RESULTS

Introduction

B-cell therapy has great potential in targeting a variety of diseases. With engineered B-cells we could tackle chronic viral infections, such as HBV, prevent cognitive decline in Alzheimer’s disease, fight life-threatening infections with MRSA, or locally treat long-term cancer patients ^1,2. We wanted to discover the variety of opportunities B-cell engineering provides and thereby adapted our antibodies to several targets. We furthermore made progress in antibody engineering, by equipping our antibodies with blood-brain barrier transport shuttles for potential treatment of brain diseases, as well as introduction of additional scFvs (single chain variable fragments) providing additional target binding sites ³. Moreover, fusion of light and heavy chains in several of our constructs was supposed to prevent mispairing of endogenous light chains when integrated and expressed in the B-cell endogenous antibody locus. ⁴ With these approaches, we aimed to improve the novel concept of B-cell engineering and expand the fields of application for B-cell therapy.

Our constructs

Integrating our engineering approaches, we designed, synthesized and compared binding affinities of the different antibody constructs. The following tables provide an overview over the various targets, additional modifications, such as linkers, scFvs (single chain variable fragments) or blood-brain barrier shuttles of our antibodies.

1. Viral target: Hepatitis b surface protein
2. Bacterial target: α hemolysin
3. Cancer target: CAIX surface marker of renal cancer
4. Neurodegenerative disease: aβ plaques which occur in Alzheimer’s disease

As an example, we displayed the concept of our antibody construct design in the following figure. The shown constructs are based on the α-hemolysin targeting antibody Tosatoxumab. However, we adapted exactly this construct design for all of our antibodies, also directed against HBV or Renal Cell Carcinoma (CAIX).

Experimental workflow

1. Cloning of constructs

To implement our Proof of Concept in the lab, we followed a particular experimental workflow which is depicted in Figure 1. We started by amplifying antibody fragments by PCR and cloning of our constructs using Gibson assembly. The assembled plasmids were then transfected in E. coli, mini-prepped and analyzed via sequencing. Successfully cloned constructs were furthermore maxi-prepped in order to ensure the removal of endotoxins for transfection.

2. Transfection and transient antibody expression in Expi293

The successfully assembled plasmids were subsequently transiently transfected in Expi293 cells as depicted in Figure 2. After 5 to 7 days of incubation, the secreted antibodies were harvested from the cell supernatant and purified using HPLC.

3. Validating correct expression and determining binding affinity of our antibodies

After successful purification of our antibodies, we conducted an SDS-PAGE to validate the correct expression of the genetic constructs. As shown in the following figure, we also aimed to validate functionality of the expressed antibodies. For this purpose, we designed affinity assays for all our targets, HBV, MRSA as well as CAIX (Renal Cell Carcinoma). The blood-brain barrier transport of our integrated shuttles, melanotransferrin peptides or TfR binding sites, were supposed to be validated via a transcytosis assay.

4. Stable integration of antibody constructs into the B-cell genome

After identification of the most successful antibodies, we were aiming for stable integration into the B-cell genome. For this purpose, we designed homology arms as well as gRNAs targeting the endogenous B-cell locus as previously described (in the Engineering section). The engineered cells were then supposed to be analyzed by sequencing to validate correct incorporation of the construct. Furthermore, we were aiming for validation of correctly expressed BCRs against our targets via FACS.

Cloning

The cloning workflow was further followed as described above, using Gibson assembly. Overall almost all samples could be synthesized and sequenced (Table 1). Therefore, we were able to proceed with these constructs and transfect them for transient expression in Expi293 cells.

Antibody purification – HPLC

Following the successful cloning of 15 of our antibody constructs, we first validated the correct assembly by sequencing. These validated plasmids were then Maxi-prepped to remove bacterial endotoxins. In the next step, we transfected Expi293 cells with our assembled plasmids, aiming for transient protein expression. After 5 to 7 days of incubation, we harvested the cell supernatants and purified the secreted antibodies using HPLC. Based on the resulting elugrams, we selected several fractions containing the highest measured protein concentration.

These fractions containing the highest protein levels were then applied onto an SDS-PAGE to validate the size and thereby correct expression of our antibody constructs. In the following Table 2, expected molecular weights of the purified antibodies as well as their targets, such as α-hemolysin (MRSA), CAIX (Marker for Renal Cell Carcinoma) or the HBV surface antigen (HBsAg) are displayed. Besides the variety of target antigens, these antibodies were further engineered: several of our constructs carried blood-brain barrier transport shuttles such as a binding site for the transferrin receptor or an additional melanotransferrin-ligand. We furthermore introduced different Glycin-Serine-based linkers connecting light and heavy chains to prevent mispairing with the endogenous light chains.

