Improved antibody production

Abstract

The present invention relates to methods and cell lines for increasing yield of proteins such as antibodies, and also methods of producing proteins in modified mammalian cells.

Description

10672-W001-SEC Electronically Filed:. August 12, 2025IMPROVED ANTIBODY PRODUCTIONFIELD OF THE INVENTION

[0001] The present invention relates to methods and cell lines for increasing yield of proteins such as antibodies, and also methods of producing proteins in modified mammalian cells.BACKGROUND OF THE INVENTION

[0002] Mammalian cells are used to express complex biological molecules such as multispecific antibodies. While multispecifics enable promising new strategies for treating human disease, their production at high purity and yield remains a challenge.

[0003] The two most common mammalian cell lines used in industry for antibody production are Chinese hamster ovary (CHO) and human embryonic kidney derived cells (HEK293), which are favored for being readily adaptable to growing at high density as a suspension culture in serum-free media. Additionally, both lines have established processes for fast transient expression and powerful genetic tools for stable recombinant protein expression using transposase integration and selectable markers with adjustable stringency. While there are strong arguments in favor of using mammalian expression to produce antibodies and multispecifics, processes tend to be slower, more complex and more expensive when compared to prokaryotic expression systems used to produce other classes of recombinant proteins. To overcome some of these issues, numerous strategies have been pursued to improve mammalian expression systems through optimization of process, media, vectors, cell line development strategy, and through genetic engineering of host cells. Despite these efforts, further studies are needed to improve mammalian expression systems.SUMMARY OF THE INVENTION

[0004] The present invention provides a modified mammalian cell comprising a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, HY0U1, INHBE, HIST2H2BF, CCDC30, AKNA, YARS2, KMT5C, PADI3, CALY,10672-W001-SECSREBF1, ISG20, SEC61A1, NCAN, ATF4, SH3BGR, DGAT2, SERPINA10, CITED2, HSPA1A, MAPK7, ABL2, or DKKl.

[0005] In an embodiment, the modified mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, HYOU1, INHBE, HIST2H2BF, CCDC30.

[0006] In an embodiment, the modified mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, or HYOU1.

[0007] In an embodiment, the modified mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C or INHBE.

[0008] In an embodiment, the modified mammalian cell comprises a reduction of the HIST2H3C gene.

[0009] In an embodiment, the modified mammalian cell comprises a reduction of the INHBE gene.

[0010] In an embodiment, the modified mammalian cell comprises a reduction of the SCFD1 gene.

[0011] In an embodiment, the modified mammalian cell comprises a reduction of the MDH2 gene.

[0012] In an embodiment, the modified mammalian cell comprises a reduction of the RRP12 gene.

[0013] In an embodiment, the modified mammalian cell comprises a reduction of the HYOU1 gene.

[0014] In an embodiment, the modified mammalian cell comprises a reduction of the HIST2H2BF gene.

[0015] In an embodiment, the modified mammalian cell comprises a reduction of the CCDC30 gene.

[0016] In an embodiment, the modified mammalian cell comprises a reduction of the AKNA gene.

[0017] In an embodiment, the modified mammalian cell comprises a reduction of the YARS2 gene.

[0018] In an embodiment, the modified mammalian cell comprises a reduction of the KMT5C gene.

[0019] In an embodiment, the modified mammalian cell comprises a reduction of the PADI3 gene.10672-W001-SEC

[0020] In an embodiment, the modified mammalian cell comprises a reduction of the CALY gene.

[0021] In an embodiment, the modified mammalian cell comprises a reduction of the SREBF1 gene.

[0022] In an embodiment, the modified mammalian cell comprises a reduction of the ISG20 gene.

[0023] In an embodiment, the modified mammalian cell comprises a reduction of the SEC61A1 gene.

[0024] In an embodiment, the modified mammalian cell comprises a reduction of the NCAN gene.

[0025] In an embodiment, the modified mammalian cell comprises a reduction of the ATF4 gene.

[0026] In an embodiment, the modified mammalian cell comprises a reduction of the SH3BGR gene.

[0027] In an embodiment, the modified mammalian cell comprises a reduction of the DGAT2 gene.

[0028] In an embodiment, the modified mammalian cell comprises a reduction of the SERPINA10 gene.

[0029] In an embodiment, the modified mammalian cell comprises a reduction of the CITED2 gene.

[0030] In an embodiment, the modified mammalian cell comprises a reduction of the HSPA1A gene.

[0031] In an embodiment, the modified mammalian cell comprises a reduction of the MAPK7 gene.

[0032] In an embodiment, the modified mammalian cell comprises a reduction of the ABL2 gene.

[0033] In an embodiment, the modified mammalian cell comprises a reduction of the DKKlgene.

[0034] In an embodiment, the modified mammalian cell comprises a reduction of both of the following genes: INHBE and HIST2H3C; HIST2H3C and HYOU1; HIST2H3C and HIST2H2BF; HIST2H3C and RRP12; HIST2H3C and CCDC30; or RRP12 and HIST2H2BF.10672-W001-SEC

[0035] In an embodiment, the modified mammalian cell comprises a reduction of either of the following genes: INHBE and HIST2H3C.

[0036] In an embodiment, the modified mammalian cell comprises a reduction of either of the following genes: HIST2H3C and HYOU1.

[0037] In an embodiment, the modified mammalian cell comprises a reduction of either of the following genes: HIST2H3C and HIST2H2BF.

[0038] In an embodiment, the modified mammalian cell comprises a reduction of either of the following genes: HIST2H3C and RRP12.

[0039] In an embodiment, the modified mammalian cell comprises a reduction of either of the following genes: HIST2H3C and CCDC30.

[0040] In an embodiment, the modified mammalian cell comprises a reduction of either of the following genes: RRP12 and HIST2H2BF.

[0041] In an embodiment, the modified mammalian cell produces a protein at a higher titer compared to a mammalian cell which comprises the gene. In an embodiment, the modified mammalian cell produces a protein at a higher titer compared to a mammalian cell which comprises at least one of the genes. In an embodiment, the modified mammalian cell produces a protein at a titer that is at least 10% more than the titer of the protein produced in a mammalian cell which comprises the gene. In an embodiment, the modified mammalian cell produces a protein at a titer that is at least 10% more than the titer of the protein produced in a mammalian cell which comprises at least one of the genes.

[0042] In an embodiment, the gene or genes is reduced by CRISPR.

[0043] In an embodiment, the gene or genes is reduced by CRISPR-Cas9.

[0044] In an embodiment, the mammalian cell is a mammalian expression cell host.

[0045] In an embodiment, the mammalian cell is a HEK293 cell.

[0046] In an embodiment, the mammalian cell is a CHO cell.

[0047] In an embodiment, at least 50% of cells in a pool have the gene(s) reduced. In an embodiment, at least 60% of cells in a pool have the gene(s) reduced. In an embodiment, at least 70% of cells in a pool have the gene(s) reduced. In an embodiment, at least 80% of cells in a pool have the gene(s) reduced. In an embodiment, at least 90% of cells in a pool have the gene(s) reduced.

[0048] The present invention provides a method of increasing titer of a protein, comprising producing the protein in modified a mammalian cell of the present invention.10672-W001-SEC

[0049] The present invention also provides a method of producing a protein, comprising expressing the protein in a modified mammalian cell of the present invention.

[0050] The present invention provides a method of producing a protein, comprising expressing the protein in a modified mammalian cell, which modified mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, HY0U1, INHBE, HIST2H2BF, CCDC30, AKNA, YARS2, KMT5C, PADI3, CALY, SREBF1, ISG20, SEC61A1, NCAN, ATF4, SH3BGR, DGAT2, SERPINA10, CITED2, HSPA1A, MAPK7, ABL2, or DKK1. In an embodiment, the mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, HYOU1, INHBE, HIST2H2BF, CCDC30. In an embodiment, the mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, or HYOU1. In an embodiment, the mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C or INHBE. In an embodiment, the mammalian cell comprises a reduction of the HIST2H3C gene. In an embodiment, the mammalian cell comprises a reduction of the INHBE gene. In an embodiment, the mammalian cell comprises a reduction of both of the following genes: INHBE and HIST2H3C; HIST2H3C and HYOU1; HIST2H3C and HIST2H2BF; HIST2H3C and RRP12; HIST2H3C and CCDC30; or RRP12 and HIST2H2BF. In an embodiment, the mammalian cell produces a protein at a higher titer compared to a mammalian cell which comprises at least one of the genes. In an embodiment, the mammalian cell produces a protein at a titer that is at least 10% more than the titer of the protein produced in a mammalian cell which comprises at least one of the genes. In an embodiment, the gene or genes is reduced by CRISPR. In an embodiment, the modified mammalian cell is a mammalian expression cell host. In an embodiment, the mammalian cell is a HEK293 cell. In an embodiment, the mammalian cell is a CHO cell. The present invention also provides a protein obtainable by expressing the protein in a modified mammalian cell of the present invention.BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figures 1A, IB, and 1C. An expression panel of multispecific antibodies highlights limitations of antibody production. (A) Five antibody formats were selected for evaluation with molecule schematics depicted under their VERITAS antibody format nomenclature (Biswas et al., mAbs 2023, 15 (1), 2207232). (B, C) Three sets of 10 molecules were composed from three Fv target pairs (i.e. Fv targets [1 or 2], [3 or 4], [5 or 6]). Fv targets10672-W001-SEC for each molecule are numbered and indicated in square brackets on the Y-axis. For bispecific molecules with two Fv sequences, targets are comma separated with the first and second number respectively corresponding to the light- and dark-gray portions of the schematics depicted in panel B. (B) Batch normalized titers relative to the best expressing parental mAb (mAb [1]) are shown as mean estimates with 90% prediction-interval error bars (N = 16, Batches = 4). The gray shaded region is the range of parental mAb titers for each set of Fv pairs. Reference titer measurements for mAb [1] averaged to 113 mg / L with batch effects on the order of ±50%. (C) The median percent of purified antibody found to be in the expected main peak (MP, light-gray), or as pre-MP (black) and post-MP (dark-gray) impurities by SEC. Molecules and Fv targets are sorted and arranged to match panel C. SEC purity data are missing for one molecule ([Fab*scFv] -heteroFc [6, 5]) due to insufficient yield.

