Alternative strategies are needed for patients with B-cell malignancy relapsing after CD19-targeted immunotherapy. Here, cell surface proteomics revealed CD72 as an optimal target for poor-prognosis KMT2A/MLL1-rearranged (MLLr) B-cell acute lymphoblastic leukemia (B-ALL), which we further found to be expressed in other B-cell malignancies. Using a recently described, fully in vitro system, we selected synthetic CD72-specific nanobodies, incorporated them into chimeric antigen receptors (CAR), and demonstrated robust activity against B-cell malignancy models, including CD19 loss. Taking advantage of the role of CD72 in inhibiting B-cell receptor signaling, we found that SHIP1 inhibition increased CD72 surface density. We establish that CD72-nanobody CAR-T cells are a promising therapy for MLLr B-ALL.
Patients with MLLr B-ALL have poor prognoses despite recent immunotherapy advances. Here, surface proteomics identifies CD72 as being enriched on MLLr B-ALL but also widely expressed across B-cell cancers. We show that a recently described, fully in vitro nanobody platform generates binders highly active in CAR-T cells and demonstrate its broad applicability for immunotherapy development.
This article is highlighted in the In This Issue feature, p. 1861
Surface antigen-targeted immunotherapies have shown great promise for treating B-cell malignancies, particularly CD19 chimeric antigen receptor T (CAR-T) cells for B-cell acute lymphoblastic leukemia (B-ALL) and diffuse large B-cell lymphoma (DLBCL; ref. 1). However, despite impressive initial responses, many patients relapse, often with loss of CD19 antigen (2). Although second-line therapies targeting alternative cell surface proteins, such as CD22, have shown efficacy in treating CD19-negative relapses, “antigen escape” has also been observed (3). In this context, several additional B cell–restricted receptors have emerged as potential targets, including CD20 (4), CD23 (5), CD79b (6), CD37 (7), and BAFFR (8). However, it remains unclear which is the optimal target after CD19 failure. Discovery of additional targets may further enable new strategies to overcome resistance and engineer novel approaches for cellular therapy.
Thus far, patients with B-ALL have shown the most promising responses to CD19 CAR-T therapy (9). However, both pediatric and adult patients with B-ALL who harbor translocations of the mixed-lineage leukemia (MLL1/KMT2A) gene (MLLr) continue to demonstrate the poorest outcomes, either in the context of standard chemotherapies (10) or, in emerging data, after CAR-T therapy (11–13). Here, our initial goal was discovering new therapeutic targets particularly enriched in MLLr patients, as immunotherapeutic efficacy frequently correlates with surface antigen density (14). Although most new immunotherapy targets are found by transcriptome analysis, RNA-based data may not accurately predict surface protein abundance (15). We hypothesized that unbiased quantification of plasma membrane protein density would be most fruitful for uncovering novel targets. Furthermore, although the “surfaceome” comprises ∼2% of the total proteome by abundance, it plays key roles in determining cellular fate by mediating external signaling, governing microenvironment interactions, driving cellular migration, and modulating immune surveillance.
Using these methods, we identified CD72, an ITIM-bearing inhibitor of B-cell receptor (BCR) signaling, as being highly expressed on MLLr B-ALL versus other B-ALL genotypes. Further investigation also demonstrated consistent CD72 expression in DLBCL. To our knowledge, this antigen has not previously been targeted with cellular therapies. We utilized a recently described, fully in vitro nanobody yeast display library (16) to develop highly specific binders to CD72. We showed for the first time, to the best of our knowledge, that CAR-T cells developed from these nanobodies, generated without llama immunization, have potent efficacy both in vitro and in vivo. Finally, we demonstrated that modulating BCR signaling via small molecules can upregulate surface CD72. Based on these results, we propose targeting CD72 as an effective genotype-driven immunotherapy strategy for patients with MLLr B-ALL. We further validate in vitro nanobody selection as a versatile platform for cellular therapy development.
Cell Surface Proteomics of MLLr versus Other B-ALL Subtypes Define Distinct Surfaceome Signatures
To define the B-ALL cell surfaceome, we enriched N-glycoproteins using a modified version of the Cell Surface Capture method (Fig.1A and Methods) followed by quantitative mass spectrometry (MS; ref. 17). As this method requires sample input of 30 to 200 million cells, it is not routinely amenable to primary sample analysis; therefore, we performed our analyses on cell lines. We profiled eight B-ALL lines with distinct driver translocations including MLL–AF4 (n = 3), MLL–ENL (n = 1), BCR–ABL (n = 3), and ETV6–RUNX1 (n = 1), plus Epstein–Barr virus (EBV)–immortalized B cells derived from normal donor umbilical cord blood as a nonmalignant comparator (Supplementary Table S1), all performed in biological triplicate. Using label-free quantification (LFQ) in MaxQuant (18) and filtering for Uniprot-annotated membrane or membrane-associated proteins, we quantified 799 membrane proteins (Supplementary Fig. S1A). Biological replicates demonstrated excellent reproducibility (Supplementary Fig. S1B). Using a 2-fold cutoff and P < 0.05 derived from a Welch t test between MLLr and non-MLLr cell lines, we identified 25 unique membrane proteins specifically enriched on the MLLr surfaceome with 39 downregulated (Fig.1B, Supplementary Fig. S1C, and Supplementary Data File S1). As positive controls, our analysis identified known hallmarks of MLLr including PROM1 and FLT3 upregulation as well as loss of CD10 (19). Principal component analysis showed marked separation of MLLr B-ALL lines from BCR–ABL B-ALL and EBV-immortalized B cells, implying a distinct cell surfaceome (Fig.1C). Gene Ontology analysis of differentially expressed membrane proteins indicated enrichment of cell adhesion proteins while interestingly showing a marked decrease of MHC class I and II receptors (Supplementary Fig. S1D).
To investigate surface protein regulation, we performed parallel RNA sequencing (RNA-seq). We found a modest correlation between upregulated surface-annotated proteins found by both RNA-seq and surface proteomics, consistent with prior studies (Fig.1D, Supplementary Fig. S1E and S1F, and Supplementary Data File S2; ref. 20). We did further note that several adhesion molecules, such as NCAM1, L1CAM, and ITGAV, showed significant upregulation only at the protein level. Similar adhesion proteins have been proposed to play a major role in B-ALL proliferation within the bone marrow niche (21). Furthermore, MHC downregulation is not prominently noted in RNA-seq data. These results reinforce that transcriptome-only analyses may miss biologically and clinically relevant findings.