In our research endeavors, we achieved the successful expression and purification of ten distinct antibody constructs. To validate the effectiveness of the prior antibody purification, we conducted SDS-PAGE analysis on the corresponding HPLC fractions under both reducing and non-reducing conditions (refer to Fig. 1 - 5).

Initially, we purified the antibodies denoted as A-B, U-V, and P-Q, each resulting from the transient expression of light and heavy chains from separate plasmids (Fig. 1-3). As expected, the reducing SDS-PAGE revealed well-defined protein bands at approximately 22 to 23 kDa, corresponding to the antibodies’ light chains. Additionally, we observed a band at around 50 kDa for each construct, which corresponded to the heavy chains. Notably, the band intensity was quite comparable between light and heavy chains included in each sample. Thus, the chains were expressed and secreted in equimolar amounts, as anticipated in antibody synthesis.

In contrast, the SDS-PAGE performed under non-reducing conditions showed prominent protein bands at about 250 kDa, which exceeded the expected bands of approximately 150 kDa by far. However, the non-reducing environment preserves the antibodies’ disulfide bonds, which in turn partially preserves their 3D structure. This is why the proteins’ migratory behavior deviates from that of linear proteins. Thus, both reducing and non-reducing SDS-PAGE conditions confirm the successful expression of the antibodies A-B, P-Q and U-V (Fig. 1-3).

Purification of the antibody constructs K, F and I yielded similar results. As shown in Fig. 4, well-defined protein bands at the expected height of 25 kDa and 50 kDa were observed in the purified fractions of antibody K under reducing conditions. These certainly correspond to the antibody’s light and heavy chain. In contrast, heavy and light chains of construct F were connected by a GS-linker and therefore exhibit a larger weight of approximately 75 kDa. In the purified fractions of antibody I, a construct composed of an scFv-fragment linked to the IgG heavy chain, a protein band at approximately 62.5 kDa could be observed. Under non-reducing conditions, the proteins deviated in their migratory behavior due to disulfide bond linkage. Thus, antibody constructs K, F and I exhibited the expected protein sizes in the SDS-PAGE, demonstrating successful purification.

Equally successful was the purification of the Guselkumab antibody (our negative control) as well as of the constructs C and L (Fig. 5). Remarkably, the SDS-PAGE on the one hand showed presence of the light chains in the correct size of approximately 22 to 25 kDa. On the other hand, the bands corresponding to the antibodies’ heavy chains could be detected at twice the height (approximately 100 kDa) we would have anticipated. This can be explained by an inefficient reduction of disulfide bonds through β-mercaptoethanol treatment: Presumably, only the single light chain bonds were reduced, while the double bonds of the heavy chains stayed intact. Corresponding to the non-reducing band pattern, we conclude that expression and purification of Guselkumab, as well as antibody C and L was successful as well.

As we could already detect in the varying band intensity of the conducted SDS-PAGEs, the protein quantity and therefore efficiency of purification varied between the samples. These differing yields can on the one hand be explained by initially differing transfection efficiencies of the Expi293 cells, resulting in overall lower protein expression. On the other hand, as shown in the wash fraction of Fig. 1, a considerable amount of protein did not bind to the Protein A column used for purification. Therefore, buffer conditions as well as the column volume could be refined in following experiments to increase the antibody yield from purification.

Affinity assays

Besides validating the correct expression of our antibody constructs in Expi293 cells, we also aimed to determine functionality of the produced antibodies. To achieve this, several assays were established to determine antibody binding affinity to the different targets. To analyze binding to HBsAg (the HBV surface antigen), to CAIX (a surface marker on Renal Cell Carcinoma) as well as to α-hemolysin (exotoxin secreted by MRSA). In this context, we also aimed to compare how engineering of the antibodies, for example through linkers between light and heavy chains or through attached scFv fragments, influences their binding capacity.

HBV ELISA

To assess the binding affinity of the antibody construct U-V for the HBV surface antigen HBsAg, we established an ELISA. To differentiate the specificity of the antibody interaction from nonspecific associations, we compared the binding of a negative control, the antibody Guselkumab, with the U-V antibody-mediated binding to HBsAg (Fig. 6).

In the course of this ELISA experiment, the binding affinity of the U-V antibody to the HBV surface antigen was notably higher compared to our negative control. Nevertheless, we observed some degree of unspecific association with HBsAg, prompting us to repeat the ELISA twice to enhance the statistical significance of our findings. Despite the unspecific binding observed with the negative control, the U-V antibody exhibited a significant increase in affinity. Based on these results, we can conclude that the expressed and purified U-V antibody is indeed fully functional.

