Neon^® Transfection System starter pack: ThermoFisher Scientific, cat # MPK5000.

Flow cytometer and general lab supplies include but are not limited to pipettes and cell culture flasks.

Geneious Prime 2022.2.2 (www.geneious.com) for gRNA design, FlowJo v10 (www.flowjo.com) for analysis of flow cytometry data. Figures were generated using a licensed version of Biorender. Graphs and statistical analysis were performed using GraphPad Prism v10.

B cells from the donor mice were mechanically extracted from the spleen and resuspended in washing buffer. Following erythrocyte depletion, B cells were enriched using the B cell isolation kit. Untouched B cells were then cultured for 24 hours in B cell stimulation medium, which is critical as RNP transfection does not work on resting cells. In our setup, 5 μg/mL of CpG was used to stimulate B cells as it does not necessarily differentiate the cells but rather puts them in cell cycle (27).

For adoptive transfer experiments, we used CD45.1 B1-8^lo and CD45.2 B1-8^hi mice, both obtained from M. Nussenzweig through F. Nimmerjahn (FAU, Erlangen), as well as CD45.1/CD45.2 heterozygous 33.C9gl mice ( Figure 1A ).

As donor mice, we used B1-8 mice, which have an immunoglobulin heavy chain knock-in derived from a 4-hydroxy-3-nitrophenyl (NP) hapten binding antibody with an affinity (K_a) of 5×10⁵ M⁻¹. A tryptophan to leucine substitution at codon 33 of the B1-8 heavy chain results in a tenfold increase in NP affinity (K_a = 5×10⁶ M⁻¹, referred to as B1-8^hi, Figure 1B ). Conversely, amino acid substitutions with glycine, threonine, glutamine, and threonine at position 24, 31, 35, and 98, respectively, decrease NP-binding by a factor of four (K_a = 1.25×10⁵ M⁻¹, denoted as B1-8^lo, Figure 1B ) (25, 28, 29). These two iterations of B1-8 mice, which we used as our donor B cell mice, were bred to be on different allelic variants of the pan-hematopoietic cell marker CD45: B1-8^lo on CD45.1 and B1-8^hi on CD45.2.

For B cells from B1-8 mice to bind the NP hapten in a subsequent immune response, they must pair with a lambda light chain during VDJ recombination. If they pair with a kappa light chain instead, NP binding is unlikely to occur (30). With this in consideration, we used an additional in-house mouse model, 33.C9gl, as recipient mice. 33.C9gl mice have a heavy chain and kappa light chain knock-in of an anti-DNA antibody derived from a systemic lupus erythematosus patient (26). Because these recipient mice have both a transgenic human heavy chain and a kappa light chain, they cannot mount adequate B cell responses to the NP hapten, justifying them as our NP non-responsive recipient mice.

Donor B cells from B1-8^lo and B1-8^hi mice were cultured for 24 hours in B cell stimulation medium, mixed in predetermined proportions, and adoptively transferred into 33.C9gl mice immunized with KLH with alum as an adjuvant a week before adoptive transfers. The KLH immunization serves as a T cell prime, which enables a much more robust T cell-mediated immune response upon NP-KLH immunization. After 24 hours, the recipient mice were immunized with 100 μg of alum-adjuvanted NP-KLH and then sacrificed on days 6, 9, or 14 to study the GC response ( Figure 2 ).

Our approach utilized the CRISPR toolbox, primarily consisting of two essential elements: CRISPR-associated (Cas9) proteins and a guide RNA (gRNA). The gRNA comprises two synthetic subcomponents: a trans-activating CRISPR RNA (tracrRNA) that is annealed to a CRISPR/targeting RNA (crRNA). Combining the Cas9 proteins with gRNAs generates RNPs that were then used to transfect B cells. A Neon Electroporation Device™ was used to electroporate B cells with the RNPs of predesigned crRNA. Electroporation efficiency was then determined by flow cytometry or next-generation amplicon sequencing ( Figures 3A, B ).

