Reprogramming the antigen specificity of B cells using genome-editing technologies

We have developed a method to introduce novel paratopes into the human antibody repertoire by modifying the immunoglobulin (Ig) genes of mature B cells directly using genome editing technologies. We used CRISPR-Cas9 in a homology directed repair strategy, to replace the heavy chain (HC) variable region in B cell lines with that from an HIV broadly neutralizing antibody (bnAb), PG9. Our strategy is designed to function in cells that have undergone VDJ recombination using any combination of variable (V), diversity (D) and joining (J) genes. The modified locus expresses PG9 HC which pairs with native light chains (LCs) resulting in the cell surface expression of HIV specific B cell receptors (BCRs). Endogenous activation-induced cytidine deaminase (AID) in engineered cells allowed for Ig class switching and generated BCR variants with improved HIV neutralizing activity. Thus, BCRs engineered in this way retain the genetic flexibility normally required for affinity maturation during adaptive immune responses. Peripheral blood derived primary B cells from three different donors were edited using this strategy. Engineered cells could bind the PG9 epitope and sequenced mRNA showed PG9 HC transcribed as several different isotypes after culture with CD40 ligand and IL-4.

HRs that would be universally present on either side of the VDJ region in polyclonal 101 human B cells. This is because each human B cell undergoes genomic 102 rearrangements during its genesis to assemble the HC VDJ gene from; 1 of at 103 (Ran et al., 2013b). Cutting activity was assessed for several different guide RNAs 135 targeting each of the three sites using the pCAG-eGxxFP recombination assay in 136 293 cells (Mashiko et al., 2013). Once the highest efficiency cutters were detected 137 using this assay (Figure 2-S1 suggesting the method may work universally (Kato et al., 2017). C108 SOSIP was 161 used to sort successfully engineered HIV-specific cells to produce enriched 162 subpopulations for further experiments. Genomic DNA extracted from these PG9-163 enriched cells was PCR amplified using primers that annealed upstream and 164 downstream of the expected insertion sites and outside of the donor DNA HRs 165 (out/out PCR, Figure 2B). Sanger sequencing of these products confirmed that the 166 new PG9 gene was grafted as expected between Cas9 cut sites within the IGHV 167 locus by HDR ( Figure 2F, Figures S6A-C). 168 169 Amplification-free whole genome sequencing of C108 SOSIP-selected cells 170 (engineered with either the V434 or V781 strategy) was performed using 171 fragmented gDNA from these lines using the Illumina HiSeq X. Reads were 172 mapped to the human reference genome hg19 (International Human Genome with the universal strategy, but not from the excluded allele which is not expressed 177 due to a chromosomal translocation of the IGHV region to chr8 in this Burkitt's 178 lymphoma line (Klein et al., 1975). Coverage depth analysis showed a deletion 179 between the V3-11 and V3-7 genes in the excluded allele (relative to the reference 180 genome), and a V-J recombination event in this allele between D3-9 and the J6 181   Figure 3A. Ig variable genes were then amplified from 245 cDNA using gene specific primers as previously described (Briney et al., 2016).  reads with identical UMIs (molecular identifier groups, MIGs), were constructed 287 using the previously referenced method. We analyzed only MIGs derived from >29 288 reads and eliminated unique sequences from the data set to reduce or eliminate 289 chimeric PCR product and any polymerase-introduced mutations from the 290 analysis. We compared data sets coming from two cell lines that had spent the 291 longest time in culture since the original transfection with VDJ gene editing 292 reagents (about 2.5 months). One of these underwent three consecutive selections 293 steps using MGRM8 SOSIP (M3), and one was simply passaged after the initial 294 C108 enrichment and experienced no further selection pressure for SOSIP binding 295 We then wanted to test the universal editing strategy in human primary polyclonal 305 B cells that have undergone a diversity of VDJ recombination events, use a variety 306 of different LCs, and which are readily available in the periphery. We changed the 307 system from one which uses the V7-81 promoter, to one which uses V3-74, (the V 308 gene immediately downstream of V7-81 in the IGHV locus, Figure 2A, Figure 4-309 S1). In engineered Ramos cells, PG9 expressed using the V3-74 promoter stained 310 more brightly with the HIV envelope probe than did those using V7-81, presumably Cells were mostly static for the first three days in culture but went through a burst 322 of division by day 5 under these conditions ( Figure 4A). We therefore chose to 323 engineer cells on day 4 of culture. We first tested AMAXA nucleofection of a 7.7Kb 324 GFP plasmid (HR210-PA) into cells at different concentrations ( Figure 4B). We 325 found that 2 or 3μgs of plasmid per million cells resulted in acceptable levels of 326 transfection allowing us to continue forward testing our engineering reagents in 327 plasmid format. We tested nucleofection of our three-plasmid system at different We also stained with a version of ZM233 SOSIP (bound to streptavidin-BV421) 343 with a deletion of the N160 glycan, which is a critical component of the PG9 HC 344 binding epitope, in order to gate-out cells binding to other epitopes on the SOSIP. 345 All three engineered donor samples had at least 10-fold more cells in the 346 APC+/BV421-/FITC+ 'PG9 chimeric BCR gate' than did non-engineered controls. 347 After a further 7 days in CD40L/IL-4 culture, cells were harvested, and mRNA was 348 purified ( Figure 4E). RT-PCR using PG9-specific primers amplified PG9 349 (confirmed by Sanger sequencing) both as IgM and IgG from engineered cell 350 mRNA but not from negative controls ( Figure 4F). To quantify PG9 transcripts, 351 RACE-PCR was also performed in order to amplify all transcribed HCs from 352 engineered cell samples (Turchaninova et al., 2016). NGS sequencing of these 353 amplicons showed PG9 mRNA-read frequencies similar to the PG9 chimeric BCR-354 supernatants were loaded overnight onto beads at 4°C. Beads were washed with 397 PBS (250ml) and eluted with 12ml 50mM citric acid buffer pH2.2 into 2ml 398 neutralization buffer (1M Tris pH 9.0). Eluted IgG was buffer exchanged into PBS 399 using 50KDa Vivaspin concentrators (Sigma-Aldrich). Each antibody sample was 400 concentrated to 500μl and purified by size exclusion chromatography (SEC) using 401 an S200 10/30 column (GE Healthcare) in PBS buffer and the 150KDa peak was 402 pooled and concentrated. IgG concentrations were measured by Nanodrop 403 (ThermoFisher) and stored at 4°C. Non-reducing SDS PAGE gels were run using 404 5μg of protein to confirm purity and quality of the IgG produced within one week of 405 purification. Three independently prepared batches of each antibody were 406 generated to confirm reproducibility. SEC profiles and SDS PAGE gels were 407 assessed from each batch to confirm reproducibility. SEC curves and SDS PAGE 408 gels from the third batch (representative of all) are given in Figure S1. 409