[0052] Figures 2A, 2B, 2C: Antibody expression induces ER stress proportional to titer. Figure 2: Differentially expressed genes (DEGs) were identified by sequencing RNA captured two days after transfection with recombinant antibody expressing plasmids and compared to an empty vector control. (A) Five molecules, composed from Fv targets 1 (black) or 2 (gray), were selected for their wide-ranging impact on titer (Figure 1). Molecule schematics are arranged from highest to lowest titer. (B) Normalized titers from Figure 1 B are shown (left) for reference to illustrate the correlated change in gene expression (middle) for all observed down- and up-regulated DEGs; shown as blue and red box plots respectively. Expression values are log2 fold-change in normalized gene counts relative to an empty vector control. On the right, the bar chart shows the count of DEGs that are specific to the set of molecules highlighted in the matrix below. For example, more than 750 DEGs were uniquely associated with mAb [2] expressing cells, whereas 364 DEGs were associated with all five molecules. All sets of molecules with more than 50 specific and shared DEGs are shown alongside counts of DEGs that were uniquely associated with a single molecule. (C) Gene set enrichment for up- and down-regulated DEGs are shown for the top three enriched terms from the Reactome (REAC) and Gene Ontology Biological Process (GO:BP) gene set databases (Ashbumer et al., Nat Genet 2000, 25 (1), 25-29; Jassal et al., Nucleic Acids Research 2020, 48, D498-D503; and Aleksander et al., Genetics 2023, 224 (1)).

[0053] Figures 3A, 3B, 3C: A single knockout arrayed CRISPR screen for impact on titer. Figure 3: (A) Starting with a transposase integrated stable pool of HEK293-6E cells expressing mAb [2], de-novo knockouts (KO) were generated by lipofection of CRISPR RNP complexes. After recovery of edited cells to >95% viability, cultures were normalized to le610672-W001-SEC viable cells / mL to begin production. Viable cell density was measured on day 2 of production to estimate the impact of KO on growth (analysis in Figure 9). CM was harvested on production day 8 to measure secreted antibody titers. (B) An arrayed CRISPR KO library of 425 gene targets was designed to include 206 DEGs identified from the transcriptomics experiment presented in Figure 2, 76 published knockdowns (KD), 1 published overexpression (OE), and 146 genes selected for their known function or other mechanism of interest. (C, right) All KO targets found to have >10% impact on antibody titers and controls ( gray) are shown and sorted from largest increase (black) to largest decrease (light gray) in productivity. Plate normalized titers are presented as a mean estimate with 95% CI error bars (N > 3, Batches > 3). The gray shaded area highlights the null effect region (±10% impact on titer). (C, left) The reason(s) for inclusion in the KO library, as shown in panel B, are annotated as squares in the matrix to the left of normalized titers.

[0054] Figures 4A, 4B, 4C, 4D, 4E, 4F: Double knockout screen for impact on titer. (A) An arrayed CRISPR double knockout (DKO) screen was performed with the same cell line and protocol as Figure 3 with sgRNAs for two gene targets pooled and the volume of RNP doubled. (B) Legend for panels C-F indicating the color coding for DKO (gray), reference titers for KOI & KO2 (white circle and plus sign, respectively), and full dose single KO (black). Reference titers represent the expected effect as a single KO when sgRNA is mixed and diluted with the known null effect CDC42BPB control. (C-F) Normalized titers are shown as mean estimates with 95% CI error bars (N = 3, Batches = 3). The gray shaded area highlights the null effect region (±10% impact on titer). (C, D) DKO results for the top two single KO hits from Figure 3 are presented alongside results for reference, and full-dose KO. Additional controls include the DKO of TRAC and CDC42BPB null-effect genes (see Fig. 3C) and the reported Bax and Bakl DKO (Arena et al., mAbs 2019, 11 (5) 977) (E) Normalized titers for reference (white circle) and full-dose (black circle) single KO are shown for all targets included in the DKO screen to highlight the impact of sgRNA dosage on resulting titer. (F) All observed additive DKOs are presented. Additivity was defined as a DKO result (gray) that was greater than both reference titers for KOI and KO2 (white circle and plus sign, respectively).

[0055] Figures 5A, 5B, 5C, 5D, 5E: Generalization and product quality. (A) Two different HEK293F clones, expressing either mAb [7] or IgG-scFv [7, 8], were assessed for impact of KO / DKO on titers and product quality. RNP delivery was performed as described for the DKO screen in Figure 4. (B-E) Knockout conditions are color coded as DKO (purple), KO (orange), or control (green). The parental strain is annotated in the gray box on top of each10672-W001-SEC figure panel. Measurements are shown as mean estimates with 95% CI error bars (N = 4). (B, C) Normalized titers relative to the CDC42BPB KO for each parental strain are shown, and the region of null effect is shaded gray (±10% impact on titer). Average titer of the CDC42BPB KO control for the mAh [7] and IgG-scFv [7, 8] expressing strains were 106 and 82 mg / L respectively. (D, E) The percent of ProA purified protein associated with the SEC main peak (MP) is presented with a black vertical line marking the average MP % of the CDC42BPB reference control.

[0056] Figures 6A and 6B: Principal components analysis of DEGs. Figure 6: A principal components analysis was performed on normalized gene counts for all differentially expressed genes (DEGs) identified in this study to identify patterns of differential gene expression across the conditions tested. (A) Normalized gene expression counts are shown for the top explanatory gene of the top 9 principal components. Points are individual samples and conditions are sorted from highest to lowest titer with No DNA and empty vector controls shown first . PC rank number, variance explained, and explanatory gene are annotated in the gray boxes above each faceted panel. (B) Principal components 1 vs. 2 with samples delineated by molecule type.

[0057] Figures 7 A, 7B, 7C, 7D: Optimization of arrayed CRISPR KO by lipofectamine delivered RNP. Figure 7: (A, B) Three example strains were chosen for a CRISPR KO optimization study. (A) During passaging, strain 3 was noted to form clumps when counting cells. Each point represents a measurement taken prior to passaging, with a “Clumpiness” score computed as the number of clusters counted per le6 cells / mL. (B) A titration of CRISPR RNP was performed to determine an optimum amount for efficient insertion & deletion (indel) formation. Indels were quantified using Inference of CRISPR Edits (ICE) described in methods. (C, D) After optimizing the amount of RNP, cells and lipofectamine, a mock arrayed CRISPR screen was performed twice, with strain 1 & 2, using 9 example gene targets to assess the consistency of KO formation across batches, strains and sgRNA target genes. Before running the experiment, and based on results from panel B, a target of >70% KO formation was set to be suitable for screening. This benchmark is annotated on each figure as a dotted line. (C) A summary of ICE KO estimates observed across the two batches is summarized as cumulative distributions for each batch with individual measurements shown as points in the inset below with colors corresponding to sgRNA target gene. (D) Individual ICE KO estimates are shown for 4 replicates in each condition with a crossbar marking the mean of these replicates. Points are colored based on sgRNA target gene to match panel C.10672-W001-SEC

[0058] Figures 8A. 8B, 8C, 8D. 8E: Normalization of arrayed CRISPR screen titers. The arrayed single KO CRISPR screen was performed in three waves of three replicates each. Each wave utilized a freshly thawed vial of the parental HEK293-6E stable pool expressing mAh [2] and included a new set of sgRNA target plates. Each replicate within a wave screened the same set of sgRNA targets plus controls resulting in N = 3 spread across three consecutive weeks of independent RNP transfections. Some hits, such as HIST2H3C, and all controls were carried forward to later waves for further replication of an initial finding, hence N > 3 in Figure 3. (A, B) Significant batch and plate effects were observed. (A) The stable pool had reduced average titers over time in later waves and replicates. Initial expression levels for wave and replicate 1 were approximately 55 mg / L, but declined to 25 mg / L by the final replicate. This represents a roughly 50% reduction in productivity over the course of the experiment. (B) Plate-level effects for wave 2 are shown and were typically less than 20%, but were frequently more than 10%. (C-E) Comparative distributions of Batch and Plate normalized titers shows narrower, more symmetrical shape with fewer extreme values for the Plate normalization procedure. Plate normalization is thus performed by dividing raw titers by the mean of titers for each measured plate, or simply including plate and batch as a term in a linear regression. Note that this procedure assumes the typical effect is null, which aligns with what was observed for three control conditions (TRAC, CDC42BPB and No sgRNA) presented in Figure 3.

