Key Points
CRISPR interference screening and patient epigenetic analysis reveal regulators of CD38 surface expression including XBP1 and SPI1.
CD38 knockdown does not lead to broad myeloma cell surface remodeling.
Visual Abstract
CD38 is a surface ectoenzyme expressed at high levels on myeloma plasma cells and is the target for the monoclonal antibodies (mAbs) daratumumab and isatuximab. Pretreatment CD38 density on tumor cells is an important determinant of mAb efficacy. Several small molecules have been found to increase tumor surface CD38, with the goal of boosting mAb efficacy in a cotreatment strategy. Numerous other CD38-targeting therapeutics are currently in preclinical or clinical development. Here, we sought to extend our currently limited insight into CD38 surface expression by using a multiomics approach. Genome-wide CRISPR interference screens integrated with patient–centered epigenetic analysis confirmed known regulators of CD38, such as RARA, while revealing XBP1 and SPI1 as other key transcription factors governing surface CD38 levels. CD38 knockdown followed by cell surface proteomics demonstrated no significant remodeling of the myeloma “surfaceome” after genetically induced loss of this antigen. Integrated transcriptome and surface proteome data confirmed high specificity of all-trans retinoic acid in upregulating CD38, in contrast to the broader effects of azacytidine and panobinostat. Finally, unbiased phosphoproteomics identified inhibition of MAP kinase pathway signaling in tumor cells after daratumumab treatment. Our work provides a resource to design strategies to enhance efficacy of CD38-targeting immunotherapies in myeloma.
Introduction
Harnessing the immune system to treat myeloma has rapidly become the most exciting therapeutic frontier in this disease. The first such immunotherapy agent to achieve the US Food and Drug Administration approval was the monoclonal antibody (mAb) daratumumab.1 Daratumumab targets CD38, a cell surface ectoenzyme highly expressed on myeloma plasma cells. Daratumumab is currently used as either monotherapy or combination therapy in the relapsed/refractory setting or frontline therapy in combination with other small molecule agents.1 A second mAb targeting CD38, isatuximab, was also recently approved for relapsed/refractory myeloma; at least 15 additional CD38-targeting agents are in development.2 Extensive and encouraging clinical data have already been obtained with daratumumab, although resistance appears to inevitably occur.3,4 Biologically, this process appears to be quite complex, with determinants of resistance ranging from alteration of surface antigens on tumor cells4-6 to dysfunction of the tumor immune microenvironment.7,8 Although it remains unclear whether CD38 downregulation on tumor cells after mAb treatment is a marker of resistance5,9 or, instead, successful therapy,10 compelling preclinical and clinical data suggest that CD38 surface antigen density before treatment strongly correlates with mAb efficacy.5,11
This latter observation has led to numerous efforts to identify small molecules that can increase tumor surface antigen density of CD38, representing potential cotreatments with CD38-targeting mAbs. The first such example of a CD38-boosting small molecule was all-trans retinoic acid (ATRA).12 Subsequent studies identified the pan-histone deacetylase (HDAC) inhibitor panobinostat,13 the thalidomide analog lenalidomide,14 the JAK inhibitor ruxolitinib,15 and the DNA methyltransferase (DNMT) inhibitor azacytidine (Aza)16 as agents that could lead to myeloma surface CD38 increase. A clinical trial combining ATRA with daratumumab has led to encouraging outcomes in patients previously refractory to daratumumab.17
Although these published strategies suggest ways to improve CD38 mAb outcomes, they also leave many questions unanswered. Most notably, we do not yet have a broad global sense of the transcriptional or posttranscriptional networks that most strongly affect CD38 expression. Bispecific and trispecific antibodies18 and chimeric antigen receptor T cells19 targeting CD38 are also in clinical development. As seen for similar modalities against other targets,20 efficacy of these novel agents, in addition to mAbs, is likely to also be affected by CD38 antigen density on tumor cells. Furthermore, prior studies showed that CD38 downregulation after daratumumab treatment was accompanied by increases in the complement inhibitors CD55 and CD59.5 Are there other features of myeloma surface remodeling driven by CD38 downregulation? For the small molecules noted above, it is unknown how they more generally affect the makeup of the myeloma cell surface proteome beyond CD38. The tumor cell surface not only harbors opportunities for immunotherapeutic targeting but also is the interface for communication between tumor and microenvironment, potentially leading to other alterations in myeloma biology after changes in surface proteomic profile. To address these questions, here, we have taken advantage of CRISPR interference (CRISPRi)–based functional genomic screens, cell surface proteomics, epigenetic analyses, and phosphoproteomics to provide a multiomic perspective on CD38 regulation and tumor cell consequences of targeting CD38 in myeloma.