Biolayer Interferometry 410
Kinetic measurements were obtained with an Octet Red instrument immobilizing 411 IgGs on PBS hydrated anti-human IgG Fc sensors (Fortebio, Inc.). A new sensor 412 was used for each sample. PGT145 purified SOSIP trimers were prepared as 413 previous described (Voss et al., 2017) and used as free analytes in solution (PBS, 414 pH 7.4). Briefly, the biosensors were immersed in PBS-T, containing IgG at a 415 concentration of 10μg/ml for 2 min and at 1,000 rpm prior to the encounter with the 416 analyte. The SOSIP analytes were concentrated to 500nM in PBS-T. The IgG-417 immobilized sensor was in contact with the analyte in solution for 120s at 1,000 418 rpm and then removed from the analyte solution and placed into PBS for another 419 250s. These time intervals generated the association and dissociation binding 420 curves reported in this study. The experiment was repeated with three independent 421 Ig preparations to confirm reproducibility of results. Data from the final reproduction 422 were included in Figure 1-S2. Analysis was performed within one week of antibody 423 purification. 424

Polyreactivity assay: HEp-2 cell staining assay 425
The HEp-2 cell-staining assay was performed using kits purchased from Aesku 426 Diagnostics (Oakland, CA) according to manufacturer's instructions. These Aesku 427 slides use optimally fixed human epithelial (HEp-2) cells (ATCC) as substrate and 428 affinity purified, FITC-conjugated goat anti-human IgG for the detection. Briefly, 429 25μl of 100μg/ml mAb and controls were incubated on HEp-2 slides in a moist 430 chamber at room temperature for 30 min. Slides were then rinsed and submerged 431 in PBS and 25μl of FITC-conjugated goat anti-human IgG was immediately applied 432 to each well. Slides were incubated again for 30 min. and washed as above before 433 mounting on coverslips using the provided medium. diluted with media (to achieve greater than 100,000 RLUs after 48hr culture with 465 TZM-bls in the TCID50 Bright-glo assay), was mixed with 25μl of PG9 chimeric 466 antibody and incubated for 1hr at 37°C. Then 0.5x10 4 trypsinized TZM-bl cells in 467 50μls supplemented media and 20μg/ml dextran was added and further incubated 468 for 48hrs at 37°C. Neutralization was measured as above using the Bright-Glo 469 assay. In Figure 1, rather than reporting neutralization as the half maximal 470 inhibitory concentration (IC50s) from a titration of antibody with virus, % virus 471 neutralization using 10μg/ml of chimeric IgG was reported as a heat map (<10%-472 100% neutralization) using the average value from three separate experiments 473 using the different batches of antibodies. Select antibodies were also titrated in a 474 2-fold serial dilution with media starting at 100ug/ml and diluting down 6 wells to 475 obtain the neutralization as a function of Ig concentration (and IC50s derived by 476 fitting data points by non-linear regression using Prism). Antibodies were tested 477 within 1 week of purification. 478