[0059] Figures 9 A, 9B. 9C, 9D. 9E: Analysis of recovery and production growth. (A-E) All KO targets were binned by median absolute deviation (MAD) of normalized titers and color coded with higher titers shown in red, and lower titers shown in blue to highlight correlation between titer and growth characteristics. All points are averages of at least three independent expression batches from the experiment presented in Figure 3 and Figure 8. (A) Relative cell density after recovering from CRISPR editing, and relative cell density on day 2 of production were roughly correlated. Dashed lines are ±20% of the identity line. All three controls are annotated along with PTEN, which was noted for causing slow growth during recovery, but faster growth during production. (B, C) Correlation between growth in early production and normalized titers is presented as a scatter plot and box plots grouped by MAD bins with top hits, controls and other targets annotated for discussion. (D, E) Percent viability and cell diameter are compared to normalized titer MAD bins.

[0060] Figures 10A and 10B: Design and optimization of DKO screen. (A) A pilot experiment to assess DKO efficiency was performed using BAX and BAK1 as example targets. Editing was quantified by ICE using two separate primer sets for each target. Individual ICE10672-W001-SECKO estimates for each sample are quantified and presented in the figure on the left. PCR products used for Sanger sequencing and subsequent ICE analysis are shown in the agarose gel on the right with ICE PCR primer set and intended KO annotated above each lane. (B) Four TRAC CDC42BPB controls were distributed in independent rows and columns of each plate as depicted in the plate schematic to enable estimation and normalization of plate level effects. This is necessary in the context of the DKO screen where the overall plate mean cannot be used to estimate plate effects because many samples are expected to have an impact on titer.DETAILED DESCRIPTION

[0061] Mammalian cells are used in pharmaceutical drug discovery for producing many classes of recombinant proteins. In particular, mammalian cells are the preferred host for producing antibody -based therapeutics due to their ability to express, assemble and secrete soluble molecules with multi-peptide chains that have fewer immune-response liabilities caused by non-mammalian post-translational modifications. Notably, six of the top ten grossing pharmaceutical products of 2023 are monoclonal antibodies (mAbs) with combined global sales of over 80 billion US dollars. This immense cost is in part a reflection of both the value of these biotherapeutics for patients and the combined costs of research, development, and manufacturing of such drugs. Thus, research into strategies to reduce costs is important for future development and access to antibody -based medicines.

[0062] More recently, the pharmaceutical industry has invested heavily in developing multispecifics, the next generation of antibody-based therapeutics. Multispecifics are molecules engineered to bind multiple targets through combining the fragment variable (Fv) domains of multiple antibodies, or other target binding domains, into a single molecule. While multispecifics have enabled promising new mechanisms and strategies for treating diseases, their complexity compounds development costs, resulting in a vast increase in the amount of required screening, optimization, and validation needed to produce one manufacturable product. Unfortunately, a multispecific with desirable functional properties may not be manufacturable simply due to reduced purity and yield relative to a typical mAb. To ensure effective molecules reach the clinic, and to keep multispecific development economically viable, further investment is needed to improve mammalian expression systems.

[0063] While some evolutionary approaches to host-cell engineering have been attempted, most published efforts focus on hypothesis-driven approaches to genetically alter a10672-W001-SEC specific cellular process, such as reducing apoptosis, modifying glycosylation, or altering the unfolded protein response (UPR). Such hypothesis-driven approaches have been greatly enabled by CRISPR genome editing tools that allow researchers to modify mammalian cell genomes through efficient generation of genetic knockins (KI) and knockouts (KO) - the respective introduction, or removal, of genetic information from the host genome. Additionally, RNA interference for gene knockdown (KD) has been used for both hypothesis driven and semi-unbiased screening for gene targets that modulate recombinant protein production.

[0064] Recent advances in direct sgRNA / Cas9 ribonucleoprotein complex (RNP) delivery into naive expression hosts, and availability of custom arrays of single-guide RNA (sgRNA) libraries, have now motivated the use of CRISPR for semi-unbiased candidate screening approaches, as was nicely demonstrated by Bauer et al. (2022) who screened 187 literature motivated candidate gene KOs for impact on antibody productivity in CHO cells (Bauer et al., Synthetic Biology 2022). While 187 single KO gene targets is a notable beginning, this set of candidate genes represents only a small fraction of the roughly 20,000 protein-coding genes, and potentially thousands of genes specifically involved in the secretory pathway of mammalian cells.

[0065] In pursuit of novel targets for host-cell genome engineering, a candidate gene screening approach was performed toward identification knockout combinations that improve complex multispecific antibody productivity in HEK293 cells. A critical limitation for future development of complex antibodies is that not all antibodies express well, and poor expression is retained in multispecifics that carry the Fv sequence of a difficult-to-express parental mAb. Differentially expressed genes (DEGs) were identified that respond to recombinant antibody expression and a common signature was found that correlates with secreted antibody titer regardless of antibody format, or multispecificity. An arrayed CRISPR screen of 425 target gene KOs, including 206 KOs of DEGs, was performed to assess their impact on productivity of a difficult-to-express mAb. This screen resulted in eight KOs identified with greater than 20% and up to 80% increased productivity. These eight hits were then multiplexed in a doubleknockout (DKO) screen to identify additive KO combinations. An assessment of three promising KO targets in an alternate HEK293 lineage was also performed, and it was determined that CRISPR mediated KO of the HIST2 gene cluster and INHBE have additive benefits that are robust to changing the HEK293 genetic background. Importantly, no observed negative impacts on product purity were found, and benefits were observed for both mAbs and a multispecific molecule.10672-W001-SEC

[0066] To understand how cells respond to complex antibody expression, five molecules were selected for bulk RNA-seq analysis early after transfection of a naive HEK293- 6E host. All five molecules shared a differential expression signature of ER, secretory and protein folding stress, but this signature was stronger for molecules with low productivity. To identify the genetic determinants of mammalian host productivity, an arrayed CRISPR knockout (KO) screening protocol was developed to assess the impact of single-and double- KOs on productivity, achieving typical de-novo KO efficiency of 70-90%. An arrayed library of sgRNAs was then designed to generate KOs of 206 differentially expressed target genes and 223 literature motivated targets. Screening of this sgRNA library identified KO combinations that increase in the productivity of a difficult-to-express mAb by up to 100%. No meaningful impact on product purity was observed, and KO impacts on productivity were observed in an alternate HEK293 background expressing a multispecific antibody, underscoring the generalizability of select findings.

[0067] Accordingly, knockout (reduction) of at least one of the following genes in a cell demonstrated increased titer of a protein compared to protein titer from a cell having said gene(s): HIST2H3C, SCFD1, MDH2, RRP12, HY0U1, INHBE, HIST2H2BF, CCDC30, AKNA, YARS2, KMT5C, PADI3, CALY, SREBF1, ISG20, SEC61A1, NCAN, ATF4, SH3BGR, DGAT2, SERPINA10, CITED2, HSPA1A, MAPK7, ABL2, DKK1.

[0068] Conversely, at least one of the following genes were reduced in a cell and demonstrated no increase in protein titer: HSPA5, INPP5D, PDIA4, CDC42BPB, DNAH17, PTEN, TNFRSF10D, No sgRNA (TE), CALCB, SNAP25, PDCD1, PIM2, POU4F2, ATF3, RUNDC3A, ACVRL1, ECEL1, FGF13, H0XA2, NHLH1, ASPHD1, SLC7A11, DAB2, ONECUT3, ATF6B, RRBP1, SEC16A, or MANF.