Methods
CRISPRi screening and hit validation
Genome-wide CRISPRi screening was performed as described previously.21 Briefly, RPMI-8226 cells stably expressing dCas9-KRAB were transduced with a genome-wide library comprising 5 single guide RNAs (sgRNAs) per protein-coding gene. After 14 days, cells were stained for surface CD38 and flow sorted to enrich for populations of cells expressing low or high cell surface levels of CD38. Cell populations were then processed for next-generation sequencing as previously described22 and sequenced on a HiSeq-4000 (Illumina). Reads were analyzed by using the Model-based Analysis of Genome-wide CRISPR/Cas9 Knockout pipeline as previously described.23 Further validation was performed by knockdown with individual sgRNAs extracted from the genome-wide library with conformation by flow cytometry or western blotting. Antibody-dependent cytotoxicity assays were performed using NK92-CD16 cells as described previously.16 Additional details are provided in supplemental Methods.
Epigenetic analysis and machine learning for CD38 transcriptional regulation
Publicly available Assay for Transposase-Accessible Chromatin with sequencing (ATAC-Seq) data from primary myeloma samples24 were analyzed with the Homer tool findPeaks. Motif binding in the identified ATAC peak regions was called with PROMO tool.25 Newly diagnosed patient tumor RNA-sequencing (RNA-seq) data in the Multiple Myeloma Research Foundation CoMMpass trial were used to correlate expression of predicted transcription factors with CD38 expression. To build a predictive model for CD38 expression as a function of transcription factor expression, we developed an XGBoost (Extreme Gradient Boosting) model with randomized search with crossvalidation to find optimal parameters. Eighty percent of CoMMpass data were used for training and the remainder for model testing. Additional details are provided in supplemental Methods.
Cell surface proteomics and phosphoproteomics
Cell surface proteomics was performed using an adapted version of the N-glycoprotein cell surface capture26 method, as we have described previously.27 Unbiased phosphoproteomics was performed using immobilized metal affinity chromatography using methods described previously.28 All samples were analyzed on a Thermo Q-Exactive Plus mass spectrometer with data processing in MaxQuant.29 Additional details are provided in supplemental Methods.
Institutional review board approval was obtained for all animal studies in this work (approval number UCSF IACUC: AN194778-01). No human subjects research is included.
Results
A CRISPRi-based screen reveals regulators of CD38 surface expression
We first sought to use an unbiased approach to identify regulators of surface CD38 in myeloma tumor cells. We specifically used genome-wide screening with CRISPRi, which leads to much higher specificity of knockdown than short hairpin RNA while avoiding potential toxicity of double-strand breakage with CRISPR deletion.30 We recently used this approach to characterize regulators of surface B-cell maturation antigen (BCMA) in myeloma.21 Here, we used an RPMI-8226 cell line with the dCas9-KRAB machinery, required for CRISPRi, as described previously.21 We confirmed that this RPMI-8226 cell line robustly expressed CD38 (supplemental Figure 1A).
The genome-wide screen was performed as shown in Figure 1A. Briefly, RPMI-8226 cells were transduced with a pooled genome-wide sgRNA library. After 14 days, the cells were then stained with fluorescently labeled anti-CD38 antibody and flow sorted into low- and high-CD38 populations. Frequencies of cells expressing each sgRNA was quantified using next-generation sequencing. As an important positive control, increasing confidence in the screen results, we first noted that knockdown of CD38 itself strongly decreased surface CD38 expression (Figure 1B). On the contrary, several dozen genes, when repressed, did indeed lead to increased surface CD38 (right side of volcano plot in Figure 1B; supplemental Table 1). As another positive control, one of these top hits included RARA, whose degradation is catalyzed by ATRA treatment to drive CD38 increase.12
To find pathways that may be useful for pharmacologic targeting, we first applied gene ontology and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis to the list of genes that, when inhibited, significantly increased CD38 (Figure 1C). We were intrigued to find that many of the strongest effects appeared to be driven by transcriptional or other epigenetic factors. These specifically included pathways such as “DNA replication,” “messenger RNA (mRNA) processing,” “DNA-templated transcription,” and “spliceosome.”