The in-silico design of the crRNA followed three rules. First, preference was given to crRNAs that target the first exons of the Coding Sequence (CDS). Secondly, the crRNAs should have relatively high on-target efficiency. Finally, they should have a low predicted off-target activity. Since RNPs are quickly degraded by cells, off-targets are not a concern in an RNP approach (31). Once appropriate crRNA sequences had been identified, they were commercially sourced. To reduce the cost of experiments, tracrRNA was bought in bulk.

To generate 100 μM gRNA, 10 nmol crRNA were resuspended in 50 μL nuclease-free duplex buffer to achieve a final concentration of 200 μM. Subsequently, 50 μL of 200 μM tracrRNA were added to the crRNA. The resulting mixture was denatured at 95°C for 5 min and then slowly cooled to facilitate annealing. As control experiments, a custom-design crRNA targeting an intron was also utilized ( Figure 3C ).

The transfection components were assembled as follows: Electrolysis Buffer E1 and Transfection Buffer T were allowed to reach room temperature. To create an RNP complex suitable for transfecting 2x10⁶ B cells, 186 pmol Cas9 (equivalent to 3 μL from a 62 μM Cas9 solution) were combined with 220 pmol pre-designed gRNA (equivalent to 2.2 μL from the 100 μM gRNA solution). 4.8 μL 1X IDTE Buffer was then added to this mixture and incubated at room temperature for 15 minutes to allow the formation of an RNP.

Next, stimulated B cells were washed with pre-warmed PBS. The cells were then counted and resuspended in Buffer T to achieve a final concentration of 40x10⁶ cells/mL. To 53 μL of Buffer T, 10 μL of the RNP complex, 5 μL of 100 μM Cas9 Electroporation Enhancer, and 50 μL of cells resuspended in Buffer T were added ( Table 1 ). This mixture was then incubated for 5 minutes at room temperature.

Using the Neon Electroporation Device, 100 μL of the B cell mix were electroporated using a single 20 ms, 1600 V pulse. The B cells were then immediately transferred to pre-warmed B cell stimulation medium and incubated at 37°C. Phenotyping or adoptive transfers were done after 24 or 48 hours.

Both flow cytometry and sequencing-based approaches were employed to assess the efficiency of CRISPR-based genome edits. Flow cytometry was performed to analyze surface molecules such as the MHC class I molecule or the Fas receptor. In scenarios where flow cytometry was impractical or the need for an additional confirmatory approach arose, sequencing-based approaches were also used.

Next-generation amplicon sequencing was employed to quantify gene editing efficiency in knock-out cells. This analysis utilized the Amplicon-EZ commercial service from Genewiz (Azenta Life Sciences). Following PCR amplification, the amplicons were normalized to a 20 ng/µL concentration and purified. Purified PCR products ranging from 150 to 500 base pairs containing partial Illumina^® adapter sequences were submitted for sequencing.

Additionally, a reference sequence and the sequences of primer binding sites were provided to facilitate mutation quantification through paired-end Illumina sequencing. Quality control, adapter sequence trimming, reference sequence alignment, and mutant sequence quantification were performed using NGS Genotyper v1.4.0 (Illumina Inc.).

Each experiment consisted of an experimental arm and a control arm. In the control arm, wild-type (WT) B1-8^lo B cells mixed with WT B1-8^hi B cells were adoptively transferred into the recipient mice. In the experimental arm, while maintaining the same ratio of B1-8^lo to B1-8^hi B cells, the gene-of-interest was knocked out in a proportion of B1-8^lo B cells. In the case of Fas knock-out experiments, the starting proportion of knock-out B cells was intentionally adjusted to around 40% by adding WT B1-8^lo B cells to the electroporated cell mix to enable the detection of an increase in the proportions of B1-8^lo CRISPR-edited GC B cells.

B cell mixtures for the control and experimental arm were stimulated for 48 hours, followed by adoptive transfer into KLH-primed 33.C9gl recipient mice. These recipient mice were immunized 24 hours later with alum-adjuvanted NP-KLH, and the GC response was subsequently analyzed on days 6, 9, and 14 post-immunization.

