formula for each sample: x'i=log2(xijnxj1000000+1)

[0155] Where

[0156] • xi is the raw count of the gene /

[0157] • x'i is the normalized count of the gene /

[0158] • n is the total number of genes

[0159] In both cases we excluded the counts of the transfected monoclonal antibody genes to consider the host as an independent system. Forthe normalization approach, R v4.3.0 and DESeq v1 .4.2 were used.

[0160] Discussion

[0161] This study represents one of the largest and most diversified studies of CHO transcriptomic datasets to date. Data from 892 different monoclonal cell lines, producing 11 different recombinant proteins, derived from 23 different gene configurations, are utilized to identify gene targets directly linked to CHO productivity. Three robust feature selection methods, one traditional differential expression analysis method and 2 machine learning approaches introduced to transcriptomics data, give insights on possible biomarkers associated with cellular productivity performance in bioreactors.

[0162] Several significant genes have been identified (Table 3), with Cnpy3 showing the strongest correlation to overall project productivity. Linear correlation of average titre per project versus average gene expression per project demonstrates a negative relationship between the two variables, with Pearson correlation coefficient r2= 0.94. The results are further verified by the average gene expression and titre of two unseen hold-out data sets expressing two different formats. Similarly, we observe a negative relationship between the average gene expression and titre per pool with Pearson correlation coefficient r2=0.83. The selection of pools transfected with different gene configurations while expressing the same protein format is furtherly validated by a pairwise t-test which leads to the conclusion that the expression of Cnpy3 is significantly upregulated in less productive pools. The protein produced by Cnpy3 is an endoplasmic reticulum protein that is reported to play an important role in the TLR maturation and inflammasome activation. The role of Cnpy3 activity in the recombinant protein folding / secretion pathway is unclear. References

[0163] All references cited herein are incorporated herein by reference in their entirety and for all purposes to the same extent as if each individual publication or patent or patent application was specifically and individually indicated to be incorporated by reference in its entirety. The specific embodiments described herein are offered by way of example, not by way of limitation. Any sub-titles herein are included for convenience only, and are not to be construed as limiting the disclosure in any way.

[0164] 1. Wang, B., T. Albanetti, G. Miro-Quesada, L. Flack, L. Li, J. Klover, K. Burson, K. Evans, W. Ivory, M. Bowen, R. Schoner, and P. Hawley-Nelson (2018) High-throughput screening of antibody-expressing CHO clones using an automated shaken deep-well system. Biotechnol. Prog. 34: 1460-1471.

[0165] 2. Tihanyi, B., and L. Nyitray (2021) Recent advances in CHO cell line development for recombinant protein production. Drug Discov Today Technologies 38: 25-34.

[0166] 3. Lai, T., Y. Yang, and S.K. Ng (2013) Advances in Mammalian Cell Line Development Technologies for Recombinant Protein Production. Pharm 6: 579-603.

[0167] 4. Bolisetty, P., G. Tremml, S. Xu, and A. Khetan (2020) Enabling speed to clinic for monoclonal antibody programs using a pool of clones for IND-enabling toxicity studies. Mabs 12: 1763727.

[0168] 5. Gross, A., J. Schoendube, S. Zimmermann, M. Steeb, R. Zengerle, and P. Koltay (2015) Technologies for Single-Cell Isolation. Int J Mol Sci 16: 16897-16919.

[0169] 6. Yang, W., J. Zhang, Y. Xiao, W. Li, and T. Wang (2022) Screening Strategies for High-Yield Chinese Hamster Ovary Cell Clones. Frontiers Bioeng Biotechnology 10: 858478.

[0170] 7. Tejwani, V., M. Chaudhari, T. Rai, and S.T. Sharfstein (2021) High-throughput and automation advances for accelerating single-cell cloning, monoclonality and early phase clone screening steps in mammalian cell line development for biologies production. Biotechnol Progr 37: e3208.

[0171] 8. Hsu, W.-T., R.P.S. Aulakh, D.L. Traul, and I.H. Yuk (2012) Advanced microscale bioreactor system: a representative scale-down model for bench-top bioreactors. Cytotechnology 64: 667-678.