We considered whether any hits associated with these pathways may be “druggable,” with the goal of expanding our repertoire of small molecules that enhance surface CD38 in myeloma. SS18, a component of the BAF (BRG1/BRM associated factor) chromatin remodeling complex, scored highly as a hit. However, treatment with the proposed BAF inhibitor caffeic acid phenol ester32 did not lead to consistent increases in surface CD38 (supplemental Figure 1B). Similarly, the lysine demethylase KDM4A was a prominent hit, but treatment with the inhibitory metabolite (R)-2-hydroxyglutarate33 also had no effect (supplemental Figure 1B).
The strongest hits for genes whose knockdown increased surface CD38 were 2 transcription factors, HEXIM1 and TLE3. Validation studies using individual sgRNA knockdown confirmed increased surface CD38 (Figures 1D and 2A; supplemental Figure 2A), as well as functional impact in natural killer (NK)–cell antibody-dependent cellular cytotoxicity (ADCC) assays with daratumumab (Figure 2B). However, these proteins are known to be widespread negative regulators of transcription,34,35 suggesting little scope for specific therapeutic targeting at the CD38 locus.
We were surprised that other targets proposed to increase CD38 expression after pharmacologic inhibition, such as HDACs,3 or catalyzed degradation, such as IKZF1/3,25 did not appear as prominent hits (Figure 1B). However, this result may reflect a limitation of functional genomic screens. A pharmacologic agent may inhibit multiple members of a protein class to drive a phenotype, whereas, with single gene knockdown, functional redundancy may prevent this phenotype from appearing36 (ie, multiple HDACs may need to ablated at once, or both IKZF1 and IKZF3 degraded simultaneously, to drive increased CD38). We speculate this is the case with DNMTs. We previously showed that treatment with the DNMT inhibitor Aza, which promotes degradation of all cellular DNMTs,37 could robustly increase surface CD38.16 Here, however, we found that knockdown of any individual DNMT only led to minor CD38 increase (Figure 1B).
Given these findings, we therefore shifted our focus to genes that, when knocked down, led to CD38 decrease (left side of volcano plot in Figure 1B). We reasoned this approach could still reveal important biological inputs that regulate the surface expression of CD38. Examining specific genes, we found that the transcription factor SPI1 was the strongest hit besides CD38 that, when knocked down, repressed surface CD38 expression (supplemental Figure 2B). We also noted that NFKB1 and NFKB2 knockdown appeared to drive CD38 decrease. This finding was intriguing given the known importance of NF-κB signaling in myeloma proliferation and survival.38 KEGG and gene ontology analysis of genes whose knockdown significantly decreased CD38 showed enrichment for MAPK pathway and protein phosphorylation more broadly (supplemental Figure 1C), suggesting key roles for intracellular signaling in regulating surface CD38.
Validation experiments with individual sgRNAs confirmed that SPI1 knockdown strongly decreased CD38 surface expression by flow cytometry, with a lesser decrease in surface CD38 with NFKB2 knockdown (Figures 1D and 2C; supplemental Figure 2D-E). These alterations also led to functional impacts. RPMI-8226 cells with knockdown of these genes showed significantly decreased NK-cell lysis in ADCC assays (Figure 2D). We further probed this dynamic in vivo, finding that RPMI-8226 cells with SPI1 knockdown were relatively resistant to daratumumab in a murine model (Figure 2E; supplemental Figure 2F). We note that we attempted to expand these results to additional cell lines. However, our 4 other myeloma cell lines harboring the CRISPRi machinery21 all express extremely low levels of SPI1 (supplemental Figure 2C), and attempted knockdown in 2 of them (AMO1 and KMS12PE) did not elicit any phenotype (not shown). Therefore, this finding suggests that SPI1 may play an important role in regulating CD38 expression in some myeloma tumors, but it is less likely to be a universal regulator.