[0069] Mammalian cell lines include HEK293, CHO, and Expi293 cells. HEK293 cell lines are known in the art. HEK293 cell lines include, for example, 293 -6E, 293 T, 293H and 293F (see e.g. Abaandou et al. Cells 2021 10(7): 1667; and Malm et al., Sci Rep 10, 18996 (2020). The HEK293 cell line 6E is of the 293E lineage. In an embodiment, the cell line is HEK293. In an embodiment, the cell line is 6E. In an embodiment, the cell line is 293F. CHO cell lines are known in the art. CHO cell lines include, for example, CHO-ori, CHO-Prominus, CHO-Pro-, CHO-K1, CHO-DG44, CHO-DUK, CHO-DXB-11, ExpiCho, and CHO-S.

[0070] As used herein, “protein” refers to a secreted protein. While it is envisaged that a protein, as it relates to the present invention, refers to any secreted protein, non-limiting examples of proteins include antibodies, heterodimeric antigen-binding proteins, multispecific10672-W001-SEC antigen-binding proteins, IgG-scFvs, asymmetric IgG (heteroIgG), heteroIgG with a common light-chain (heteroIgG-cLC), IgG with a C-terminal single-chain Fv (IgG-scFv), and a heteroIgG with one scFv binding arm ([Fab*scFv]-heteroIgG).

[0071] As used herein, an “antibody” is an immunoglobulin molecule comprising 2 heavy chains (HCs) and 2 light chains (LCs) interconnected by disulfide bonds. The amino terminal portion of each LC and HC includes a variable region of about 100-120 amino acids primarily responsible for antigen recognition via the CDRs contained therein. The CDRs are separated with regions that are more conserved, termed framework regions (“FR”). Each LCVR and HCVR is composed of 3 CDRs and 4 FRs, arranged from amino-terminus to carboxyterminus in the following order: FR1, CDR1, FR2, CDR2, FR3, CDR3, FR4. The 3 CDRs of the LC are referred to as “LCDR1, LCDR2, and LCDR3,” and the 3 CDRs of the HC are referred to as “HCDR1, HCDR2, and HCDR3.” The CDRs contain most of the residues which form specific interactions with the antigen. The functional ability of an antibody to bind a particular antigen is, thus, largely influenced by the amino acid residues within the six CDRs. Assignment of amino acids to CDR domains within the LCVR and HCVR regions of the antibodies of the present invention is based on the well-known Kabat numbering convention (Kabat, et al., Ann. NY Acad. Sci. 190:382-93 (1971); Kabat et al., Sequences of Proteins of Immunological Interest, Fifth Edition, U.S. Department of Health and Human Services, NIH Publication No. 91-3242 (1991)). It is understand that other numbering conventions may also be used, such as, for example, Chothia (Chothia et al., “Canonical structures for the hypervariable regions of immunoglobulins”, Journal of Molecular Biology, 196, 901-917 (1987); Al-Lazikani et al., “Standard conformations for the canonical structures of immunoglobulins”, Journal of Molecular Biology, 273, 927-948 (1997)), and / or North (North et al., “A New Clustering of Antibody CDR Loop Conformations”, Journal of Molecular Biology, 406, 228-256 (2011)).

[0072] Titer” refers to a measure of all secreted material that binds Protein A. For at least in the context of an antibody, binding to Protein A is driven mostly by the Fc portion. Titer can be determined by determining the amount of protein bound to Protein A (or ProL, for example) by methods known in the art. In addition, one can visualize all secreted proteins present in conditioned media by running a gel (PAGE) and seeing how strong a band corresponding to the desired size product is relative to a non-expressing cell line. An affinity tag might also be added to the protein. The KOs of the present invention increase protein bound to ProA. Since there is little to no background signal in conditioned media, this material10672-W001-SEC is essentially all generated from the expression constructs (i.e. there is very little if any measurable non-specific binding). In an embodiment, the ProA bound purified material is >95% desired protein. In an embodiment, the ProA bound purified material is >95% desired antibody. In an embodiment, the ProA bound purified material is >95% desired IgG-scFv.

[0073] “Purity” refers to a measurement of how much of the Protein A purified material corresponds to the desired product. Purity can be determined by the percent Main Peak. There are a number of ways to measure purity, but generally, a sample run on a column will have the desired product run together as a "Main Peak" and fractions are collected from that peak and typically assessed by e.g. mass-spec to confirm the main peak was the desired product and identify potential impurities.

[0074] “Yield” refers to how much of the desired product was produced after purification, which is a function of titer and purity. To determine yield, there might be some measure of purity that one meets, e.g. >95%. One can measure the protein concentration after meeting said benchmark which allows one to calculate how much material was produced, which when compared to the volume of conditioned media that the material was purified from, one can calculate a volumetric yield (e.g. mg / L).

[0075] As HIST2H3C, INHBE and HY0U1 KOs did not impact SEC purity, it is envisaged that gains in titer from knocking out genes of the present invention can be realized as increased yield (i.e. increased productivity). An increase in titer is a benefit if the purity is not reduced.

[0076] “Reduction” (also referred to herein as a knockout) refers to a cell having one or both alleles of a gene inactivated in addition to the use of RNA interference (RNAi) tools such as small interfering RNA (siRNA) for inducing the silencing of protein coding genes. In an embodiment, the gene will be endogenously present in a cell, and a method(s) will be used to reduce the gene. This results in a cell not having a particular gene or multiple genes, and is referred to as a modified mammalian cell. A gene can be knocked out (reduced) by a method known in the art. For example, CRISPR is a common approach to knocking out genes. Cas9 may be used, but any double-strand DNA break generating site-specific nuclease would work (e.g. TALENs, Casl2a). Also, one can use CRISPR systems to insert sequence, which can also be used to disrupt genes (e.g. Knockin using HDR with a repair template, and "prime" editing). Cytosine base editors can also be used to introduce nonsense mutations to KO genes (Billon et al., Molecular Cell, vol. 67;6, pgs. 1068-1079.E4, Sept. 21, 2017). A pool of cells refers to multiple cells typically edited (e.g. by CRISPR) at the same10672-W001-SEC time. A pool of cells may be heterogeneous, in that some of the cells have a gene reduced and some contain the gene. In an embodiment, at least 70% of gene copies in a pool of cells have a gene reduced. In an embodiment, at least 70% of gene copies in a pool of cells have an insertion or deletion that causes a frameshift mutation. For diploid cells, that means the pool is a probabilistic mixture of homozygous, heterozygous and wildtype. In an embodiment, the mRNA levels of the genes are lowered significantly using RNAi tools (Weng et al., Biotechnology Advances, vol. 37;5, pgs 801-825, 2019).

[0077] Sequences of single guide RNA (sgRNA) can be designed or purchased. Kits may include multiple guides that target an early common exon in the gene to increase the likelihood that DNA sequence is lost between two cut sites, and to increase the likelihood that a frameshift mutation will result in nonsense mediated decay (complete loss of the transcript). Some genes, like histones, are too small to design multiple guides, so one is provided (guides require a minimum distance from each other, or they can interfere and reduce cutting efficiency). sgRNA sequences are disclosed herein. It is envisaged that knocking out of any and all transcripts that contain at least one of the sgRNA sequences are part of the present invention.

[0078] Proteins can be produced in mammalian cells comprising gene KOs of the present invention. Cells are cultured using techniques well known in the art.

[0079] Vectors containing the polynucleotide sequences of interest (e.g., the polynucleotides encoding the protein and expression control sequences) can be transferred into the cell by well-known methods, which vary depending on the type of cellular host. Examples of vectors include, but are not limited to, plasmids, viral vectors, non-episomal mammalian vectors and expression vectors, for example, recombinant expression vectors.

[0080] Recombinant expression vectors can comprise a nucleic acid in a form suitable for expression of the nucleic acid in a host cell. The recombinant expression vectors include one or more regulatory sequences, selected on the basis of the host cells to be used for expression, which is operably linked to the nucleic acid sequence to be expressed. Regulatory sequences include those that direct constitutive expression of a nucleotide sequence in many types of host cells (e.g., SV40 early gene enhancer, Rous sarcoma virus promoter and cytomegalovirus promoter), those that direct expression of the nucleotide sequence only in certain host cells (e.g., tissue-specific regulatory sequences, see Voss et al., 1986, Trends Biochem. Sci. 11 :287, Maniatis et al., 1987, Science 236: 1237, incorporated by reference herein in their entireties), and those that direct inducible expression of a nucleotide sequence in10672-W001-SEC response to particular treatment or condition (e.g., the metallothionin promoter in mammalian cells and the tet-responsive and / or streptomycin responsive promoter in both prokaryotic and eukaryotic systems (see id.). It will be appreciated by those skilled in the art that the design of the expression vector can depend on such factors as the choice of the host cell to be transformed, the level of expression of protein desired, etc. Expression vectors can be introduced into host cells to thereby produce proteins encoded by nucleic acids.