[0172] 9. Moses, S., M. Manahan, A. Ambrogelly, and W.L.W. Ling (2012) Assessment of AMBRTM as a model for high-throughput cell culture process development strategy. Adv Biosci Biotechnology 2012: 918— 927.

[0173] 10. Pybus, L.P., G. Dean, N.R. West, A. Smith, O. Daramola, R. Field, S.J. Wilkinson, and D.C. James (2004) Model-directed engineering of “difficult-to-express” monoclonal antibody production by Chinese hamster ovary cells. Biotechnology and Bioengineering. 11. Tharmalingam, T., H. Barkhordarian, N. Tejeda, K. Daris, S. Yaghmour, P. Yam, F. Lu, C. Goudar, T. Munro, and J. Stevens (2018) Characterization of phenotypic and genotypic diversity in subclones derived from a clonal cell line. Biotechnol. Prog. 34: 613-623.

[0174] 12. Schlatter, S., S.H. Stansfield, D.M. Dinnis, A.J. Racher, J.R. Birch, and D.C. James (2005) On the Optimal Ratio of Heavy to Light Chain Genes for Efficient Recombinant Antibody Production by CHO Cells. Biotechnol. Prog. 21 : 122-133.

[0175] 13. Porter, A. J., A.J. Dickson, and A.J. Racher (2010) Strategies for selecting Recombinant CHO cell lines for cGMP manufacturing: Realizing the potential in bioreactors. Biotechnol. Prog. 26: 1446-1454.

[0176] 14. Kyriakopoulos, S., and C. Kontoravdi (2014) Insights on biomarkers from Chinese hamster ovary ‘omics’ studies. Pharmaceutical Bioprocessing.

[0177] 15. Vishwanathan, N., H. Le, N.M. Jacob, Y.-S. Tsao, S.-W. Ng, B. Loo, Z. Liu, A. Kantardjieff, and W.- S. Hu (2014) Transcriptome dynamics of transgene amplification in Chinese hamster ovary cells. Biotechnology and Bioengineering.

[0178] 16. Clarke, C., P. Doolan, N. Barron, P. Meleady, F. O’Sullivan, P. Gammell, M. Melville, M. Leonard, and M. Clynes (2011) Large scale microarray profiling and coexpression network analysis of CHO cells identifies transcriptional modules associated with growth and productivity. J. Biotechnol. 155: 350-359.

[0179] 17. Povey, J.F., C.J. O’Malley, T. Root, E.B. Martin, G.A. Montague, M. Feary, C. Trim, D.A. Lang, R. Alldread, A.J. Racher, and C.M. Smales (2014) Rapid high-throughput characterisation, classification and selection of recombinant mammalian cell line phenotypes using intact cell MALDI-ToF mass spectrometry fingerprinting and PLS-DA modeling. J. Biotechnol. 184: 84-93.

[0180] 18. Benjamin, M. (2015) Prevention and prediction of production instability of CHO-K1 cell lines by the examination of epigenetic mechanisms. Dissertation, Ludwig-MaximiliansUniversitat Munchen at <https: / / edoc.ub. uni-muenchen.de / 18845 / 1 / Moritz_Benjamin.pdf>.

[0181] 19. Yao, G., K. Aron, M. Borys, Z. Li, G. Pendse, and K. Lee (2021) A Metabolomics Approach to Increasing Chinese Hamster Ovary (CHO) Cell Productivity. Metabolites 11 : 823.

[0182] 20. Popp, O., D. Mueller, K. Didzus, W. Paul, F. Lipsmeier, F. Kirchner, J. Niklas, K. Mauch, and N. Beaucamp (2016) A hybrid approach identifies metabolic signatures of high-producers for Chinese hamster ovary clone selection and process optimization. Biotechnology and Bioengineering.

[0183] 21. Chakrabarti, L., A. Mathew, L. Li, S. Han, J. Klover, T. Albanetti, and P. Hawley-Nelson (2019) Mitochondrial membrane potential identifies cells with high recombinant protein productivity. J. Immunol. Methods 464: 31-39. 22. Meng, Y.G., J. Liang, W.L. Wong, and V. Chisholm (2000) Green fluorescent protein as a second selectable marker for selection of high producing clones from transfected CHO cells. Gene 242: 201- 207.