CRISPRi Screen of the MLLr Cell Surfaceome to Reveal Potential Vulnerabilities
We reasoned that cell surface receptors relevant to the growth and survival of MLLr B-ALL may serve as promising immunotherapy targets. We therefore performed a CRISPR-based interference functional genomic screen (22) of the MLLr cell line SEM, using a single-guide RNA (sgRNA) library targeting 5973 genes (5 sgRNAs/gene) encoding all Uniprot-annotated membrane-spanning proteins (ref. 20; Supplementary Methods and Supplementary Fig. S2A). We found 60 membrane proteins for which knockdown resulted in significant growth disadvantage after 30 days (>12 doublings; Supplementary Fig. S2B). However, the only protein that showed specific upregulation on MLLr B-ALL, as well as significant genetic dependence, was FLT3 (Supplementary Fig. S2C). Although FLT3 is being investigated by some groups as a potential immunotherapy target, this protein has marked therapeutic liabilities due to high expression on hematopoietic stem and progenitor cells (HSPC; ref. 23). Therefore, we explored alternative avenues to find novel targets with more favorable characteristics.
Triage of MLLr Cell Surface Proteins Identifies CD72 as an Optimal Immunotherapy Target
We next bioinformatically triaged our cell surface markers. We sequentially considered upregulated cell surface markers in MLLr versus other (63/799 proteins); relatively abundant proteins to find markers with high antigen density (LFQ intensity > 25; 27/799 proteins); and single-pass membrane proteins to facilitate development of in vitro antibodies (17/799 proteins; Fig.2A). To avoid “on-target, off-tumor” toxicity, we eliminated protein-encoding genes with a median transcript per million (TPM) > 10 in normal tissue (excluding spleen) per the Genotype-Tissue Expression (GTex) database, or any detectable immunohistochemical staining in non-hematopoietic tissues per the Human Protein Atlas (24). This left eight of 799 proteins (Supplementary Fig. S3A). Finally, to avoid sensitive hematopoietic compartments, we eliminated proteins with any detectable RNA expression in CD34+ stem and progenitor cells (HSPCs) or T cells in the DMAP resource (25) and Human Blood Atlas (Supplementary Fig. S3B and S3C; ref. 26). After completing this triage, one membrane protein target stood out as best fulfilling our criteria: CD72.
Reanalysis of multiple cell line and patient sample transcriptome datasets also confirmed that CD72 is upregulated at the transcript level in MLLr versus other B-ALL subtypes (Supplementary Fig. S3D–S3F). In addition, our own RNA-seq data suggested an ∼4-fold upregulation of CD72 in MLLr versus other (Supplementary Fig. S3G), nearly identical to that found in our surface proteomic screen. This integrated transcriptome–proteome analysis therefore suggested that CD72 surface expression is primarily governed by transcriptional control.
CD72, also known as Lyb-2 in murine biology, is a single-pass type II membrane protein with an extracellular C-type lectin domain and cytoplasmic ITIM motifs. The ITIM motifs on CD72, similar to CD22, serve as scaffolds for inhibitory phosphatases to counteract BCR signaling (27). These proteins, as well as CD19, demonstrate highly similar expression patterns across hematopoietic cell types per the Human Blood Atlas (Supplementary Fig. S4A), beginning expression early in B-cell development but becoming lost on plasmablasts and plasma cells (Supplementary Fig. S4B). Genetic ablation of CD72 in mice demonstrated no lethality but some increase in immune system activation (28). Although CD72 knockdown did not affect SEM growth in vitro, the same was true of well-validated targets CD19 and CD22. Taken together, our results suggest that CD72 could be a promising immunotherapeutic target for MLLr B-ALL.
CD72 Is Expressed Not Only in MLLr B-ALL but Also across B-cell Malignancies
A literature search revealed one prior report from 1992 investigating CD72 as a potential biomarker for B-cell malignancies (29). Supporting our results, the authors found positivity by immunohistochemistry (IHC) of CD72 in both MLLr B-ALL patient samples tested. However, they also noted that essentially all B-ALL and B-cell lymphoma samples examined, with the exception of plasmablastic lymphoma, also showed CD72 expression. The Cancer Cell Line Encyclopedia (CCLE) confirmed CD72 to be highly expressed in B-cell leukemia and lymphoma cell lines (Supplementary Fig. S4C). These findings suggested that targeting CD72 may find broader utility beyond MLLr B-ALL.
To independently verify these results, we examined CD72 surface expression on B-ALL patient-derived xenografts (PDX) from the Public Repository for Xenografts (PRoXe) biobank (30) and viably frozen primary pediatric samples from our institution (Supplementary Tables S2 and S3). By quantitative flow cytometry, we found CD72 to be expressed at several thousand copies per cell in MLLr PDX samples, similar to and sometimes greater than CD19 (Fig.2B). Primary sample analysis suggested higher CD72 in MLLr cells than non-MLLr cells but, notably, revealed CD72 expression even in non-MLLr disease (Fig.2C). IHC on fixed adult B-ALL bone marrow aspirate found uniformly high CD72 on MLLr B-ALL blasts compared with variable, but still present, expression in other genomic subtypes (Fig.2D and E). We also examined peripheral blood samples from patients undergoing autologous stem cell harvest and confirmed no evidence of detectable CD72 expression on CD34+ HSPCs (Supplementary Fig. S4D). Analysis of the Multiple Myeloma Research Foundation CoMMpass dataset (research.themmrf.org) confirmed no detectable CD72 mRNA expression in malignant plasma cells (Supplementary Fig. S4E).
We next validated CD72 expression in DLBCL. We analyzed several publicly available DLBCL patient transcriptome datasets (31) and found that CD72 mRNA was highly expressed, similar to levels of CD19 and generally higher than CD22 (Fig.3A). This analysis suggested that poorer prognosis activated B-cell (ABC) patients may have increased CD72 expression versus germinal center B cells (Fig.3B). Whereas CD72 IHC on an internal cohort of lymph node biopsies did not confirm this hypothesis, we did verify that 93% of the biopsies (26 of 28) showed detectable CD72 expression (Fig.3C and D).
Because our combined proteomic and RNA-seq results suggested that transcription is the primary determinant of CD72 antigen density, we sought to identify potential epigenetic regulators of CD72 by analyzing publicly available datasets (Supplementary Methods; refs. 32, 33). Combining ENCODE chromatin immunoprecipitation sequencing (ChIP-seq) data with Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) transcription factor motif analysis suggested that central B cell–identity transcription factors, including PAX5 and EBF1, have direct roles as regulators of CD72 (Supplementary Fig. S4F–S4J and Supplementary Data File S3). Overall, these results support the notion that CD72 is largely restricted to the B-cell compartment, further underscoring its promise as an immunotherapy target.
Development of a CD72-Directed CAR-T with Nanobody Yeast Display
To generate CD72-specific binding reagents for use in CAR-T cells, we employed a recently developed, fully in vitro nanobody yeast display screening platform (ref. 16; Fig.4A). Nanobodies are variable heavy-chain–only immunoglobulins derived from camelids that, due to their simple format, small size, and highly modular nature, are finding increasing utility in therapeutic applications. Although others have recently shown the potential of nanobodies for CAR-T development (34), to our knowledge all such binders have been generated via llama immunization, a slow, expensive process with limited availability. This fully synthetic yeast display library, which is widely available for academic use, was initially designed for enabling structural biology studies. Here, we demonstrate that this ready-to-use, low-cost in vitro approach can also generate binders efficacious in immunotherapy applications.