To assess the efficiency of knocking out Fas in the input B cells, the 40LB culture system was utilized to achieve a GC-like state since resting B cells and CpG-activated B cells do not express high levels of Fas. Aliquots of the input B cells from the experimental and control arm were co-cultured with 40LB cells for 5 days to enable a phenotypic quantification of Fas knock-out efficiency ( Figure 4 ). 1x10⁵ of the input B cells were co-cultured for 5 days with radiation-inactivated 40LB cells in B cell medium supplemented with 1 ng/mL IL-4. An in vitro time-series experiment was also performed to confirm that Fas^KO B cells had no competitive advantage or disadvantage over Fas^WT B cells ( Figure 5 ).

To confirm that Fas^KO and Fas^WT B cells had similar growth rates and that there was no in vitro selection of the Fas^KO B cells, a time-series experiment was performed. B cells were isolated from donor mice and cultured in B cell stimulation medium for 24 hours before CRISPR editing. The mix of Fas^KO and Fas^WT B cells were cultured for 24 hours in B cell stimulation medium before being co-cultured with 40LB cells. To determine the temporal effect of the gene knock-out in vitro, next-generation amplicon sequencing and flow cytometry was performed on days 3, 4, and 5. For next-generation amplicon sequencing, induced GC (iGCs) B cells were enriched using CD19 microbeads, and their DNA was subsequently extracted for sequencing ( Figure 5 ).

To assess and potentially optimize the efficiency of our electroporation, we first designed a crRNA targeting the disruption of the MHC class I molecule. MHC class I has been shown to have a relatively short half-life, which explains why the disruption very quickly reveals a phenotypic effect (32). Due to the high polymorphism observed in the MHC class I gene in humans and mice, we targeted the conserved Beta-2-Microglobulin (B2M) gene (33). The B2M gene encodes the B2M protein, which is necessary for cell surface expression of MHC class I and the stability of the peptide-binding groove (34). To target the B2M gene locus, we used a commercially available, pre-designed crRNA targeting the exon and a custom intron-targeting crRNA as a control. We then used flow cytometry to quantify the proportion of B cells that had no detectable amount of MHC class I molecules 48 hours post-electroporation with appropriate crRNAs. We used this experiment to optimize for electroporation conditions. A single 20 ms, 1600 V pulse was found to have the optimal cell viability with the highest knock-out efficiency. In the control electroporation, all B cells had detectable expression of MHC class I, while in the B2M targeted electroporation, 59% of the B cells did not have any detectable levels of MHC class I ( Figure 6 ).

To study GC dynamics, we developed an experimental approach in which we could adoptively transfer high-affinity and low-affinity B cells and put them in competition with each other in recipient mice. Two iterations of the classical, NP-reactive B1-8 mouse model were employed as donor mice: a high-affinity mouse model (B1-8^hi) and a low-affinity mouse model (B1-8^lo) (25). For recipient mice, we adapted an in-house 33.C9gl mouse model; any mouse model carrying a non-NP-reactive B cell receptor can also be utilized.

Briefly, 24-hour stimulated 3x10⁶ B1-8^lo B cells and 3x10³ B1-8^hi B cells were adoptively transferred into each alum-adjuvanted KLH immunized recipient mouse. 24 hours later, the recipient mice were immunized with 100 µg alum-adjuvanted NP-KLH and sacrificed 6 or 9 days later for an analysis of the GC response.

Despite being significantly outnumbered at a ratio of approximately 1:100 on the input day, B1-8^hi B cells could still outcompete the B1-8^lo B cells. By day 6, the number of B1-8^hi B cells had already begun to surpass the B1-8^lo B cells in the GC reaction. This competitive advantage continued to intensify, and by day 9, B1-8^hi B cells had almost completely overtaken the B1-8^lo in the GC ( Figure 7 ). It is important to note that the B1-8^lo B cells were always outcompeted by the B1-8^hi B cells for selection in the GC in multiple experiments that we conducted. We next sought to investigate what factors might mediate the observed selection in vivo.