[0081] Typically, expression vectors used in a host cell (comprising at least one KO of the present invention) will contain sequences for plasmid maintenance and for cloning and expression of exogenous nucleotide sequences. Such sequences, collectively referred to as “flanking sequences” in certain embodiments will typically include one or more of the following nucleotide sequences: a promoter, one or more enhancer sequences, an origin of replication, a transcriptional termination sequence, a complete intron sequence containing a donor and acceptor splice site, a sequence encoding a leader sequence for polypeptide secretion, a ribosome binding site, a polyadenylation sequence, a polylinker region for inserting the nucleic acid encoding the polypeptide to be expressed, and a selectable marker element. The leader sequence may comprise SEQ ID NO: 1 (MDMRVPAQLLGLLLLWLRGARC) which can be encoded by SEQ ID NO: 2 (atggacatgagagtgcctgcacagctgctgggcctgctgctgctgtggctgagaggcgccagatgc). The leader sequence may comprise SEQ ID NO: 3. (MAWALLLLTLLTQGTGSWA) which can be encoded by SEQ ID NO: 4 (atggcctggg ctctgctgct cctcaccctc ctcactcagg gcacagggtc ctgggcc).

[0082] Various methods of protein purification may be employed to purify proteins, and such methods are known in the art.

[0083] Proteins can be biosynthesized, purified, and formulated for administration by well-known methods. For example, an appropriate host cell, such as HEK 293 or CHO, comprising at least one KO(s) of the present invention, is either transiently or stably transfected with an expression system for secreting proteins using a predetermined HC:LC (in the context of an antibody, for example) vector ratio if two vectors are used, or a single vector system encoding both heavy chain and light chain. Vectors suitable for expression and secretion of proteins are well-known. Following expression and secretion of an antibody, for example, the medium is clarified to remove cells and the clarified medium is purified using any of many commonly-used techniques. For example, the medium may be applied to a Protein A or G column that has been equilibrated with a buffer, such as phosphate buffered saline (pH 7.4). The column is washed to remove nonspecific binding components. The bound antibody is10672-W001-SEC eluted, for example, by a pH gradient (such as 0.1 M sodium phosphate buffer pH 6.8 to 0.1 M sodium citrate buffer pH 2.5). Antibody fractions are detected, such as by SDS-PAGE, and then are pooled. Further purification is optional, depending on the intended use. The antibody may be concentrated and / or sterile filtered using common techniques. Other materials than the antibody, such as host cell and growth medium components, and soluble aggregates and multimers of the antibody, may be effectively reduced or removed by common techniques, including size exclusion, hydrophobic interaction, cation exchange, anion exchange, affinity, or hydroxyapatite chromatography. The purity of the antibody after these chromatography steps is typically greater than 95%. The product may be frozen at -70 °C or may be lyophilized.EXAMPLESEXAMPLE 1: Fv sequence and format limit multispecific antibody productivity

[0084] Mammalian cells are used to express complex multispecific antibodies, which can be engineered from antibody binding domains (Fv targets) in many different arrangements, or “formats”. The diverse array of potential multispecific formats, and the unique expression liabilities they may have, presents an additional challenge for choosing an appropriate molecule to begin optimizing an expression host. To assist in selection of a suitable molecule for downstream genetic screening, an expression panel of thirty molecules representing five different multispecific formats constructed from six Fv sequences was created. Antibody formats included parental IgG (mAb), asymmetric IgG (heteroIgG), heteroIgG with a common light-chain (heteroIgG-cLC), IgG with a C-terminal single-chain Fv (IgG-scFv), and a heteroIgG with one scFv binding arm ([Fab*scFv]-heteroIgG) (Figure 1 A). The six Fv targets were then paired to represent sets of Fv sequences from parental mAbs that expressed well (Fv targets 3 and 4), poorly (Fv targets 5 and 6), or mixed (Fv targets 1 and 2) (Figure 1 B). Each Fv was engineered to be in both possible positions of each antibody format. For example, IgG- scFv [1, 2] has Fv 1 and Fv 2 formatted as the Fab and scFv binding arms respectively, and vice versa for IgG-scFv [2, 1], All thirty molecules were then transiently expressed in HEK293- 6E host cells and secreted antibody titer was measured by Octet using Protein A (ProA) biosensors. ProA purified material was then assessed for purity by size-exclusion chromatography (SEC) using methods described previously.

[0085] For details on antibody production from HEK293-6E and purification, see Example 10 and Li et al., STAR Protocols 2022, 3 (2), 101428.10672-W001-SEC

[0086] This experiment was over-replicated with sixteen replicates for each molecule distributed across four independent expression batches to gain insight into the statistical properties of our processes at the planned 24-well plate expression scale to be used for genetic screening. Titers were determined to be log-normally distributed with batch effects on the scale of 50%. Therefore, titers were presented as normalized estimates relative to the best expressing molecule, mAb [1], from a model that accounts for expression batch (Figure 1 B). Also shown is 90% prediction intervals to illustrate the range of typical individual measurements, which were approximately within ±15% of the mean normalized titer for a given molecule. These statistical properties were taken into consideration when designing the arrayed CRISPR knockout (KO) screen presented in Figure 3.

[0087] When comparing titers across molecules composed from the same Fv sequences, there was a striking pattern that nearly all multispecific molecules expressed within the observed range of their parental mAbs (Figure 1 B). For example, parental mAbs for Fv targets 1 and 2 (mAb [1] and mAb [2]) had relative titers of 1 and 0.3 respectively, and multispecifics composed from these Fv sequences had wide-ranging relative titers from 0.2 to 0.8 (Figure 1 B, left). In contrast, six of eight multispecifics containing good expressing Fv 3 and 4 performed well with relative titers ranging from 0.7 to 0.9 (Figure 1 B, middle). And, all but one multispecific containing poor expressing Fv 5 and 6 had low relative titers below 0.5 (Figure 1 B, right). The one exception, [Fab*scFv] -heteroFc [5, 6], failed to outperform its parental mAbs when accounting for large impurities found by SEC, and the reverse orientation molecule [Fab*scFv] -heteroFc [6, 5] was the worst expressing molecule in the entire panel (Figure 1 A & B, right). For the high expressing panel of Fv 3 and 4 containing molecules, two molecules underperformed, which may be due to Fv 3 failing to convert to an scFv module. While no clear correlation was found between Fv sequence and product quality, scFv containing formats consistently struggled with SEC pre-main peak impurities (Figure 1 C). These impurities, which typically correspond to aggregates and other large species, consistently resulted in 20 to 50% loss of purified material.

[0088] Taken together, the experiment presented in Figure 1 highlights two critical limitations for engineering multispecifics. First and foremost, not all parental mAbs can be expected to express well, and the expression liability of their Fv target binding domains are likely inherited by engineered multispecifics harboring a poor performing Fv. Second, formatspecific impurities limit final yields of some multispecific designs. These findings also suggest a few opportunities for improving the likelihood of success when developing future10672-W001-SEC multispecific molecules. One opportunity would be at the Fv discovery stage, where likelihood of success can be improved by having a large diversity of Fv sequences to choose from before designing a multispecific molecule. However, this does not solve the format purity problem, which may be difficult to address because of the relationship between format geometry and required biological function. A second opportunity stems from the observation that multispecifics do not necessarily express at far lower levels than a standard mAb, and thus increasing overall antibody productivity through optimizing the expression system would be a viable strategy for improving success in multispecific development programs.EXAMPLE 2. Antibody expression induces an ER stress response that correlates with titer

[0089] Genes were identified that are down- and up-regulated in response to recombinant antibody expression. While these differentially expressed genes (DEGs) would not necessarily be known to improve, or otherwise alter, productivity of a host cell, regulated expression by the host is circumstantial evidence that the gene is involved in antibody production in some capacity. Therefore, the hypothesis was that DEGs would be enriched for genes with a phenotypically relevant KO. From the expression panel in Figure 1 five molecules were selected that ranged from low to high relative titer. The molecule set is composed from Fv 1 and 2, with a focus on the parental mAbs and scFv-containing multispecifics that were prone to impurities (Figure 2 A). Bulk RNA was extracted two days after transfection with recombinant antibody expressing plasmids to identify RNA expression changes relative to an empty vector control at an early point in time when cells are responding to transgene expression. A “No DNA” control was also included to identify changes in gene expression caused by the vector alone.

[0090] After classifying down- and up-regulated DEGs, DEGs were compared and assigned and shared among each of the five molecules. The first and most obvious pattern that emerged was that the most difficult to express molecule, mAb [2], had over 800 uniquely assigned DEGs, whereas all other molecules had fewer than 40 (Figure 2 B, right). When considering all combinatorial sets of molecules that had at least 50 uniquely assigned DEGs, every set included mAb [2], and 364 DEGs were observed across all five molecules (Figure 2 B, right). When visualizing the expression changes for all observed DEGs there was a clear correlation between normalized titers and the distribution of expression values for down- and up-regulated DEGs (Figure 2 B, middle). This correlation resulted in a DEG detection bias in10672-W001-SEC favor of the most difficult to express molecule. Indeed, a principal components analysis found that the dominant expression signature, principal component 1 (PCI), explained 85% of the variance in expression of all classified DEGs, and corresponds to correlated expression changes with normalized titers ( Figure 6).