We expressed in mammalian cells a recombinant fusion protein comprised of the C-terminal extracellular domain of CD72 (amino acids 117–359) fused to a biotinylated human Fc domain to enable in vitro nanobody panning (Fig.4B; ref. 35). After six rounds of magnetic bead and flow cytometry-based selection (Fig.4C and Methods), >50% of the remaining nanobody-expressing yeast specifically bound CD72. Ultimately, we identified eight unique clones that displayed high affinity for recombinant CD72 and no binding to Fc domain only (Fig.4D). CDR3, the major binding determinant for both nanobodies and antibodies (36), possessed a wide range of length and sequence variability. Through the use of on-yeast affinity assays (37), measured clones were estimated to possess dissociation constant (KD) values in the low-nanomolar range for recombinant CD72 (Fig.4E). Taken together, these results indicate that this in vitro–evolved approach can generate nanobody binders to CD72.
In Vitro Activity of CD72-Directed CAR-T against MLLr B-ALL and Other B-cell Malignancies
We cloned our eight unique nanobody sequences into a second-generation CAR-T format to screen activity in vitro. Notably, the lentiviral backbone (Fig.5A) is identical to that used in tisangenlecleucel, an FDA-approved CD19 CAR-T. We first transduced Jurkat cells with our nanobody-based CARs to assess their antigen-independent and antigen-dependent activation during coculture with either a CD72-negative cell line (AMO1, multiple myeloma) or a CD72-positive cell line (RS411, MLLr B-ALL). For all assays, we used the tisangenlecleucel single-chain variable fragment (scFv) CD19 binder as a positive control. At a 1:1 effector:tumor (E:T) ratio, by CD69 staining we found that clone NbD4 possessed superior antigen-dependent activation while demonstrating low antigen-independent activation (Fig.5B). We therefore utilized this clone as our lead binder candidate in all subsequent experiments. CD72(NbD4) CARs were generated using normal donor CD4+ and CD8+ T cells using standard methodology and screened for cytotoxicity against multiple B-cell malignancy cell lines in vitro (see Methods). CD72(NbD4) CARs demonstrated potent cytotoxicity against CD72-bearing B-ALL and lymphoma cell lines when cocultured with tumor cells across varying E:T ratios and incubation times, mirroring the efficacy of CD19 CAR-T cells (Fig.5C and D). Use of the proliferation stain CellTrace Violet indicated that CD72(NbD4) CAR-T cells also demonstrated robust proliferation after 1:1 coculture for 72 hours with the SEM cell line, equivalent to CD19 CAR-T cells (Fig.5E). We observed robust degranulation of CD72(NbD4) CAR-T cells against these same cell lines during 6-hour coculture at a 2:1 E:T ratio, as well as against primary B-ALL patient samples, although the degree of degranulation was somewhat less than that of CD19 CAR-T cells (Supplementary Fig. S5A–S5C). Similarly, cytokine profiling of CD72(NbD4) CAR-T cells after 1:1 coculture for 24 hours with the SEM B-ALL cell line showed patterns of cytokine release comparable to those for CD19 CAR-T cells, although of lower magnitude (Fig.5F).
To evaluate the utility of CD72 CAR-T cells as a therapeutic option after CD19 failure, we suppressed CD19 in SEM cells using CRISPRi to generate a model of CD19 antigen escape. CD72(NbD4) CAR-T cells were equally efficacious against CD19-negative SEM cells as parental (Fig.5G), whereas CD19 CAR-T cells showed greatly diminished activity. Additionally, we knocked down CD72 and showed that CD72(NbD4) CAR-T cells had no detectable activity against these cells, whereas CD19 CAR-T cells retained robust killing (Fig.5H). Thus, CD72(NbD4) CAR-T therapy is highly specific and potent against CD72-bearing B cells, and effective targeting of CD72 is independent of CD19 surface density.
In Vivo Activity of CD72(NbD4) CAR-T against MLLr B-ALL
Finally, we examined the in vivo efficacy of our CD72(NbD4) CAR-T cells against an MLLr B-ALL cell line (SEM) and an MLLr B-ALL PDX in NOD/SCID gamma (NSG) mice. We engineered both cells to express luciferase for noninvasive bioluminescence imaging (BLI). We implanted 1E6 cells via tail-vein injection and engraftment confirmed by BLI at either 3 or 10 days for SEM and PDX, respectively. Each cohort of mice (n = 6 per arm) received 5 million total CAR-T cells (a 1:1 mixture of CD4:CD8 primary T cells) engineered with either an “empty” CAR backbone, CD72(NbD4) CARs, or CD19 CARs. MLLr PDX-injected mice that received CD72(NbD4) CAR-T cells showed a strong response and undetectable leukemic burden by BLI, comparable to CD19 CAR-T cells, and significantly increased survival versus the empty CARs (Fig.6A). CD72(NbD4) CAR-T cells performed similarly to CD19 CAR-T cells against wild-type SEM, significantly prolonging survival compared with empty CARs (Fig.6B). Notably, CD72(NbD4) CAR-T cells robustly controlled CRISPRi CD19-knockdown SEM tumor cell growth in vivo (Fig.6C). These results underscore the orthogonal nature of these two antigens and verify that targeting CD72 could be effective against CD19-negative relapsed leukemia.
CD72(NbD4) CAR-T Displays No Toxicity against Normal Tissues
To evaluate potential off-tumor toxicity in key normal tissue compartments, we performed coculture experiments combining CD72(NbD4) CAR with normal donor peripheral blood mononuclear cells (PBMC), as well as other representative normal tissue, including human umbilical vein endothelial cells (HUVEC), human bone marrow stromal cells (BMSC), human induced pluripotent stem cell (iPSC)–derived neural progenitor cells (NPC), human embryonic stem cell (hESC)–derived cardiomyocytes, and macrophages differentiated in vitro from donor monocytes (Supplementary Methods). After 1:1 coculture for 24 hours, flow cytometry evaluation of the PBMC population subtypes demonstrated B-cell depletion by both CD72(NbD4) CAR-T cells and CD19 CAR-T cells, but no effects on T cells, neutrophils, natural killer (NK) cells, monocytes, or dendritic cells (Supplementary Fig. S6A). In 6-hour degranulation assays, CD72(NbD4) CARs showed strong activation against SEM MLLr B-ALL cells, but no discernable degranulation over empty CAR control in the presence of all normal tissues tested (Supplementary Fig. S6B). Given the recent evidence of CD19 expression in neuronal vascular pericytes, which may explain neurotoxicity associated with CD19-directed CAR-T therapy (38), we additionally analyzed human brain single-cell RNA-seq datasets to assess CD72 expression. In contrast to significant CD19 expression in CD248-expressing pericytes, our analysis demonstrated no meaningful CD72 expression in this population or anywhere else in human brain tissue (Supplementary Fig. S7). Taken together, these analyses suggest that targeting CD72 is likely to lead to limited, if any, off-tumor toxicity outside the B-cell compartment.