[0184] 23. Bailey, C.G., A.S. Tait, and N. Sunstrom (2002) High-throughput clonal selection of recombinant CHO cells using a dominant selectable and amplifiable metallothionein-GFP fusion protein. Biotechnol. Bioeng. 80: 670-676.

[0185] 24. Lang, S., 1 Delia Drewello, 1 Johannes Wichter, A. Nommay, B. Wilms, H.-P. Knopf, and T. Jostock (2016) Surface display vectors for selective detection and isolation of high level antibody producing cells. Biotechnology and Bioengineering.

[0186] 25. Kang, S., 1 Da Ren, 1 Gang Xiao, 1 Kristi Daris, 2 Lynette Buck, 2 Atim A. Enyenihi, 3Roman Zubarev, 4 Pavel V. Bondarenko, and R. Deshpande (2014) Cell line profiling to improve monoclonal antibody production. Biotechnology and Bioengineering.

[0187] 26. Doolan, P., N. Barron, P. Kinsella, C. Clarke, P. Meleady, F. O’Sullivan, M. Melville, M. Leonard, and M. Clynes (2012) Microarray expression profiling identifies genes regulating sustained cell specific productivity (S-Qp) in CHO K1 production cell lines. Biotechnol. J. 7: 516-526.

[0188] 27. Carlage, T., M. Hincapie, L. Zang, Y. Lyubarskaya, H. Madden, R. Mhatre, and W.S. Hancock (2009) Proteomic Profiling of a High-Producing Chinese Hamster Ovary Cell Culture. Anal. Chem. 81 : 7357- 7362.

[0189] 28. Edros, R., S. McDonnell, and M. Al-Rubeai (2014) The relationship between mTOR signaling pathway and recombinant antibody productivity in CHO cell lines. BMC Biotechnol. 14: 15.

[0190] 29. Ng, D., M. Zhou, D. Zhan, S. Yip, P. Ko, M. Yim, Z. Modrusan, J. Joly, B. Snedecor, M.W. Laird, and A. Shen (2021) Development of a targeted integration Chinese hamster ovary host directly targeting either one or two vectors simultaneously to a single locus using the Cre / Lox recombinase-mediated cassette exchange system. Biotechnol. Prog. 37: e3140.

[0191] 30. Carver, J., D. Ng, M. Zhou, P. Ko, D. Zhan, M. Yim, D. Shaw, B. Snedecor, M.W. Laird, S. Lang, A. Shen, and Z. Hu (2020) Maximizing antibody production in a targeted integration host by optimization of subunit gene dosage and position. Biotechnol Progr 36: e2967.

[0192] 31. Rupp, O., M.L. MacDonald, S. Li, H. Dhiman, S. Polson, S. Griep, K. Heffner, I. Hernandez, K. Brinkrolf, V. Jadhav, M. Samoudi, H. Hao, B. Kingham, A. Goesmann, M.J. Betenbaugh, N.E. Lewis, N. Borth, and K.H. Lee (2018) A reference genome of the Chinese hamster based on a hybrid assembly strategy. Biotechnol. Bioeng. 115: 2087-2100.

[0193] 32. Robinson, M.D., D.J. McCarthy, and G.K. Smyth (2010) edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26: 139-140. 33. Ritchie, M.E., B. Phipson, D. Wu, Y. Hu, C.W. Law, W. Shi, and G.K. Smyth (2015) limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43: e47-e47.

[0194] 34. Kuhn, M., and K. Johnson (2013) Applied Predictive Modeling. Springer DOI: 10.1007 / 978-1-4614- 6849-3.

[0195] 35. Zou, H., and T. Hastie (2005) Regularization and variable selection via the elastic net. Journal of the Royal Statistical Society.

[0196] 36. Ho, T.K. (1995) Random Decision Forests. Proceedings of 3rd international conference on document analysis and recognition.

[0197] 37. Tibshirani, R. (1996) Regression Shrinkage and Selection Via the Lasso. Journal of the Royal Statistical Society DOI: 10.1111 / j.2517-6161 ,1996.tb02080.x.

[0198] 38. Kuhn, M. (2008) Building Predictive Models in R Using the caret Package. Journal of Statistical Software.