[0091] To identify more subtle gene expression patterns, the next major principal components were examined. PC2 and PC3 together explained only 7% of the remaining variance, and corresponded to DEGs that are down-regulated relative to the vector control, but not relative to the No DNA control (Figure 6). PC4 and PC5, explained less than 3% of the remaining variance and had an expression pattern that differed from PCI in that the poorly expressed IgG-scFv format induced less of an expression change relative to the heteroFc containing molecules (Figure 6 A). Interestingly, the top ten explanatory genes for PC4 and PC5 share similar functions in the secretory pathway and reside in the endoplasmic reticulum (ER). These ten genes include four chaperones (HSPA5 [BiP], HY0U1, CALR & HSP90B1), two dissulfide isomerases (PDIA4 & CRELD2), two ER associated degradation genes (HERPUD1 & DERL3), one ER stress response gene (MANF), and three genes with a role in calcium homeostasis (CALR, CRELD2 & CALCB). The well studied antibody chaperone, HSPA5 (BiP), suggests the heteroFc containing molecules may require more up-regulation of antibody assembly factors relative to the IgG-scFv format. That said, mAb [2] induced the largest increase in gene expression for this signature, which demonstrates that an Fv sequence in its native mAb format can still dramatically impact host cells beyond what was observed with more complex multispecific formats.

[0092] The notion that difficult-to-express antibodies struggle to fold and assemble appropriately in the ER was also supported by an analysis of enriched gene sets. Up-regulated DEGs were enriched for genes involved in an ER stress response to unfolded proteins, and negative regulation of transcription (Figure 2 C, right). Conversely, down-regulated genes were enriched for secreted extracellular matrix components (Figure 2 C, left). This pattern supports the theory that recombinant antibody expression results in an ER associated unfolded protein response (UPR), restricted transcription, and down-regulation of proteins that transit through the secretory pathway. Additionally, the correlation of differential gene expression with antibody productivity indicates that the strength of this stress response is a hallmark of low producing molecules.EXAMPLE 3. Design of an arrayed CRISPR KO library10672-W001-SEC

[0093] There are hundreds of proteins with known roles in the secretory pathway, and over 2,300 DEGs were identified as potential candidates for an arrayed KO screen. To reduce the candidate gene target list to a manageable size for screening, additional filtering was applied. 146 genes were selected based on known function or other roles relevant to secretion, and adding an additional set of 77 gene targets that had previously been identified to impact antibody productivity in CHO cells when knocked down (KD), or overexpressed (OE). After reserving multiple positions in the library for two editing controls with expected null effects on antibody titer (TRAC & CDC42BPB), and a “No sgRNA” control, the library was completed by including 206 DEGs. These DEGs were selected for being in the top 20 down- and up- regulated genes for each study molecule presented in Figure 2 A, or were DEGs observed for four of the five studied molecules. The library was arrayed across 5 96 well plates, which mapped to 18 24DWB expression plates. The resulting arrayed CRISPR KO library represents 425 total targets and is summarized in Figure 3 B.EXAMPLE 4. Optimization of an arrayed CRISPR KO screen for impact on antibody titer

[0094] In preparation for the arrayed CRISPR KO screen, delivery of CRISPR enzyme / sgRNA ribonucleoprotein (RNP) complex was optimized. Electroporation, sometimes referred to as nucleofection, is typically considered to be the gold-standard for RNP delivery into primary cell lines, and Bauer et al. (2022) had succuss using this method to efficiently generate KO pools in their screen of 187 candidate gene targets in CHO cells. However, reagents required for electroporation are considerably more expensive than a lipofection based approach, which requires less guide RNA, and no expensive single-use electroporation cassettes. Therefore, CRISPR editing conditions for lipofection delivered RNP in HEK293 cells was sought to be optimized.

[0095] Starting from a Synthego protocol for lipofection delivery of CRISPR components, a serial dilution of RNP was used. Using Inference of CRISPR Edits (ICE), a Sanger sequence based approach to quantify insertion and deletions (indels), it was determined that doubling the recommended RNP amount increased indel generation for TRAC, a positive editing control target (Figure 7 B). The RNP titration showed a clear dosage dependency across three separate HEK293 suspension lines, suggesting that indel formation using this method is quantitatively sensitive to the amount of RNP provided by the lipofectamine complex. Also observed was one strain that had lower indel estimates, which is thought to be due to the10672-W001-SEC propensity of this strain to form cell clusters, thus sheltering a population of cells from the transfection complex (Figure 7 A & B).

[0096] After further optimization of lipofectamine and RNP amounts, the maximum dose of lipofectamine was identified that could be used to deliver the highest amount of RNP without dramatically inhibiting growth of our cells due to lipofectamine induced toxicity. With these parameters fixed, a mock arrayed screen was run with nine example targets for two different strains across two independent transfection batches. While some variation was observed attributable to batch, strain, and target, KO estimates were consistently above our goal of 70% (Figure 7 C & D). With this experiment we were also able to gauge the typical recovery time after lipofection required for cultures to return to over 97% viability. Using this information, a timeline was established for editing, recovery and seeding edited cells for a batch production of a stable expressed antibody. The end result of this optimization is the protocol outlined in Figure 3 A.EXAMPLE 5. Quality control and normalization of arrayed CRISPR screen results

[0097] The arrayed CRISPR KO sgRNA library, which mapped to 18 24-well expression plates, was screened across three separate waves of up to eight plates each. Lipofection of CRISPR RNP for each wave was replicated three times on separate weeks. The starting cell line was a transposase integrated stable pool expressing the difficult-to-express mAb [2] in HEK293-6E host cells under hygromycin selection. When analyzing the average titer observed in the final harvest, a reduction was observed in productivity from 60 mg / L to 25 mg / L over time with each passing wave and replicate, likely reflecting a selection pressure against high expression of mAb [2] within the stable pool (Figure 8 A). Accounting for this systematic batch effect is necessary to compare KO effect sizes across waves and replicates.

[0098] In addition to batch effects, also observed was random systematic plate-level biases. Such effects can arise from subtle differences in plate handling during the experiment and ranged in size from 10 to 20% (Figure 8 B). To account for the observed systematic batch and plate-level effects, the distributions of relative titers after batch and plate normalization were compared. Plate normalization resulted in a distribution that was symmetrical around 1.0 and minimized spuriously large observations (Figure 8 C-E). It is important to note that plate normalization, which centers observations relative to the plate mean, assumes the average KO across each plate has no effect on titer. This assumption was valid in this case because effects10672-W001-SEC for each plate varied randomly around 0.0 across replicates, and because most KO targets did not have a measurable impact on titer (Figure 8 B).

[0099] To monitor KO impacts on growth during the screen, culture densities were measured after recovery and after two days of growth in production as an estimate of growth rate. Growth after recovering from CRISPR editing was correlated with growth in production, which suggests phenotypic impacts of gene editing were observable during recovery and production (Figure 9 A). Also observed was a weak correlation between growth during production, percent viability, and titer (Figure 9 B & C). This effect could be seen among the three controls where small differences in growth rate could account for small differences in final titers. For example, the “No sgRNA” control grew marginally faster (9% higher density on production day two) and had lower normalized titers relative to the rest of the library (9% decrease). This correlation was not universal as KO targets like PTEN, a well known tumor suppressor, grew considerably faster in production (23% higher density on production day 2), but had no impact on titer.EXAMPLE 6. Single KO screen identifies targets that increase antibody titer up to 80%

[0100] After completing the arrayed screen and accounting for systematic plate and batch effects, twenty KO targets were identified that decreased titer at least 10% and up to 25% (Figure 3 C). The ER stress response gene, MANF, when targeted for KO, caused the largest decrease in titer. MANF was included in the screen as a DEG that was identified to be up- regulated in response to expression of the four lowest producing molecules studied in Figure 2. MANF is known to have a protective role within the UPR, specifically in responding to inappropriate formation of disulfide bonds, and the expression pattern of MANF in our transcriptomic study was typical of other ER resident chaperones and disulfide isomerases37,38. The second largest observed decrease in titer resulted with the SEC16A KO target, which was included in the screen as a positive control for its role in the secretory pathway. SEC16A is part of the Secl6 complex which mediates COPII vesicle formation to enable transport from the ER to the Golgi apparatus - an essential step of the secretory pathway.