CD72 Loss Alters the B-ALL Membrane Proteome
Based on our CRISPRi screen in SEM cells and clinical data with CD19 and CD22 CARs (2), we anticipated that CD72 antigen escape is a potential form of resistance to CD72(NbD4) CAR-T cells. In support of this hypothesis, we noted that bone marrow–isolated MLLr B-ALL PDX cells showed a population with decreased CD72 in NbD4-treated versus empty CAR-treated mice (Supplementary Fig. S8A–S8C). Given the role of CD72 in modulating BCR signaling, we hypothesized that other components of this pathway, including other known immunotherapy targets (i.e., CD22 and CD79a), may have altered expression levels in the context of CD72 antigen escape. We modeled this scenario by coupling CRISPRi CD72 knockdown in RS411 MLLr B-ALL cells to quantitative cell surface proteomics. Contrary to our hypothesis, we did not find significantly altered regulation of the aforementioned well-known immunotherapeutic targets nor other BCR signaling components. However, we did find multiple surface proteins downregulated in parallel with CD72 that are primarily involved in cell adhesion, migration, and desmosome junctions (BST2, DSG2, and CD99L2; Fig.7A). Several of these genes were predictive of poor overall survival when highly expressed (Supplementary Fig. S8D). DSG2 downregulation was confirmed by flow in both SEM and RS411 CD72-knockdown cell lines (Supplementary Fig. S8E). Although thoroughly validating this proposal is beyond the scope of our work here, in conjunction with prior studies (39) these findings suggest a hypothesis that loss of CD72 may lead to decreased B-ALL adhesion within the marrow niche, thereby increasing vulnerability to standard chemotherapeutic regimens.
Modulating CD72 Antigen Density through SHIP1 Inhibition
Although CD72 is known to be a negative regulator of BCR signaling in normal B cells, its role in oncogenic BCR signaling remains mostly unexplored. We were inspired by recent work where treatment with ibrutinib, a Bruton tyrosine kinase (BTK) inhibitor, led to downregulation of several negative regulators of BCR signaling, including CD22, CD72, and PTPN6, presumably to reachieve a baseline level of BCR signaling strength (40). We reasoned that the opposite may also be true, whereby pharmacologically agonizing BCR signaling may increase surface CD72. Recent work in BCR–ABL B-ALL has suggested that pharmacologically inhibiting SHIP1 can agonize the BCR pathway, leading to tumor cell death via activation of the autoimmunity checkpoint (AIC; ref. 41). More broadly, upregulating immunotherapy targets using pharmacologic intervention has generated great interest as a strategy to mitigate antigen escape (42–44). To test this hypothesis, we treated the MLLr cell lines SEM and RS411 with the SHIP1 inhibitor 3AC and found a strong dose-dependent increase in CD72 surface abundance at 72 hours (Supplementary Fig. S8F). In addition, we found that 3AC showed some degree of cytotoxicity to MLLr cell lines, potentially suggestive of AIC-mediated cell death (Supplementary Fig. S8G and S8H). Using unbiased phosphoproteomics, we notably confirmed increased phosphorylation of BTK and LYN substrates (among others) in MLLr vs. BCR–ABL B-ALL cell lines, corroborating active BCR signaling pathways (Supplementary Fig. S8I). We next evaluated the ability of 3AC to enhance CD72 CAR-T cytotoxicity. RS411 (MLLr) and NALM6 (non-MLLr) B-ALL cells, with moderate to low baseline CD72 copy number (Fig.5C), indeed showed significant increases in CD72 CAR-T cytotoxicity after SHIP1 inhibition (Fig.7B and C). Notably, this increased cytotoxicity was in line with a relative increase in CD72 surface antigen density after 3AC (Fig.7B). These findings are also in agreement with our results in Fig.5C, where we saw a correlation between CD72 antigen density and CAR-T cytotoxicity across cell lines. In contrast, SEM cells, which express high (Fig.5C) levels of CD72, did not show any impact on CAR-T cytotoxicity after 3AC. This result suggests a possible “saturation limit” at which CD72 antigen density no longer affects efficacy, although this conclusion will require more investigation in the future. We do note that relatively high doses of the tool compound 3AC were required to achieve both CD72 upregulation and tumor cell death in these models. However, more potent, ideally clinical-grade inhibitors of SHIP1 are under active investigation. Taken together, our results suggest that combining SHIP1 inhibition with CD72-directed therapy could be a promising rational combination strategy, particularly in the setting of lower baseline CD72 antigen density.
Our unbiased assessment of the membrane proteome demonstrates that in B-ALL different oncogenic drivers can generate markedly different cell surface signatures. These results underscore the power of surface proteomic profiling to identify novel immunotherapeutic targets. We found that MLLr B-ALL possesses a unique cell surface signature, marked by increases in receptors related to cell adhesion with significantly depleted MHC I and II. Notably, many of these surface signatures would not be detected by RNA analysis alone. Although our results uncover several potential approaches for targeting MLLr biology via the cell surfaceome and thereby provide a useful resource for the community, here we focused on CD72 as a particularly promising target for MLLr disease. To our knowledge, this poorest-prognosis subtype of B-ALL does not have any previously known preferential immunotherapy targets. Notably, our strategy here may also be applicable to identifying other genotype-enriched surface targets in other malignancies.
Bioinformatic analysis and patient sample examination further indicated that CD72 is highly expressed in other B-cell malignancies and driven by critical B cell–related transcription factors but is minimally expressed in other tissues. Given the current clinically available armamentarium, CD72 is most likely to be placed as a second-line target for patients relapsing after CD19-directed therapy. However, there may be some advantages to targeting CD72 instead of CD19. These include potential for less neurotoxicity, based on our brain single-cell RNA-seq analysis (Supplementary Fig. S7). Therefore, we propose CD72 as an attractive immunotherapeutic target for this poorest-prognosis subtype of B-ALL, and future work may more firmly validate CD72 nanobody CAR-T cells as having broader utility in other B-cell malignancies.
Prior studies have described the development of antibody–drug conjugates (ADC) targeting CD72 for various B-cell cancer indications (45–47). However, to our knowledge, no prior work has proposed any specific targeting of CD72 on MLLr B-ALL, as we found here. Furthermore, based on available data this ADC strategy appears limited by modest efficacy, possibly due to poor internalization after antibody binding. In contrast, this property may be beneficial for cellular therapies. We are particularly excited that the fully in vitro nanobody selection platform we employ here can develop robust CAR-T–active binders. The ability to rapidly generate nanobodies for CAR-T cells, without the need for llama immunization and using a library widely available for academic use, promises to greatly expand the utility of this approach. In particular, the small size and strictly monomeric fold of nanobodies make them advantageous for immunotherapy applications. For example, almost all current CAR-T cells employ scFv binders, which suffer from well-known protein engineering problems, including the selection of chain orientation (VH–VL vs. VL–VH) and linker length, both of which heavily influence CAR behavior, tonic signaling, and clinical responses (48). The inherent simplicity of nanobodies eliminates these protein engineering difficulties while also potentially facilitating the design of CAR constructs simultaneously targeting multiple antigens.