Ramos (RA I) and (2G6) cells were obtained from ATCC (CRL-1596 and CRL-516
1923 respectively) and cultured as directed in IMDM media (Invitrogen) 517 supplemented with 10%FBS, L-glutamine and penicillin/streptomycin. Cells were 518 kept between 0.2 and 2 x10 6 cells/ml. Growing cells to high densities disrupted their 519 ability to be transfected. Optimal nucleofection parameters for Ramos RA1or 2G6 520 lines (from ATCC) were identified using a GFPmax (Lonza) plasmid as described 521 for the Neon transfection System (Life Technologies). Cells were recovered in 522 antibiotic free media and GFP expression was assessed using the Accuri analyzer 523 48h post nucleofection. Settings which gave the highest GFP nucleofection 524 efficiencies (1400V, 20ms, 2 pulses and 1600V, 20ms, 1 pulse for the RA1 and 525 2G6 subclone respectively), were used to nucleofect 10μg of HR110PA-1 PG9 526 donor DNA along with 2.5μg of each gRNA/Cas9 plasmid (pX330-U6-527 Chimeric_BB-CBh-hSpCas9) into 5x10 6 cells using the 100ul tip according to the 528 manufacturer's instructions. Cells were recovered in antibiotic free media and 529 grown for 72h (antibiotics were added back 24h after nucleofection). 530 for APC fluorophore were selected using the FACS ARIA III (BD Biosciences) at 540 the TSRI FACS core. Selection gates were made using non-engineered cell 541 controls incubated with the same probe. 542

Ramos gDNA sequence analysis 543
For analysis of PCR amplicons generated from genomic DNA: gDNA was isolated 544 from 3 x 10 6 cells using the AllPrep DNA/RNA Mini Kit (Qiagen) for use as template 545 in a PCR reaction using 3 forward and reverse primer sets specific for genomic 546 regions beyond the 5' and 3' homology regions found within the donor DNAs. 547 These primer sets were designed using the NCBI Primer BLAST server and are 548 listed in Table S1 (section I) where 'a' are forward and 'b' are the reverse 549 complement primers of a pair. The reaction was carried out using Phusion HF 550 Polymerase (NEB), 200ng template, 0.4μ M each primer, 200μM each dNTP in a 551 total volume of 100μl. After denaturing at 98 C for 30s 34 cycles were performed 552 at 98 C for 10s, 63 C for 30s, and then 72 C for 3.5 min. followed by a 30 min. hold 553 at 72 C. The 5.5kb product was purified on 1% agarose and the DNA extracted 554 using the QIAquick Gel Extraction Kit (Qiagen). The PCR product was sequenced 555 using Sanger sequencing (Eton Bioscience) with several primers such that the 556 complete 5.5kb sequence contig could be assembled (Figure 2-S3). The 557 sequencing primer sequences are listed in were aligned or nBLASTed to identify specific sequences). 571 To assess off-target editing events by PCR, genomic DNA was isolated from 4x10 6 572 cells using the AllPrep DNA/RNA Mini Kit (Qiagen) for use as template in a PCR 573 reaction. Primers 1-4 (Table S1.IV) were designed using the NCBI Primer BLAST 574 server. Three primers for each location were tested to select the best for 575 amplification. Primers 5-6 (Table S1.IV) are standard primers for the donor back 576 bone extended to match Tm of paired primers. The reaction was carried out using 577 Phusion HF Polymerase (NEB), 10ng template, 0.2μM each primer, 200μM each 578 dNTP in a total volume of 25μl. After denaturing at 98°C for 30s performed 34 579 cycles at 98°C for 10s, 63°C for 30s, then 72°C for 3min followed by a 30min hold 580 at 72°C. The products were visualized on 0.5% agarose. 581 582 Sanger Sequencing of Ramos Ig mRNA 583 To confirm the presence of PG9 mRNA in Ramos RA 1 engineered cells and to 584 detect isotype switching from PG9 IgM to IgG in engineered Ramos 2G6 cells, total 585 RNA was isolated from 3 x 10 6 pelleted cells using the AllPrep DNA/RNA Mini Kit 586 (Qiagen) as template for reverse transcription and amplification using the OneStep 587 RT-PCR Kit (Qiagen) with forward primers (Integrated DNA Technologies) specific 588 to the Ramos/PG9 variable regions and reverse primers specific to the IgM/IgG 589 constant regions (Table S1, section II). The reactions contained 400μM each 590 dNTP, 0.6μM each forward and reverse primer, 10ng RNA template, 5U RNasin 591 Plus (Promega) in a total volume of 50μl. The conditions were 50 C for 30min, 95 C 592 for 15min and then 30 cycles of 94 C for 30s, 58 C for 40s and 72 C for 60s 593 followed by an additional 10min at 72 C. Products were visualized on 1% agarose 594 and purified using the QIAquick PCR Purification Kit (Qiagen). The PCR products 595 were sequenced using Sanger sequencing with the same primers used for the 596 PCR (Eton Bioscience). 597