Epigenetic analysis suggests XBP1 as a key determinant of CD38 in primary myeloma tumors
Our CRISPRi results suggest that epigenetic and/or transcriptional regulation is a critical driver of surface CD38 levels. However, we do not know whether these specific hits in a myeloma cell line will extend to primary myeloma tumors. We therefore took a complementary approach to find potential transcriptional regulators of CD38. Using ATAC-seq data from 24 primary myeloma tumor samples,24 we extracted open chromatin motifs near the CD38 promoter (supplemental Figure 3A) to identify a list of 46 transcription factors with potential binding sites at this locus (supplemental Table 2). We then correlated expression (via Pearson R) of these transcription factors with CD38 expression across 664 primary patient tumors at diagnosis in the Multiple Myeloma Research Foundation CoMMpass database (release IA13). In this analysis, we found the transcription factor most negatively correlated with CD38 expression was RARA (Figure 3A), consistent with our CRISPRi screen data and underscoring the promise of ATRA as a cotreatment to increase CD38. Intriguingly, the transcription factor with the strongest positive correlation was XBP1 (Figure 3A-B), a central driver of plasma cell identity.39,XBP1 also showed strong positive correlations with CD38 in 2 other patient tumor gene expression data sets40,41 (supplemental Table 2).
To further extend this analysis, we sought to build a predictive model that could estimate CD38 transcript level as a function of transcription factor expression. We used an XGBoost method applied to CoMMpass mRNA-seq data to find weights of transcription factor expression that most influence CD38 levels in patient tumors. We first tested this analysis on 80% of patient data as a training set, with 20% left out as test set. We found this model, solely based on transcription factor expression, could predict about half of the variance (coefficient of variation R2 = 0.49 using fivefold cross validation) in test set CD38 levels (Figure 3C). Using model weights and Shapley additive explanations analysis (see supplemental Methods) to determine transcription factors that have the greatest impact, either positive or negative, on CD38 expression, we found that XBP1 played the strongest role overall. Other strong hits from both of our analyses included IRF2, ATF1, and STAT1. SPI1 also appeared in the top 10 most relevant transcriptional regulators (Figure 3D; supplemental Figure 3B), consistent with our CRISPRi results, suggesting that SPI1 may play a key role in regulating CD38 in a subset of tumors.
We further evaluated XBP1, given its prominent role in these 2 complementary bioinformatic analyses. In a prior data set of shRNA knockdown of XBP1 in myeloma plasma cells,42 CD38 mRNA was decreased approximately threefold after XBP1 silencing (supplemental Figure 3C). This finding was consistent with results in our CRISPRi screen, in which XBP1 knockdown led to significant surface CD38 decrease (Figure 1B). We further validated this relationship by using a doxycycline-inducible sgRNA construct for CRISPRi vs XBP1, finding a clear dose-response between degree of XBP1 knockdown and loss of surface CD38 by flow cytometry (supplemental Figure 3D-G). Supporting relevance of this link, a recent myeloma tumor whole-genome sequencing study found that deletion of XBP1 was one of the strongest determinants of clinical response to daratumumab.43 We attempted to perform promoter activation assays to directly link XBP1 binding to CD38 expression but were unable to successfully generate a reporter construct that reflected all 8 loci in which XBP1 may bind at CD38 regulatory elements (supplemental Figure 3H). Taken together, these results nominate XBP1 as a particularly strong determinant of surface CD38 in myeloma plasma cells, although future investigation will be required to validate a direct or indirect relationship to CD38 transcription.