There are caveats to our approach. First, although our in vitro and bioinformatic analyses suggest low probability of off-tumor toxicity with CD72(NbD4), we cannot fully exclude this possibility on some other hematopoietic cells, particularly if new binders more sensitive for very low levels of CD72 are employed. Encouragingly, based on transcriptional data, CD22 demonstrates expression patterns roughly similar to those of CD72 (Supplementary Fig. S4A); fortunately, significant off-target activities have not been noted in CD22-directed clinical trials (49). Furthermore, although our preclinical results here are certainly very promising, there is likely additional room for nanobody optimization and CAR engineering. Finally, one concern is the potential for immunogenicity. However, we note that the llama VHH framework residues are >80% homologous to human VH sequences, a vast improvement over the murine scFvs currently used clinically, which possess only ∼60% homology yet still have curative potential. Notably, a bi-epitopic, anti–B-cell maturation antigen (BCMA) nanobody CAR-T, generated via llama immunization, is currently in clinical trials for multiple myeloma and thus far has displayed impressive clinical outcomes with minimal impact of binder immunogenicity (34). Although these real-world data are highly encouraging, in future work we will continue to evaluate the potential for CD72 nanobody humanization while retaining potent activity.
Importantly, we have demonstrated that CD72(NbD4) CAR-T cells eliminate tumor cells lacking CD19 (Fig.5G and Fig.6C), supporting utility as a second-line therapy after CD19 failure, at least in the absence of complete lineage switching (11). However, our results indicate that CD72, like CD19 and CD22, can be lost without any major effect on B-ALL viability. Therefore, we coupled CRISPRi knockdown with cell surface proteomics to define membrane protein alterations in the context of CD72 downregulation. This strategy, which could be considered as “antigen escape profiling,” may prove to have broad utility in the study of immunotherapy resistance when applied to other novel targets. In addition, in anticipation of reversing resistance, we investigated the mechanistic hypothesis that agonizing BCR signaling could both increase CD72 expression and directly kill tumor cells. Future pharmacologic approaches to agonize BCR signaling could potentially be combined with CD72-targeted cellular therapy as part of a rational combination strategy in patients. To our knowledge, this is the initial demonstration of CD72 as an attractive cellular therapy target. Our in vitro nanobody engineered T-cell platform has potential for immediate clinical translation in MLLr B-ALL and potential future application in an array of tumor types.
Human Cell Lines and Patient Samples
Human cell lines (Supplementary Table S1) were a generous gift from Dr. Markus Müschen (City of Hope), and were authenticated by short tandem repeat (STR) analysis prior to conducting proteomics experiments. Cells were grown in RPMI 1640 medium supplemented with 20% fetal bovine serum (FBS) and 100 U/mL penicillin–streptomycin (pen-strep) and passaged for less than 6 months before use. Cell lines were not tested for Mycoplasma. All MLLr B-ALL PDXs (Supplementary Table S2) were obtained from the PRoXe at the Dana-Farber Cancer Center except for the ICN3 PDX line (MLLr B-ALL), which was a generous gift from Dr. Markus Müschen. Primary patient samples that had been viably frozen were obtained from the UCSF Helen Diller Cancer Research Center Tissue Bank (Supplementary Table S3). All cells were cultured at 37°C in a humidified incubator with 5% CO2. Modified cell lines were generated using lentivirus as described below.
Cell Surface Protein Labeling
Cell surface proteins were labeled with biotin using the N-linked glycosylation-site biotin labeling method (17). Briefly, 3 × 107 live cells were washed twice and resuspended in 1 mL of ice-cold phosphate-buffered saline (PBS) and then treated with 1.6 mmol/L sodium metaperiodate (13798-22; VWR International) at 4°C for 20 minutes to oxidize the vicinal diols of sugar residues linked to surface proteins. The cells were then washed twice in PBS in order to remove excess sodium metaperiodate. Cells were resuspended in 1 mL of ice-cold PBS and treated with 1 mmol/L biocytin hydrazide (90060; Biotium) and 10 mmol/L aniline (242284; Sigma-Aldrich) at 4°C for 90 minutes with gentle mixing in order to biotinylate free aldehydes exposed on the sugar residues. After labeling, cells were washed three times with ice-cold PBS to removed excess biotin, frozen in liquid nitrogen, and stored at −80°C until further processing for mass spectrometry. All experiments were performed in biological triplicate with replicates harvested from consecutive passages.
Cell Lysis, Cell Surface Protein Enrichment, and Peptide Digestion
Frozen cell pellets were thawed on ice in 1 mL of radioimmunoprecipitation assay (RIPA) buffer (20-188; Millipore Sigma) with the addition of 1× Halt protease inhibitors (78442; Thermo Scientific Pierce). After incubation on ice for 10 minutes, cells were disrupted by sonication and the lysates were clarified by centrifugation at 17,000 relative centrifugal force (RCF) at 4°C for 10 minutes. Clarified lysate was mixed with 500 μL of NeutrAvidin agarose resin (29200; Thermo Fisher Scientific) and incubated at 4°C for 2 hours with end-over-end mixing. NeutrAvidin beads with captured biotinylated surface proteins were washed extensively by gravity flow to remove unbound proteins using 50 mL of 1× RIPA + 1 mmol/L ethylenediaminetetraacetic acid (EDTA), followed by 50 mL of PBS + 1 mol/L NaCl, and finally 50 mL of 50 mmol/L avidin–biotin complex (ABC) + 2 mol/L urea buffer. Washed beads were resuspended in digestion buffer, comprised of 50 mmol/L Tris (pH 8.5), 10 mmol/L Tris(2-carboxyethyl)phosphine (TCEP), 20 mmol/L 2-iodoacetamide, and 1.6 mol/L urea, with 10 μg of added trypsin protease (90057; Thermo Fisher Pierce) to perform simultaneous disulfide reduction, alkylation, and on-bead peptide digestion at room temperature overnight (16–20 hours). After digestion, the pH was dropped to ∼2 with neat trifluoroacetic acid (T6508-10AMP; Millipore Sigma), and the peptide mixture was desalted using a SOLA HRP column (60109-001; Thermo Fisher Scientific) on a vacuum manifold. Desalted peptides were eluted with 50% acetonitrile (ACN; 34998-4L; Sigma-Aldrich) and 50% water with 0.1% trifluoroacetic acid and dried down completely in a SpeedVac (Thermo Fisher Scientific). Dried peptides were resuspended in Optima LC/MS Grade water (W64; Thermo Fisher Scientific) with 2% ACN and 0.1% formic acid (FA, 94318-250ML-F; Honeywell International). Peptide concentration was measured using 280-nm absorbance on a NanoDrop (Thermo Fisher Scientific), and the peptide concentration was adjusted to 0.2 μg/μL for MS runs.