[0101] Interestingly, three KO targets (PIM2, DAB2 & ATF 6B) previously found to improve productivity when knocked down by RNA interference, were found to decrease titers in our KO screen (Figure 3 C, “Pub. KD”). While PIM2 and DAB2 were identified in a large- scale siRNA screen and not pursued in depth, ATF6B KD is a patented finding that improved10672-W001-SECIgG productivity by 15-36% in CHO-DG44 cells, so it was surprising to find the opposite effect. However, numerous differences in our experimental designs, which include differences in host cell species, KD versus KO, and different model antibodies, could all explain these opposing findings.

[0102] Six KO targets (ABL2, MAPK7, SERPINA10, CALY, KMT5C & YARS2) were identified that improved productivity more than 10%, and had previously been found to improve productivity when knocked down (Figure 3 C, “Pub. KD”). In total, twenty-six KO targets increased titers by more than 10%, eight of these by more than 20%, and the top hit increased titers by approximately 80%. Normalized titer of 1.2 is a 20% increase in titer, 1.7 is a 70% increase, 2 is a 100% increase. Seven of the eight KO targets that increased productivity by more than 20% were DEGs up-regulated in response to mAb [2] expression (Figure 3 C). When considering all KO targets that impacted titer by at least 10%, an enrichment for DEGs (p < 0.01 by proportion test) was observed, which supports our hypothesis that DEGs would be enriched for phenotypically relevant targets.10672-W001-SEC10672-W001-SECEXAMPLE 7. Double KO screen of top eight targets identifies additive KOs

[0103] Given the promising results from the single KO screen, a follow-up double KO (DKO) study was pursued to identify potential additive combinations (Fig. 4A). Previously, editing efficiency was observed to be RNP dose-dependent (Figure 7 B), a test was conducted to see whether pooling guides targeting two genes could achieve reasonable editing efficiency if we doubled the amount of RNP in the transfection complex . Unfortunately, the lipofectamine amount could not be increased as this harmed viabilities of our cultures. As our test case, BAX and BAK1 were selected as DKO targets previously shown to benefit antibody productivity in intensive CHO culture. Editing efficiency was assessed by ICE and it was found that combining the two multi-guide pools resulted in reduced KO efficiency (Figure 4 A; Figure 10 A). However, KO estimates in the DKO condition were still >50% and 80% for BAX and BAK1 respectively, which was deemed suitable for screening purposes.

[0104] Since the DKO library would be enriched for hits from our KO screen, the average effect within a plate would be expected to skew towards increased titer, and thus the mean titer within a plate would not be suitable for estimating systematic plate-level effects caused by variation in plate handling. To enable plate normalization in the DKO screen context, four DKO editing controls targeting TRAC and CDC42BPB distributed across independent rows and columns of each 24-well expression plate (Figure 10 B) were included. These controls enable estimation of plate-level effects when the average effect within a plate cannot be assumed null.

[0105] Given the reduced editing efficiency caused by pooling guide RNAs targeting multiple genes, two single KO reference conditions were included in the experimental design. The first, “KO full-dose”, represents the condition where only guides for a single KO target are included. This condition mimics the single KO screen where the ratio of lipofectamine and total guide was optimal for editing without compromising cell viability. The second, “KO reference,” represents the condition where the target KO of interest is pooled with guide RNA targeting CDC42BPB, which is known from our single KO screen not to impact titer. This condition is interpreted as the expected effect size of the single KO when editing efficiency is10672-W001-SEC reduced by dilution with guides targeting a second gene. Evidence for additivity was then interpreted as a DKO effect size that is larger than the KO reference effect of both targets (Figure 4 A & B).

[0106] The top eight KO targets that improved titer by >20% were then selected for multiplexing in a DKO screen. RNP dose dependent effects was observed for several targets (Figure 4 E). Although no benefit was observed from the previously reported BAX BAK1 positive control, the top two hits from the single KO screen had a clear and reproducible benefit to production titers (Figure 4 C, D). While SCFD1 KO full-dose and reference conditions each showed increased titers upwards of 60%, no additive DKO combinations with SCFD1 were found since no DKOs outperformed the SCFD1 KO reference (Figure 4 D). The top single KO hit, HIST2H3C, also increased titer across all combinations tested, albeit with a larger sensitivity to RNP dose as evidenced by a large gap between the HIST2H3C KO reference and full-dose effect sizes (Figure 4 C, E). Notably, five of the six DKOs with evidence for additivity included the HIST2H3C as a KO target (Figure 4 F). The two largest additive DKOs, which appeared to synergize with HIST2H3C, included a second HIST2 cluster target (HIST2H2BF), and the ER stress responsive chaperone, HY0U1 (Figure 4 F). Remarkably, the HIST2H3C HY0U1 knockout combination increased productivity of the difficult-to-express mAb [2] molecule by more than 100% on average.EXAMPLE 8. HIST2H3C and INHBE KO effects generalize to a multispecific and an alternate HEK293 host

[0107] Given our findings of KO targets with showed reproducible benefits to titer, whether the identified KO targets negatively impact product quality and whether the observations generalize to new antibodies and multispecifics was examined, in addition to whether these effects are specific to stable pools and if these observations are robust to changing the genetic background of the host.

[0108] The set of candidate KO targets were narrowed based on observations from the previous studies. Two top hits, SCFD1 and RRP12, were removed from further consideration for having consistently low viability and slow growth that resulted in unstable cultures (Figure 9). HIST2H2BF was excluded as it targets the same chromosomal region as the top hit, but has a weaker phenotype. MDH2 and CCDC30 were excluded for not reproducing well in the DKO screen. The top hit, HIST2H3C, was selected for having a strong and reproducible benefit to10672-W001-SEC titer and only a small impact on growth rates. HY0U1 and INHBE were both included as they showed additivity with HIST2H3C and minimal impacts on growth.

[0109] An experiment was designed to address the questions outlined above. Starting with an alternate HEK293 lineage, HEK293F, stable clones were produced that were selected for high expression of two previously unstudied molecules mAb [7] and IgG-scFv [7, 8], A batch expression following KO of the three selected targets of interest was performed (Figure 5 A). Antibodies were purified from conditioned media by ProA affinity, and the resulting purified material was assessed for impurities by analytical SEC.

[0110] No KO target considered in this experiment had a meaningful impact on product quality as measured by SEC Main Peak % (MP). All observations were within 1% of the average MP observed in the CDC42BPB KO control, which ranged from approximately 97% to 98% MP depending on the molecule expressed (Figure 5 D & E). Unfortunately, HYOU1 KO appeared to show no benefit to antibody titer alone, or when combined with HIST2H3C. However, HIST2H3C and INHBE both showed consistent benefit to antibody titer as single KOs and reproduced the evidence for additivity observed previously (Fig. 5B and C). The pattern of increased titers, while reduced relative to what was observed with HEK293-6E mAb [2] stable pools, was remarkably consistent between the two independent clones expressing different molecules and antibody formats (Figure 5 B & C). Given their additive benefits to antibody productivity, that is robust to changing the host genetic background, and robust to changing the expressed molecule, EHST2H3C and INHBE gene disruption was found to be a promising host-cell engineering strategy for improving productivity of antibodies and multispecifics in stable HEK293 production lines.EXAMPLE 9: Gene knockout in CHO cells

[0111] The genes HIST2H3A, HIST2H3B, HIST2H3C were knocked out in three CHO cell lines and compared to a KO the previously reported Myc gene. Titers of protein expressed in CHO cell knockouts were greater compared to Myc knockout cells on at least day 12. The single HIST2 guide targets three copies of H3.2 encoding genes in the HIST2 histone cluster in CHO cells, corresponding to HIST2H3A, HIST2H3B, HIST2H3C (a.k.a. H3C13, H3C14, H3C15). The guide employed in human cell experiments could not be exactly replicated for CHO KO experiments, and only targets HIST2H3B and HIST2H3C (a.k.a. H3C14, H3C15). In the human (293) experiments, HISTH33B and HIST2H3C are both potentially disrupted, but HIST2H3 A will be intact. In CHO cells, all three could be disrupted.10672-W001-SECEXAMPLE 10: MethodsCell culture, transient, and stable batch expression of antibodies

[0112] Culturing and expression conditions were performed as described in Li et al., STAR Protocols 2022, 3 (2), 101428. Briefly, for transient expression experiments, monocistronic plasmids encoding molecules of interest were transfected using PEImax (Polysciences Cat# 24765-1) and approximately 250 ng of plasmid DNA per le6 transfected cells. For stable pools and clones, batch expressions were initiated by seeding cells at le6 viable cells / mL. For de-novo knockout (KO) experiments, after recovery from KO generation, culture densities were normalized to le6 viable cells / mL using an automated liquid handler. All expressions were performed in 24 deep-well blocks using FreeStyle F-17 based media (Gibco Cat# A1383502) with a feed supplement of Tryptone-Nl (Organotechnie Cat# 19553) and glucose added at the beginning of production. Culture density and viability were measured with Vi-CELL BLU automated cell counters (Beckman Coulter). Conditioned media was harvested by clarifying centrifugation. Titers were measured using Octet HTX and QKe instruments (Sartorius) using Protein A (ProA) biosensors.Stable line generation