Neutralization assay

In order to validate the affinity of our antibody constructs directed against α-hemolysin, the MRSA exotoxin, a neutralization assay was performed with the antibody constructs F, I, C, L as well as A-B (Fig. 7). As control for specific binding served the antibody Guselkumab, directed against IL 23. For this assay, antibody solutions (concentration of 0.2 mg/mL) as well as harvested S. aureus supernatant were added in five different ratios to blood agar plates (90:10, 50:50, 10:90, 1:99, 0.1:99.9). Since halos indicate cell lysis, successful neutralization of the exotoxin through our antibodies should not induce halo formation surrounding the corresponding sample. As shown in Fig. X, the antibody constructs F, I, L and A-B, applied in ratios of 50:50 (antibody: S. aureus supernatant) showed a reduced halo in comparison to ratios containing higher amounts of S. aureus supernatant. This indicates a functional binding affinity to the S. aureus toxin and could thereby underline that our engineered antibody constructs still show functionality despite additional modifications: for example constructs modified with linkers between light and heavy chains, such as L, I and F, or antibody constructs containing a melanotransferrin-peptide for blood-brain barrier transport, such as L. Since the concentration of α-hemolysin is not defined in the S. aureus supernatant, we can unfortunately not directly quantify whether our modifications influence the strength of binding affinity.

However, focusing on the sample applied in a 90:10 ratio (antibody: S. aureus supernatant), a halo is clearly visible. In case of present functional binding sites in our antibody constructs, higher antibody levels should also correspond to a lower level of neutralization. Possibly, antibody concentrations used in excess might negatively influence the cells’ viability. Nevertheless, the neutralization assay including the negative control directed against IL 23, Guselkumab, also shows a reduced halo formation surrounding the 50:50 ratio sample. Therefore, one can presume that the IL 23-directed control antibody, Guselkumab, has an unspecific binding affinity to α-hemolysin and thereby also prevents cell lysis. Corresponding to the conflicting results of the applied 90:10 ratio (antibody: S. aureus supernatant), one might also presume that differences in sample handling or an improperly carried out neutralization assay caused the observed results. Reasons for these inconsistent findings could for example lie in old blood agar plates, incompatible antibody concentrations or varying exotoxin concentrations in the S. aureus supernatant. Therefore, we could not determine for certain whether our antibodies are functional and show proper binding affinity to the MRSA-exotoxin.

Conclusion

In conclusion, we successfully designed and cloned a variety of different antibody constructs, equipped with different linkers, blood-brain barrier transport systems and additional scFvs. We furthermore successfully transiently transfected and expressed our designed antibody constructs in Expi293 cells, which were then purified via HPLC. For several antibodies, we could also validate intact binding to their corresponding targets through HBV-ELISA as well as a neutralization assay. However, several experiments may indicate functionality but need to be repeated due to conflicting results. In the following sections, we provide a detailed outlook on the opportunities to expand our experimental procedures.

Outlook

Determining binding affinity of CAIX-targeting antibodies

Besides tackling chronic viral and bacterial infections as well as brain diseases, our goal was to establish B-cell engineering for cancer targets. For our proof of concept we therefore engineered and expressed antibodies targeting Carbonic Anhydrase IX (CAIX), a common surface marker for Renal Cell Carcinoma (RCC). ^5,6 To assess the functionality of our CAIX-targeting antibodies, such as P-Q, we aimed to analyze their affinity for human Caki-1 cells, which naturally express CAIX on their surface. [^7]^,7 The assay involved incubating Caki-1 cells with our antibodies, followed by staining with fluorescently-labeled secondary anti-human IgG antibodies. This would have enabled us to quantify fluorescence emissions in samples treated with anti-CAIX antibodies using flow cytometry analysis and to compare the affinity in between differently engineered constructs.

Validating blood-brain barrier transport via the transferrin receptor

To address diseases affecting the brain, such as Alzheimer’s disease, we engineered antibodies equipped with a blood-brain barrier shuttle. Several of our antibody constructs featured additional single-chain variable fragments (scFvs) with affinity for the transferrin receptor (TfR) or melanotransferrin, which is also a TfR ligand. Notably, we successfully expressed and purified several of these antibodies, including constructs K and L. Our next step would be to confirm their ability to bind to TfR, potentially facilitating blood-brain barrier transport through a transcytosis mechanism. For this purpose, we partnered with Ibidi, who provided a unique transcytosis pump system. This system would have allowed us to establish a controlled flow of antibody solutions across a HUVEC-coated cell barrier, enabling us to assess the efficiency of our antibodies’ transport systems. ⁸

Stable integration of our antibody constructs into the B-cell genome

After analyzing the correct expression and target affinity of our antibodies, we wanted to select the most promising candidates. Subsequently, we were aiming for stable integration of these selected antibodies into the B-cell genome using CRISPR/Cas9. Following the approach of one of the pioneers in B-cell engineering, Adi Barzel ^1,2, we were planning to integrate the antibody cassettes into the endogenous IgH locus, thereby preserving endogenous regulatory mechanisms and class switching. For this purpose, we already designed suitable homology templates and gRNAs for transfection into a RAMOS B-cell line. These meditate on the one hand specific genomic integration of our antibody constructs and on the other hand, terminate endogenous IgH expression upon integration. Subsequently, our goal was to check stable incorporation in the cell via sequencing and to clonally select and expand clones with the corresponding BCRs via FACS analysis.