Liquid Chromatrograpy/Mass Spectometry and Data Analysis
For each replicate, 1 μg of peptide was injected onto a Dionex UltiMate 3000 RSLCnano instrument with a 15-cm Acclaim PepMap C18 reverse-phase column (164534; Thermo Fisher Scientific). The samples were separated on a 3.5-hour nonlinear gradient using a mixture of Buffer A (0.1% FA) and Buffer B (80% ACN/0.1% FA), from 2.4% ACN to 32% ACN. Eluted peptides were analyzed with a Thermo Scientific Q Exactive Plus mass spectrometer. The MS survey scan was performed over a mass range of 350 to 1500 m/z with a resolution of 70,000, with a maximum injection time of 100 ms. We performed a data-dependent MS2 acquisition at a resolution of 17,500, AGC of 5E4, and injection time of 150 ms. The 15 most intense precursor ions were fragmented in the higher energy collisional dissociation (HCD) cell at a normalized collision energy of 27. Dynamic exclusion was set to 20 seconds to avoid oversampling of highly abundant species. The raw spectral data files have been deposited at the ProteomeXchange PRIDE repository (PXD016800).
Raw spectral data were analyzed using MaxQuant 18.104.22.168 (18) to identify and quantify peptide abundance and searched against the human Swiss-Prot annotated human proteome from UniProt (downloaded May 13, 2018, with 20,303 entries). The “match-between-runs” option was selected to increase peptide identifications, and the “fast LFQ” option was selected to calculate LFQ values of identified proteins. All other settings were left to the default MaxQuant values. The MaxQuant output data were analyzed using Perseus (50) and R 3.4.0 in R-Studio. Proteins annotated as “reverse,” “only identified by site,” and “potential contaminant” were filtered out, as were proteins that were quantified in fewer than 2 out of 3 biological replicates in at least one experimental group. Proteins were further filtered to include only membrane proteins or membrane-associated proteins using a manually curated list of surfaceome proteins (20). Missing values were imputed based on the normal distribution of the dataset as implemented by Perseus. Volcano plots were generated using output from a two-sample t-test comparing the log2-transformed LFQ protein abundance values from different cell lines with a false discovery rate set to 0.01. All proteomics results (as well as all other figures) were produced using the R package ggplot2 (51).
Total RNA from cell pellets (biological duplicates) were isolated with the QIAGEN RNeasy Kit (74104; QIAGEN), and mRNA was further purified from isolated total RNA by poly(A) separation using Oligo (dT)25 magnetic beads (S1550S; New England Biolabs) per the manufacturer instructions. Total RNA and mRNA concentrations were measured by NanoDrop. RNA-seq libraries were prepared from purified mRNA using the KAPA Stranded RNAseq Kit with Riboerase (KK8483; Roche) per the manufacturer's protocol and quantified using a 2100 Bioanalyzer High Sensitivity DNA Kit (5067-4626; Agilent). Next-generation sequencing was performed on an Illumina NextSeq High Output Kit using single-end, 75-bp reads at the Chan-Zuckerberg Biohub. Reads were mapped to the reference human (hg19) genome sequence, using TopHat2 aligner (52) with the default parameters. The mRNA expression level for each gene was represented as fragments per kilobase per million mapped reads (FPKM), called by Cufflinks (53). The raw sequence files have been deposited at the Gene Expression Omnibus repository (GSE142447).
Staining of cells was performed with either 0.5E6 or 1E6 total cells per sample unless otherwise noted, and the manufacturers' recommended amount of antibody was used in 100 μL total volume of fluorescence-activated cell sorting (FACS) buffer (PBS +2% FBS) for 30 to 60 minutes prior to washing with excess FACS buffer. Samples were immediately analyzed using either a CytoFLEX Flow Cytometer (Beckman Coulter) or FACSAria II flow cytometer (BD Biosciences). Antibodies used for flow cytometry are listed in Supplementary Table S4.
Samples were obtained from the UCSF hematopathology archive under an Institutional Review Board–approval protocol. Paraffin-embedded tissue was sectioned at 4 μm onto positively charged glass slides. The slides were incubated at 60°C for 2 to 12 hours prior to staining. All immunohistochemistry was performed using a Ventana Medical Systems DISCOVERY ULTRA automated slide preparation system (Roche). Following deparaffinization, tissue sections were conditioned using an alkaline buffer (CC1, 950-124; Roche) at 97°C for 92 minutes followed by treatment with hydrogen peroxide (760-4840; Roche) at room temperature for 12 minutes, prior to application of the primary antibody. A rabbit polyclonal antibody to CD72 (HPA044658, lot no. R40911; Atlas Antibodies), applied to the slides at a 1:250 dilution, was incubated for 32 minutes at 36°C. Primary antibody detection used HQ hapten-labeled anti-rabbit antibodies (760-4815; Roche), incubated for 12 minutes at 37°C, with the DISCOVERY HQ-HRP detection system (760-4820; Roche) and 3,3′-diaminobenzidine as the chromogen.
Expression of CD72 Fc-Fusion Antigen
DNA encoding the CD72 extracellular domain (amino acids 117–359) was polymerase chain reaction–amplified from a plasmid obtained from the Human ORFeome collection (hORFeome 8.1) and cloned into a mammalian expression vector, fused to the C-terminus of a human constant CH2–CH3 domain (Fc domain), along with a N-terminal Avidity AviTag to facilitate site-specific biotinylation during expression. For expression, 30 μg of plasmid was transiently transfected into Expi293F cells (A14527, modified to stably express ER-localized BirA; Thermo Fisher Scientific) using polyethyleneimine (Transporter 5, 26008-5; Polysciences) at a 4:1 polyethyleneimine:DNA mass ratio. Cells were cultured in Expi293 Expression Medium (A1435101; Thermo Fisher) supplemented with 100 μmol/L biotin for 5 to 7 days to allow for protein expression and biotinylation. To purify recombinant protein, cells were pelleted and the supernatant that contained the protein was recovered, filtered, and pH adjusted with PBS (pH 7.4), prior to loading onto a HiTrap Protein A HP antibody purification column (29048576; Cytiva) to capture the Fc-fusion protein. The column was washed with PBS, and the protein was eluted with 0.1-mol/L acetic acid, then buffer exchanged into PBS using an Amicon Ultra-4 10K device (10,000 MWCO; UFC503008; EMD Millipore). The concentration of Fc-fusion protein was determined by A280 on a NanoDrop, and the molecular weight was confirmed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis prior to aliquoting, snap-freezing in liquid nitrogen, and storage at −80°C until use in the yeast display.