[0113] Stable pools were generated using the same transfection protocol as transient expression, but with an additional co-transfected plasmid expressing PiggyBac transposase. The resulting pools of semi-random integration of recombinant antibody coding sequences were then selected with 100 mg / L hygromycin B (Roche Cat# 10843555001). Transformed cultures were allowed to recover until viability reached >97% and doubling times stabilized, at which point the stable pools were frozen for later use. Clones were recovered from stable pools by FACS sorting of single cells in 96-well plates.Purification and product quality

[0114] High-throughput antibody purification was performed as described in Li et al., STAR Protocols 2022, 3 (2), 101428. Cultures were incubated overnight with ProA magnetic beads (GenScript Cat# L00695), with wash, recovery, and elution steps performed using Kingfisher Flex and Apex instruments (Thermo Fisher Scientific). ProA bound material was eluted in 100 mM Sodium acetate, pH 3.6, followed by immediate neutralization in 3M Tris- HC (pH 11). Product quality was then assessed by HPLC-based analytical size-exclusion chromatography (SEC) (Agilent). SEC data were analyzed using the Chromeleon10672-W001-SECChromatography Data System (Thermo Fisher Scientific), to assign pre / post main-peak (MP) fractions.Transcriptomics

[0115] Cell samples were collected from cultures on day 2 post-transfection of a transient batch expression. Total RNA was purified using TRIzol reagent (Invitrogen Cat# 15596026), and column purified using an RNeasy 96 kit (Qiagen Cat# 74181) with on-column DNA digestion using RNase-Free DNase I (Qiagen Cat# 79254). RNA sequencing libraries were constructed with the TruSeq Stranded RNA sample preparation kit (Illumina Cat# RS- 122-9004DOC). Transcript and gene level quantification was performed with Salmon (Patro et al., Nat Methods 2017, 14 (4), 417-419). Gene level counts were normalized and assessed for differential expression using the DESeq2 R package (Love et al., Genome Biology 2014, 15 (12), 550). Differentially expressed genes (DEGs) were classified as having a log2(fold-change from a vector control) > 0.5 and an adjusted p-value < 0.001. Genes were only included for analysis if at least one sample had a count of 20 or higher. Each molecule or control condition had 4 replicates with the exception of IgG-scFv [1, 2] which had one sample excluded for failing NGS quality control. A principal component analysis was performed using normalized gene counts for all DEGs as defined above. Gene set enrichment was performed using the gprofiler2 R package with a focus on Reactome and Gene Ontology Biological Process gene set databases.CRISPR knockout generation and evaluation

[0116] CRISPR knockouts were generated using Synthego’s Immortalized Cell Lipofection Protocol optimized for our cell line and culture conditions. RNP complexes consisting of CRISPR enzyme-2NLS (Synthego) and sgRNA knockout kit multi-guide pools and custom arrayed libraries (Synthego) were transfected into 6e5 viable cells in 1 mL of media with 2 uL of CRISRPMAX lipofectamine (Invitrogen Cat# CMAX00001). A CRISPR enzyme to sgRNA molar ratio of 1 : 1.6 was used for RNP complex formation. When adding cells to preformed RNP-lipofectamine complex, samples were immediately mixed by pipetting two times. For double KO generation, guides for each targeted gene were pooled and the amount of RNP complex was doubled, while keeping the amount of lipofectamine fixed due to this reagent’s negative impact on cell growth. One day after RNP transfection, an additional 1 mL of growth media was added, and cells were given four days to one week to recover prior to collecting cells for KO assessment and beginning a batch expression run. KO was assessed using Synthego’s ICE protocol, which involves PCR amplifying the targeted region and Sanger10672-W001-SEC sequencing across the region of expected insertion and deletion (indel) formation. Sanger sequence traces were deconvolved using Synthego’s ICE analysis tool to quantify indel and KO frequency and predict the most abundant indels. Relevant Statistical analysisAll statistical analysis was performed using the R statistical programming environment. Unless otherwise stated, point estimates and error ranges are means and bootstrapped 95% confidence intervals (CI).10672-W001-SECSEQUENCES10672-W001-SEC10672-W001-SEC10672-W001-SEC

Claims

10672-W001-SECCLAIMSWhat is claimed:

1. A modified mammalian cell comprising a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, HY0U1, INHBE, HIST2H2BF, CCDC30, AKNA, YARS2, KMT5C, PADI3, CALY, SREBF1, ISG20, SEC61A1, NCAN, ATF4, SH3BGR, DGAT2, SERPINA10, CITED2, HSPA1A, MAPK7, ABL2, or DKKl.

2. The modified mammalian cell of Claim 1, which comprises a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, HY0U1, INHBE, HIST2H2BF, CCDC30.

3. The modified mammalian cell of Claim 1 or Claim 2, comprising a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, or HY0U1.

4. The modified mammalian cell of any one of Claims 1-3, comprising a reduction of at least one of the following genes: HIST2H3C or INHBE.

5. The modified mammalian cell of any one of Claims 1-4, comprising a reduction of the HIST2H3C gene.

6. The modified mammalian cell of any one of Claims 1-4, comprising a reduction of the INHBE gene.

7. The modified mammalian cell of any one of Claims 1-3, comprising a reduction of both of the following genes: INHBE and HIST2H3C; HIST2H3C and HY0U1; HIST2H3C and HIST2H2BF; HIST2H3C and RRP12; HIST2H3C and CCDC30; or RRP12 and HIST2H2BF.

8. The modified mammalian cell of any one of Claims 1-7, which produces a protein at a higher titer compared to a mammalian cell which comprises at least one of the genes.

9. The modified mammalian cell of Claim 8, which produces a protein at a titer that is at least 10% more than the titer of the protein produced in a mammalian cell which comprises at least one of the genes.

10. The modified mammalian cell of any one of Claims 1-9, in which the gene or genes is reduced by CRISPR.10672-W001-SEC11. The modified mammalian cell of any one of Claims 1-9, in which the gene or genes is reduced by TALENs.

12. The modified mammalian cell of any one of Claims 1-11, which modified mammalian cell is a mammalian expression cell host.

13. The modified mammalian cell of any one of Claims 1-12, which modified mammalian cell is a HEK293 cell.

14. The modified mammalian cell of any one of Claims 1-12, which modified mammalian cell is a CHO cell.

15. A method of increasing titer of a protein, comprising producing the protein in the modified mammalian cell of any one of Claims 1-14.

16. A method of producing a protein, comprising expressing the protein in a modified mammalian cell, which modified mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, HY0U1, INHBE, HIST2H2BF, CCDC30, AKNA, YARS2, KMT5C, PADI3, CALY, SREBF1, ISG20, SEC61A1, NCAN, ATF4, SH3BGR, DGAT2, SERPINA10, CITED2, HSPA1A, MAPK7, ABL2, or DKK1.

17. The method of Claim 16, which mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, HY0U1, INHBE, HIST2H2BF, CCDC30.

18. The method of Claim 16 or Claim 17, which mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C, SCFD1, MDH2, RRP12, or HYOUl.

19. The method of any one of Claims 16-18, which mammalian cell comprises a reduction of at least one of the following genes: HIST2H3C or INHBE.

20. The method of any one of Claims 16-19, which mammalian cell comprises a reduction of the HIST2H3C gene.

21. The method of any one of Claims 16-19, which mammalian cell comprises a reduction of the INHBE gene.

22. The method of any one of Claims 16-18, which mammalian cell comprises a reduction of both of the following genes: INHBE and HIST2H3C; HIST2H3C and HY0U1; HIST2H3C and HIST2H2BF; HIST2H3C and RRP12; HIST2H3C and CCDC30; or RRP12 and HIST2H2BF.10672-W001-SEC23. The method of any one of Claims 16-22, which mammalian cell produces a protein at a higher titer compared to a mammalian cell which comprises at least one of the genes.

24. The method of any one of Claims 16-23, which mammalian cell produces a protein at a titer that is at least 10% more than the titer of the protein produced in a mammalian cell which comprises at least one of the genes.

25. The method of any one of Claims 16-24, in which the gene or genes is reduced by CRISPR.

26. The method of any one of Claims 16-24, in which the gene or genes is reduced by TALENs.

27. The method of any one of Claims 16-26, which modified mammalian cell is a mammalian expression cell host.

28. The method of any one of Claims 16-27, which modified mammalian cell is a HEK293 cell.

29. The method of any one of Claims 16-27, which modified mammalian cell is a CHO cell.

30. A protein obtainable by the method of any one of Claims 16-29.

Images