Having successfully established an in vitro system of knocking out genes in murine B cells and an in vivo competition system of B cells of different affinities, we next sought to set up an adaptable in vivo CRISPR-Cas9 protocol that can be used to study the role of genes potentially involved in positive or negative selection in the context of a GC reaction. Since Fas is upregulated by B cells that participate in the GC reaction and its exact role in selection processes in GC cells is still not entirely clear (35), we used our experimental model to test the hypothesis that Fas plays a role in the negative selection of low-affinity B cells within the GC. Our experiment had two arms: a control arm and an experimental arm. In the experimental arm, Fas was knocked out from approximately 37.2% of the input B1-8^lo B cells, which were then mixed with B1-8^hi B cells before being adoptively transferred into alum-adjuvanted recipient mice. In the control arm, Fas^WT B1-8^lo and Fas^WT B1-8^hi B cells were mixed and then adoptively transferred into alum-adjuvanted recipient mice. In both arms, a ratio of B1-8^lo to B1-8^hi B cells of 1:6.5 was used, with a total of approximately 3x10⁶ B cells being adoptively transferred to each recipient mouse. To enable a precise determination of the proportion of Fas expressing input B cells by flow cytometry, 1x10⁵ of the input B cells were co-cultured with 40LB cells for 5 days ( Figure 8 ). This 40LB culture system leads to a significant upregulation of Fas, which enabled a precise determination of the Fas knock-out efficiency, especially in the experimental arm.

As determined by the 40LB culture system, 37.2% of the input B1-8^lo B cells in the experimental arm were Fas knock-out cells. On the other hand, B1-8^lo and B1-8^hi input donor B cells had similar Fas expression levels in the control arm ( Figure 8 ).

Upon adoptive transfer into recipient mice, Fas^KO B cells were still able to fully participate in the GC reaction, showing that our stimulation and electroporation protocol did not impair the physiological activity of the gene-edited B cells ( Figure 9A ).

Examining the total donor B cells in both the control and experimental arm, Fas^KO B1-8^lo B cells and Fas^WT B1-8^lo B cells were outcompeted by the B1-8^hi B cells, as the B1-8^lo/B1-8^hi ratio fell below 1 ( Figure 9B ).

Although the Fas^WT B1-8^hi B cells still outcompeted both Fas^KO B1-8^lo B cells and Fas^WT B1-8^lo B cells, there was a higher ratio of B1-8^lo/B1-8^hi in the experimental arm when compared to the control arm at days 9 and 14 ( Figure 9B ). Focusing on the B1-8^lo GC B cell population in the experimental arm, there was an increase in the proportions of Fas^KO B1-8^lo GC B cells from an input proportion of 37.2% to an average of 38% on day 6 to around 55% on day 9, which reached 67% by day 14 ( Figure 9C ). This indicated that a level of Fas-mediated selection was occurring within the GC.

To exclude an intrinsic effect of the gene editing on cell growth, we co-cultured Fas^KO B cells in vitro with 40LB cells to induce a GC-like phenotype in the B cells. Fas^WT B cells were co-cultured as a control with the 40LB cells ( Figure 10A ). We performed next-generation amplicon sequencing on the extracted DNA from the iGC B cells on days 3, 4, and 5 to check for mutations in the Fas gene locus that the crRNA had targeted. The frequency of mutations was then compared to the flow cytometry knock-out efficiency.

When comparing phenotypic data from flow cytometry and next-generation amplicon sequencing data, we found a strong concordance in the frequencies of knock-out iGC B cells ( Figure 10B ). Both methods confirmed that the proportion of Fas^KO B cells stayed stable over the observation period. In the control arm, the 5% of the background mutation, as determined by next-generation amplicon sequencing, was possibly due to sequencing or PCR errors.

This experiment demonstrates that within 40LB cultures, the loss of Fas expression does not confer a growth advantage or disadvantage to B cells, thus excluding the likelihood of an experimental artifact in our in vivo data.