Nanobody Selections Using Yeast Display
Nanobody yeast selections were performed as previously described (16) with some modifications. Briefly, the first magnetic-assisted cell sorting (MACS) selection round utilized 5E9 induced yeast from the naïve library with an initial pre-clear depletion step prior to specific enrichment. Depletion was performed by washing and resuspending induced yeast in selection buffer (20 mmol/L HEPES, pH 7.5; 150 mmol/L NaCl; 0.5% BSA; 20 mmol/L maltose) with 400 nmol/L recombinant Fc biotin protein plus 500 μL of Anti-Biotin MicroBeads (130-090-485; Miltenyi Biotec) and incubating for 40 minutes at 4°C. The sample was passed over a MACS LD column (130-042-901; Miltenyi Biotec) to capture nanobody yeast with affinity for the magnetic beads or the Fc protein domain. For enrichment, yeast recovered in the flow-through were pelleted, then resuspended in selection buffer with 1 μmol/L recombinant CD72 Fc biotin and Anti-Biotin MicroBeads and incubated for 1 hour at 4°C. The sample was passed over a MACS LS column (130-042-401; Miltenyi Biotec), washed extensively, then eluted with selection buffer off-magnet to recover nanobody yeast specific for the CD72 extracellular domain (ECD). Recovered yeast was grown overnight in glucose containing TRP-dropout media (D9530; United States Biological), then induced with galactose containing TRP-dropout media to begin the next round. Two rounds of MACS selection were performed with equivalent amounts of reagents used in each round. For subsequent FACS selection rounds, induced yeast from previous rounds was washed, resuspended in selection buffer, and incubated with recombinant CD72 Fc biotin protein for 1 hour at 4°C. To stain for FACS sorting, the yeast was washed, then resuspended in a fresh selection of buffer, and incubated with anti–HA-fluorescein isothiocyanate (FITC) antibodies (130-099-389; Miltenyi Biotec) and anti–biotin-allophycocyanin (APC) antibodies (130-113-288; Miltenyi Biotec) for 30 minutes at 4°C in the dark, then washed and resuspended in cold selection buffer prior to FACS sorting. Yeast was sorted on a FACSAria II flow cytometer by gating on the double-positive FITC/APC population of yeast (yeast expressing HA-tagged nanobodies on their surface and bound by recombinant CD72 Fc biotin protein, respectively). To isolate nanobody binders with increased affinity for the CD72 ECD, four rounds of FACS were conducted using decreasing amounts of recombinant protein (500 nmol/L, 100 nmol/L, 100 nmol/L, and 10 nmol/L). To isolate individual nanobody yeast clones after the final selection round, the yeast was sparsely plated on yeast peptone dextrose agar plates, and individual yeast colonies were picked, grown, and induced in 96-well deep-well microplates. Nanobody yeast clones were screened in a 96-well-plate format for CD72 ECD-specific binding by incubating with 10 nmol/L CD72 Fc-biotin protein or, in a separate plate to rule out Fc binders, 10 nmol/L Fc biotin protein. Clones were stained with antibodies and analyzed with a CytoFLEX Flow Cytometer. Nanobody yeast clones with both positive binding to CD72 Fc biotin and negative binding to Fc biotin were subsequently selected for plasmid isolation and Sanger sequencing.
Nanobody Binding Affinity Determination by On-Yeast FACS
Estimated nanobody binding affinities for the recombinant CD72 ECD were determined on-yeast using established methods as previously described (37). Briefly, nanobody-expressing yeast clones were incubated in selection buffer with multiple, decreasing concentrations of recombinant CD72 Fc biotin protein for several hours at 4°C to allow for complete binding equilibrium to occur. Yeast samples were washed, then stained with anti–HA-FITC and anti–biotin-APC for 30 minutes at 4°C in the dark. The yeast was washed and resuspended in selection buffer, then analyzed by flow cytometry on a CytoFLEX Flow Cytometer. To determine binding constants, spectra was gated on double-positive APC/FITC populations, and the median fluorescent intensity of the APC signal was plotted against the concentration of CD72 ECD recombinant protein to construct binding curves. Dissociation constants were determined by curve-fitting the data using nonlinear least-squares regression.
All CAR expression plasmids utilized identical components aside from the variable extracellular binding domains (CD72-directed nanobodies or CD19-directed scFv). The signaling components including the CD8 hinge and transmembrane domain, 4-1BB co-stimulatory domain, and CD3ζ signaling domain are identical to those utilized in the clinically approved CD19-directed CAR construct tisangenlecleucel. Nanobody binders were cloned into the CAR backbone plasmid using Gibson Assembly protocol. CAR expression vectors utilized a green fluorescent protein (GFP) marker for identification of CAR+ cells.
Jurkat CAR Activation Testing
Jurkat T cells were transduced with experimental CAR constructs including CD72-directed nanobody-based CARs, empty CARs, and CD19-directed CAR-T cells as previously described. Jurkat CAR–expressing cells were incubated with target cells at a 1:1 ratio overnight (50,000 cell each), then assessed for activation by examining CD69 cell surface upregulation by flow cytometry (anti-CD69-APC, 560967; BD Biosciences) on a CytoFLEX flow cytometer. Antigen-independent activation was assessed by incubation of Jurkat CAR cells alone or with the CD72- and CD19-negative, multiple myeloma cell line AMO1. Antigen-dependent activation was assessed by incubation of Jurkat CAR cells with the CD72-/CD19-positive B-ALL cell line SEM.
Transduction and Expansion of Human T Cells
Primary human T cells were purified from the leukapheresis products of anonymous healthy blood donors from the Blood Centers of the Pacific (now Vitalant) under an institutional review board–exempt protocol in accordance with the U.S. Common Rule (Category 4). CD8+ and CD4+ T-cell populations were isolated separately using RosetteSep Human T Cell Enrichment Cocktails (15023 and 15022; STEMCELL Technologies). T cells were cultured in X-VIVO 15 Medium (c04-418Q; Lonza) supplemented with 10% human AB serum (HP1022; Valley Medical), 10 mmol/L neutralized N-acetylcystine (A9165-5G; Sigma-Aldrich), and 55 μmol/L 2-mercaptoethanol (21985023; Thermo Fisher Scientific) and were passaged every 2 days. For expansion, T cells were stimulated with CD3/CD28 Dynabeads (11131-D; Thermo Fisher Scientific) according to the manufacturer's instructions (25 μL of beads per 1 million T cells) for 5 days and grown in the presence of 30 U/mL recombinant interleukin 2 (CYT-209; ProSpec-Tany Technogene). Transduction with CAR lentiviruses was performed 1 day after the start of bead stimulation without spinfection or the use of polybrene supplements. After the removal of activation beads, transduction efficiency was assessed by flow cytometry (GFP+), and T cells were optionally MACS enriched and then expanded for another 4 days. CAR T cells used in all experiments were normalized for CAR expression.
CAR-T In Vitro Cytotoxicity Assays
Cytotoxicity assays were conducted by mixing target cells with CAR-T cells for 4, 24, or 48 hours at ratios as described for each experiment. For measuring cytotoxicity by bioluminescence with target cell lines stably expressing effLuc, 150 μg/mL of d-luciferin (LUCK-1G; Gold Biotechnology) was added to each sample, incubated for 10 minutes at room temperature, and then read using a GloMax Explorer Plate Reader (Promega). Percent viable cells were normalized to the bioluminescence of target cells incubated alone (100% viable), and experiments were performed with technical triplicates.