No consistent large-scale remodeling of the myeloma surface proteome after CD38 downregulation
We next evaluated CD38 surface regulation from the perspective of mAb therapy. In clinical samples, CD38 loss after daratumumab was accompanied by increases in CD55 and CD59, which may inhibit complement-dependent cytotoxicity and contribute to daratumumab resistance.5 In preclinical studies, macrophage trogocytosis has been proposed as a mechanism contributing to CD38 loss after mAb treatment, which also leads to alterations in other surface antigens, including CD138/SDC1.7 However, we hypothesized that, given its enzymatic activity and role as a cellular differentiation marker,44 loss of CD38 on its own may influence surface expression of other myeloma antigens. Such alterations may reveal new biology or (immuno)therapeutic vulnerabilities of CD38 mAb–treated disease.
To test this hypothesis, we used a method we recently developed termed “antigen escape profiling.”27 We used CRISPRi to transcriptionally repress CD38 in RPMI-8226, AMO-1, and KMS12-PE myeloma cell lines, using this genetic approach to partially mimic the loss of surface antigen seen after mAb therapy (Figure 4A). We then performed cell surface capture proteomics26,27 to uncover surface proteome alterations in a relatively unbiased fashion. Across cell lines, analyzed in biological triplicate with CD38 knockdown vs nontargeting sgRNA, we quantified 897 proteins annotated as membrane-spanning in Uniprot (minimum of 2 peptides per protein; supplemental Table 3). As a positive control, in all lines, we found that the strongest signature was the decrease of CD38 itself (supplemental Figure 4A-C). However, when aggregating proteomic data, we found no significant alterations in any surface antigens beyond CD38 itself (Figure 4B). Integration with RNA-seq data revealed only THY-1/CD90 as upregulated more than threefold at both the mRNA and surface proteomic level after CD38 knockdown (Figure 4C). Intriguingly, CD90 is known as a marker of “stemness” in early hematopoietic lineage cells that is lost when CD38 expression is increased.45 CoMMpass analysis also confirmed increased THY1 expression in tumors with lower CD38 (supplemental Figure 4D). However, further validation as to whether CD90 is truly altered after CD38 mAb will require pretreatment and posttreatment clinical specimens, beyond the scope of our work here. Overall, we conclude that loss of CD38 in isolation leads to limited remodeling of the myeloma surface proteome.
Integrated surface proteomic and transcriptional analysis suggests ATRA is highly specific in CD38 upregulation
Data from our group16 and others12-15 have suggested that several small molecules can increase myeloma surface CD38. However, the broader impacts of these agents on membrane antigens beyond CD38 have not been directly compared. We performed integrated cell surface proteomics and transcriptional analysis of RPMI-8226 cells treated with 10 nM ATRA, 2 μM Aza, and 10 nM panobinostat, all treated for 72 hours, in comparison with 0.1% dimethyl sulfoxide (supplemental Table 4). These doses are chosen because they have been previously published to significantly increase myeloma surface CD38 by flow cytometry.12,13,16 Notably, the magnitude of the increase in surface CD38 after 10 nM ATRA treatment, as measured by surface proteomics, was consistent with that which we previously observed by flow cytometry.16 In this integrated analysis, we found much broader impacts of Aza and panobinostat than ATRA on the “surfaceome” of plasma cells, beyond increasing CD38 (Figure 5A). These results suggest that, at doses driving CD38 upregulation, for Aza or panobinostat, altering CD38 is just a small component of their impact on myeloma tumor cells, whereas ATRA is much more specific in driving CD38 upregulation.
Toward understanding how CD38 is modulated after drug treatment, in our previous work,16 we noted that the mechanism of CD38 increase after Aza treatment was unclear. We thus further investigated the global transcriptional response (ie, not limited to membrane) after Aza (supplemental Table 4). Prior studies have suggested that Aza antitumor effect is largely mediated by reactivation of endogenous retroviruses stimulating a tumor-autonomous interferon response.46,47 Consistent with this work, we found a pronounced increase in interferon-responsive genes after Aza, but not ATRA, including IRF1, IFITM1, IFITM2, and IFITM3 (Figure 5B). KEGG analysis also confirmed this effect (Figure 5C). Given evidence across multiple systems that interferon upregulates CD38 expression,48,49 our transcriptional profiling thus also supports an interferon-based mechanism driving surface CD38 increase in plasma cells after Aza treatment.