In vitro affinity maturation of Ramos lines 598
Ramos RA 1 cells engineered to replace the endogenous VDJ with PG9 VDJ using 599 the universal (V781) strategy and selected with C108 SOSIP in FACS was 600 passaged 8 times to allow for the introduction of mutations into the Ig variable 601 regions. Cells were titrated with biotinylated PGT145 purified WITO, MGRM8, 602 CRF-T250 or C108 SOSIP, APC-labeled streptavidin tetramers (described above). 603 Cells were incubated with a range of concentrations (3-0.0015 μg/ml SOSIP as 604 tetramer solution) for 45 minutes in FACS buffer and washed with PBS. APC+ 605 gates were set using non-engineered Ramos cells incubated with the highest 606 concentration of SOSIP probe (3μg/ml SOSIP as tetramer). Engineered cells in 607 the APC+ gate at each SOSIP concentration was plotted as a % of total cells 608 against the log of the probe concentration in μg/ml to calculate the effective 609 concentration require to stain 10% of cells (EC10) (Figure 3-S1A). MGRM8 or 610 WITO probes were incubated with either MGRM8 or WITO APC labeled tetramer 611 as previously described at their EC10 concentrations along with 1000x dilution of 612 anti-human lambda FITC-labeled antibody (Southern Biotech) for 45min. Cells 613 were washed and live single cells with the highest APC signal (top 5%) after 614 normalization for surface BCR levels (FITC) were selected for subsequent 615 expansion and further sorting with WITO or MGRM8 SOSIPs (an example of which 616 is shown in Figure 3-S2). This process was repeated twice more with EC10 617 concentrations for probes calculated before each sort. The starting C108 selected 618 engineered line was also continually passaged throughout the experiment for final 619 mRNA sequencing (PU). At the end of the experiment all cell lines were titrated 620 with C108, CRF-T250, WITO and MGRM8 probes as shown in Figure 3-S1B. 621 mRNA was harvested from cells after each sorting step and sequenced after RT-622 PCR/RACE-PCR using next generation sequencing (NGS). 623

Next Generation Sequencing of Ramos Ig mRNA 624
To characterize the selection of mutations in Ig heavy and light chain variable gene 625 regions in HIV Env SOSIP selected Ramos cells, mRNA was prepared from 3x10 6 626 cells (RNEasy kit, Qiagen) and eluted in 50uls of elution buffer. In a DNA clean 627 area, RT-PCR reactions were set up with these mRNA samples using Superscript 628 III (Invitrogen) and gene specific primers (listed in   (Table S1, section II). The reactions 731 contained 400μM each dNTP, 0.6μM each forward and reverse primer, 10ng RNA 732 template, 2.5U RNasin Plus (Promega), 0.5μl OneStep RT-PCR Enzyme Mix in a 733 total volume of 25μl. The conditions were 50 C for 30 min., 95 C for 15 min. then 734 30 cycles of 94 C for 30s, 67 C for 40s and 72 C for 60s followed by an additional 735 10 min. at 72 C. A second PCR was done to amplify the products. These reactions 736 contained 1μl 1 st PCR product, 2X HotStarTaq Plus Master Mix (Qiagen) and 737 0.5μM each forward and reverse primer in a total volume of 25μl. The conditions 738 were 95 C for 5 min. then 24 cycles of 94 C for 30s, 67 C for 30s and 68 C for 1 739 min. followed by an additional 20 min. at 68 C. Products were visualized on 1% 740 agarose. 741 742 RACE-PCR was performed using human heavy chain constant primers and a 743 'smart template switch adaptor' as described in (Turchaninova et al., 2016). 744 Libraries were pooled for Illumina sequencing adapter ligation and sequenced 745 using asymmetric 400x100 paired end reads on the MiSeq (Illumina) by the TSRI 746 sequencing core as described above for Ramos engineered HC AID mutation 747 analysis. Non-engineered and engineered samples from each of the three donors 748 were sequenced (3 million reads from each sample) and processed using 749 Migec/MiTools. The average MIG size from each dataset ranged from 3 to 11. We 750 accepted MIGs made from 3 or more reads for analysis. MIGs were processed 751 using Mixcr as described. CDRH3s were mapped to the PG9 CDRH3 reference 752 and these reads extracted for Figure 4G Isotype calls were made using 753 Mixcr/IMGT. 754 755