CAR-T Proliferation Assays
CAR-T cells were stained with CellTrace Violet (CTV; Thermo Fisher Scientific) according to the manufacturer's protocol, then cocultured with or without target cells (1:1 effector-to-target ratio) for 72 hours. CAR-T proliferation and cell doublings were measured via flow cytometry by observing dilution of the CTV signal on a CytoFLEX Flow Cytometer.
CAR-T Cytokine Release Assays
CAR-T cytokine release was assessed after 24 hours of CAR-T cells in culture alone or in coculture with target cells at a 1:1 effector-to-target ratio. Cytokine concentrations in collected supernatant were measured by Eve Technologies using a multiplex bead assay with a BioRad Bio-Plex 200 system and a Milliplex MAP Human High Sensitivity T Cell Panel (EMD Millipore) according to their protocols.
NSG mice were obtained from in-house breeding stocks at the UCSF Preclinical Therapeutics Core (PTC) facility or from The Jackson Laboratory. A mixture of male and female NSG mice, 6 to 8 weeks old, were transplanted via tail-vein injection with one million B-ALL tumor cells including the SEM cell line, SEM-CRISPRi knockdown lines, or the ICN3 B-ALL xenograft, all previously modified to stably express luciferase. Tumor burden was assessed either 3 or 10 days post-transplantation (for cell lines or xenografts, respectively) through non-invasive bioluminescence imaging at the UCSF PTC on a Xenogen In Vivo Imaging System (Caliper Life Sciences). Mice were distributed to different experimental arms such that each arm had equal initial tumor burden. One day after distribution, mice received five million CAR-T cells with different targeting domains described for each experimental arm (1:1 mixture of CD4/CD8) via tail-vein injection. Tumor burden was subsequently monitored weekly for 1 month. Mice were followed for survival, which was denoted by the time to develop overt leukemia symptoms, at which point the mice were sacrificed and their bone marrow was harvested and viably frozen for later analysis. All studies were approved by the UCSF Institutional Animal Care and Usage Committee.
The INNP5D/SHIP1 inhibitor 3-alpha-aminocholestane (3AC) was obtained from Millipore Sigma (Calbiochem 565835). Concentrated stocks were made at 10 mmol/L in sterile dimethyl sulfoxide and stored at −20°C until used.
Cell Viability Assay
Cells were aliquoted into 96-well plates in 100 μL of complete media at 1,000 to 10,000 cells per well prior to addition of 3AC diluted into complete media at varying concentrations. After 2 days, cell viability was measured using the CellTiter-Glo Luminescence Cell Viability Assay (G7573; Promega) according to the manufacturer's instructions. Luminescence was measured on a GloMax Explorer Plate Reader. The percent of viable cells at each dose was normalized to untreated wells, and all experiments were performed with technical triplicates at each dose.
CD19 and CD72 Surface Density Measured by Flow Cytometry
Quantitative measurement of CD19 and CD72 surface density was determined by flow cytometry using the Quantum MESF Kit (827; Bangs Laboratories) and was performed according to the manufacturer's recommended procedure. Flow cytometry was performed on a CytoFLEX Flow Cytometer.
All data are presented as mean ± standard deviation. Statistical significance in proteomics and transcriptomics analysis was determined by Welch t test, a two-sample t test with the null hypothesis that the difference in the log2-transformation of proteomic LFQs, transcriptome FPKMs for RNA-seq, or transcriptome microarray intensity is equal to 0. A P < 0.05 is considered statistically significant. For Kaplan–Meier survival analysis, we used the log-ranked test to determine statistical significance.
M.A. Nix reports a patent pending. K.R. Parker reports personal fees from Cartography Biosciences outside the submitted work. A.T. Satpathy reports grants from Cancer Research Institute, National Institutes of Health, and Burroughs Wellcome Fund during the conduct of the study; personal fees from Immunai and Cartography Biosciences, grants from Arsenal Biosciences, and non-financial support from 10x Genomics outside the submitted work. A. Manglik reports personal fees from Third Rock Ventures, Epiodyne, and Ligand and grants from AEi2 outside the submitted work. A.P. Wiita reports grants from the National Institutes of Health during the conduct of the study and other from Indapta Therapeutics and Protocol Intelligence outside the submitted work; in addition, A.P. Wiita has a patent application pending. No potential conflicts of interest were disclosed by the other authors.
M.A. Nix: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. K. Mandal: Investigation. H. Geng: Software. N. Paranjape: Investigation. Y.T. Lin: Investigation. J.M. Rivera: Investigation.M. Marcoulis: Investigation. K.L. White: Investigation, methodology. J.D. Whitman: Validation, investigation. S.P. Bapat: Validation, investigation. K.R. Parker: Data curation, formal analysis. J. Ramirez: Resources, investigation. A. Deucher: Investigation, methodology. P. Phojanokong: Investigation. V. Steri: Investigation. F. Fattahi: Resources. B.C. Hann: Investigation. A.T. Satpathy: Resources, data curation, formal analysis. A. Manglik: Resources, writing–review and editing. E. Stieglitz: Resources, investigation. A.P. Wiita: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.
We thank Dr. Markus Müschen for providing B-ALL cell lines and insightful comments on the manuscript; Drs. Alex Martinko and Max Horlbeck for providing CRISPRi sgRNA libraries and advice on data analysis; Drs. Axel Hyrenius-Wittstein, Kole Roybal, and Justin Eyquem for providing CAR-T backbone construct, protocols for generating CAR-T cells, and advice on murine studies; Dr. Andrew Leavitt for providing access to stem cell harvest samples; and the staff of the UCSF Helen Diller Family Comprehensive Cancer Center (HDFCCC) Tissue Core for performing immunohistochemistry. We also thank the staff of the UCSF Antibiome Center for advice and assistance with recombinant protein expression, as well as the Chan-Zuckerberg Biohub staff for assistance with sequencing. Finally, we would also like to acknowledge the family of a patient who experienced relapse due to CD19 escape for raising funds to support this project. This work was supported by the UCSF Department of Laboratory Medicine; a lymphoma pilot grant from the UCSF HDFCCC; grants from the National Institutes of Health (DP2 OD022552 and K08 CA184116 to A.P. Wiita; K08 CA230188 to A.T. Satpathy); a Career Award for Medical Sciences from the Burroughs Wellcome Fund and a Technology Impact Award from the Cancer Research Institute (to A.T. Satpathy); and a grant from the National Institutes of Health (P30 CA082103 to Preclinical Therapeutics Core managed by B.C. Hann, the HDFCCC Pediatric Malignancies Tissue Bank, and HDFCCC Tissue Core).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Note: Supplementary data for this article are available at Cancer Discovery Online (http://cancerdiscovery.aacrjournals.org/).