Introduction of a kill-switch to make B-cell engineering a safe therapy

In order to ensure safe application of B-cell therapy in the patient, we were aiming to integrate a caspase 9-inducible kill switch into our B-cells. Upon binding of a small molecule dimerizer, caspase 9, fused to a FK506 binding protein (FKBP) dimerizes and promptly induces apoptosis. This kill switch could have been efficiently expressed in B cells without impairing phenotype, function or antigen specificity of the cell. ⁹ Our goal was to transiently transfect RAMOS B-cells with plasmids carrying this caspase-inducible kill switch and prove efficient removal of the cells upon stimulation. With integrating this safety measure, we wanted to move one step closer to clinical implementation.

Outlook: scFv activational switch

In addition to terminating antibody expression, our objective was to implement a more sophisticated level of control over our engineered B-cells. We aimed to actively induce the secretion of therapeutic antibodies. To achieve this, we devised an innovative approach by designing single-chain variable fragment (scFv) constructs with an affinity for mCherry. These scFv constructs were intracellularly linked to the B-cell receptor (BCR) signaling subunit. Our plan was to introduce this BCR-scFv cassette into a human safe harbor locus, such as AAVS1, utilizing CRISPR/Cas9 technology. Upon stimulation of the scFv domain with mCherry, the downstream BCR signaling cascade would be initiated, resulting in the specific expression of antibodies by the B-cell. This pioneering technique held exciting possibilities: it allowed a single B-cell to be triggered by two entirely different stimuli, thereby enabling adaptability to various markers of a specific disease. Furthermore, in a clinical context, scFvs responsive to small molecule drugs could provide the means for targeted antibody secretion induction, allowing precise fine-tuning of therapeutic concentrations within the patient’s body.

[7]: PubMed Central (PMC). Engineering an Anti-Transferrin Receptor ScFv for pH-Sensitive Binding Leads to Increased Intracellular Accumulation; 2023 [cited 2023 October 12] Available from: URL: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4694649/

Footnotes

1. Nahmad AD, Raviv Y, Horovitz-Fried M, et al. Engineered B cells expressing an anti-HIV antibody enable memory retention, isotype switching and clonal expansion. Nature communications 2020; 11(1): 5851 [https://doi.org/10.1038/s41467-020-19649-1][PMID: 33203857] ↩ ↩²

2. Nahmad AD, Lazzarotto CR, Zelikson N, et al. In vivo engineered B cells secrete high titers of broadly neutralizing anti-HIV antibodies in mice. Nat Biotechnol 2022; 40(8): 1241–9 [https://doi.org/10.1038/s41587-022-01328-9][PMID: 35681059] ↩ ↩²

3. Bethune MT, Gee MH, Bunse M, et al. Domain-swapped T cell receptors improve the safety of TCR gene therapy. eLife 2016; 5 [https://doi.org/10.7554/eLife.19095][PMID: 27823582] ↩

4. Pasha M, Sivaraman SK, Frantz R, Agouni A, Munusamy S. Metformin Induces Different Responses in Clear Cell Renal Cell Carcinoma Caki Cell Lines. Biomolecules 2019; 9(3) [https://doi.org/10.3390/biom9030113][PMID: 30909494] ↩

5. Chen Z, Ai L, Mboge MY, et al. Differential expression and function of CAIX and CAXII in breast cancer: A comparison between tumorgraft models and cells. PLoS ONE 2018; 13(7): e0199476 [https://doi.org/10.1371/journal.pone.0199476][PMID: 29965974] ↩

6. Chen Y-W, Rini BI, Beckermann KE. Emerging Targets in Clear Cell Renal Cell Carcinoma. Cancers 2022; 14(19) [https://doi.org/10.3390/cancers14194843][PMID: 36230766] ↩

7. Im K, Mareninov S, Diaz MFP, Yong WH. An Introduction to Performing Immunofluorescence Staining. Methods in molecular biology (Clifton, N.J.) 2019; 1897: 299–311 [https://doi.org/10.1007/978-1-4939-8935-5_26][PMID: 30539454] ↩

8. Experimental Workflow of a Flow Assay With Cells | ibidi; 2023 [cited 2023 October 12] Available from: URL: https://ibidi.com/content/292-experimental-workflow. ↩

9. Straathof KC, Pulè MA, Yotnda P, et al. An inducible caspase 9 safety switch for T-cell therapy. Blood 2005; 105(11): 4247–54 [https://doi.org/10.1182/blood-2004-11-4564][PMID: 15728125] ↩