[0199] 39. S.S., S., M.B. Wilk, and H.J. Chen (1968) A Comparative Study of Various Tests for Normality. Journal of the American Statistical Association DOI: 10.1080 / 01621459.1968.10480932.

[0200] 40. Liu, B., Y. Yang, Z. Qiu, M. Staron, F. Hong, Y. Li, S. Wu, Y. Li, B. Hao, R. Bona, D. Han, and Z. Li (2010) Folding of Toll-like receptors by the HSP90 paralogue gp96 requires a substrate-specific cochaperone. Nat. Commun. 1 : 79.

[0201] 41. Ghait, M., R.A. Husain, S.N. Duduskar, T.B. Haack, M. Rooney, B. Gbhrig, M. Bauer, I. Rubio, and S.D. Deshmukh (2022) The TLR-chaperone CNPY3 is a critical regulator of NLRP3-inflammasome activation. Eur. J. Immunol. 52: 907-923.

[0202] 42. Anders, S., Huber, W. (2010) Differential expression analysis for sequence count data. Genome Biology 11 , R106

Claims

Claims1 . A method for facilitating selection of a mammalian cell pool, from among a plurality of candidate cell pools that produce a recombinant protein, the method comprising:(a1) measuring expression level of a gene selected from a group consisting of Cnpy3, Lambl , Ptgfrn, Slc38a10, Ppfibpl , Apmap, Znf768, Trim46, C1qtnf1 , Nfe2l1 , Syne3, Chidl , Erp29, Tedd , LOC113831255, Plaur and LOC113833088 in samples from each of said cell pools; and / or(a2) measuring expression level of a gene selected from a group consisting of Ppp6c, Cops2, LOC113835967, Alg6, Capzal , Mirlet7a, Tbc1d19, LOC100773044, Phyh, Znf639, LOC103158671 , LOC113833053, Arl8b, Sh3bgrl, Chptl and Tlr4 in samples from each of said cell pools; and(b) comparing expression levels measured in (a1) and / or (a2); wherein the cell pool(s) with lower expression levels as measured in (a1) and / or higher expression levels as measured in (a2) are selected for recombinant protein production.

2. The method of claim 1 , wherein the gene is selected from a group consisting of Cnpy3, Lambl , Ptgfrn, Slc38a10, Ppfibpl and Apmap, and the cell pool(s) with lower expression levels of one or more of said genes are selected for recombinant protein production.

3. The method of claim 1 or 2, wherein the gene is Cnpy3.

4. The method of any of claims 1-3, wherein the gene expression level is measured at RNA (e.g. mRNA) level.

5. The method of claim 4, wherein RNA expression level measurement is by NGS, Northern blot, dotblot, PCR, RT-PCR, array hybridization, RNase protection assay, or microarrays.

6. The method of any of claims 1-5, wherein the mammalian cells are CHO cells.

7. The method of any of claims 1 -6, wherein the recombinant protein is an antibody or a fragment thereof, and optionally wherein at least two of said plurality of cell pools express the antibody or fragments thereof in a different format.

8. The method of any of claims 1-7, wherein each of said plurality of cell pools contains more than one cell line.

9. The method of any of claims 1-8, wherein the amino acid sequence of the recombinant protein expressed by the cells within each cell pool is the same.

10. The method of any of claims 1 -9, wherein different cell pools have been transfected by different DNA constructs (e.g. expression vectors) encoding the same recombinant protein.11 . The method of any of claims 1 -9, wherein different cell pools have been transfected by different DNA constructs (e.g. expression vectors) encoding different recombinant proteins.

12. The method of any of claims 1-11 , wherein the method further comprises: based on the selection, advancing one or more cell pools of the plurality of candidate cell pools to a next step of cell pool or cell line screening.

13. The method of claim 12, wherein the next step of cell pool or cell line screening comprises a fed batch cell culture stage.

14. A computer program product comprising computer readable instructions which, when executed by one or more processors, cause the one or more processors to carry out the method of any of claims 1- 11.

15. A non-transitory computer-readable medium having stored thereon computer readable instructions which, when executed by one or more processors, cause the one or more processors to carry out the method of any of claims 1-11.

16. A system comprising: at least one processor; and at least one non-transitory computer readable medium containing instructions that, when executed by the at least one processor, cause the at least one processor to perform the method of any of claims 1-

Images