Plasma cell proliferative signaling pathways are inhibited by mAb binding to CD38
In our final set of experiments related to targeting surface CD38, we were intrigued as to whether binding of a therapeutic mAb leads to specific cellular phenotypes within myeloma plasma cells. For example, isatuximab is known to directly lead to apoptosis of plasma cells,50 and daratumumab can do so after crosslinking.51 However, the mechanism underlying this transduction of extracellular mAb binding to intracellular phenotype remains unclear. In addition, our CRISPRi screen data (Figure 1B; supplemental Figure 1C) suggest that surface CD38 expression may be strongly affected by intracellular phospho-signaling pathways.
Therefore, we used unbiased phosphoproteomics by mass spectrometry to probe intracellular signaling effects driven by CD38 mAb binding. In RPMI-8226 cells, we compared 20 μM daratumumab treatment vs IgG1 isotype control. This supraphysiological dose of daratumumab was chosen to maximize signal-to-noise in the downstream phosphoproteomics assay. We chose a time point of 20 minutes of treatment, given the known rapid alterations in signaling pathways in similar phosphoproteomic experiments.52 In total, across triplicate samples, we quantified 5430 phosphopeptides (supplemental Table 5; supplemental Figure 5A). Analyzing phosphopeptide changes by kinase substrate enrichment analysis,53 we were intrigued to find downregulation of phosphorylation motifs consistent with both cyclin-dependent kinases as well as several kinases of the MAP kinase pathway (Figure 6A). Downregulation of phosphorylation on several central nodes in the MAP kinase as well as AKT pathway was also apparent via KEGG analysis (supplemental Figure 5B). Across a time course, we further confirmed effects on MAP kinase pathway (reported by phosphorylation of MAPK [ERK1/2], a key node in this response) and AKT signaling after daratumumab treatment via western blotting in RPMI-8226 and MM.1S cell lines, respectively (Figure 6B). Although the absolute value of changes in MAPK signaling are modest, both by phosphoproteomics and western blot, these results indicate that daratumumab binding to CD38 can at least partially inhibit this central proliferative pathway within myeloma tumor cells and thus may form a component of daratumumab’s antitumor effect.
Discussion
Our studies here present a “multiomics” view of therapeutically targeting CD38 in multiple myeloma. Our integrated functional genomics and epigenetic analysis point to the central role of transcriptional regulators in governing CD38 surface expression. Using surface proteomics, we further identify that loss of CD38 in isolation is unlikely to drive large changes in the “surfaceome,” whereas known pharmacologic strategies to increase CD38 have largely divergent impacts on other surface antigens. Finally, unbiased phosphoproteomics reveals that binding of anti-CD38 mAb can impair intracellular proliferative signaling within plasma cells.
Our CRISPRi screen illustrated the central role of numerous transcription factors, such as SPI1, HEXIM1, and TLE3, in regulating surface CD38. This functional genomic study suggests that regulation of surface CD38 largely occurs at the transcriptional, as opposed to protein trafficking, level. This finding was in sharp contrast to our prior CRISPRi results with BCMA, in which we found that posttranscriptional mechanisms, such as proteolytic cleavage by γ-secretase and protein trafficking via the SEC61 translocon, played the strongest roles in determining surface BCMA levels.21
Another recent study by Anderson and colleagues used genome-wide CRISPR deletion screening to find genes that, when knocked out, could abrogate interleukin-6–mediated downregulation of surface CD38.15 The strongest hits in this prior study included the transcription factors STAT1 and STAT3, demonstrating a role for JAK-STAT signaling in regulating tumor CD38 expression within the bone marrow microenvironment.15 In support of this notion, our integrated epigenetic and machine learning analyses, extracted from bone marrow–derived patient tumor samples, also support a critical role for STAT1 in governing surface CD38. However, in our CRISPRi screen in an in vitro monoculture system, neither STAT1 nor STAT3 affected CD38 surface expression (Figure 1B). This result suggests that JAK-STAT signaling may not play a major role in CD38 regulation in the absence of exogenous tumor stimulation. This finding illustrates the complementary nature of our genome-wide screen to that previously published under the context of interleukin-6 stimulation.15 Even more recently, another study from the same group used CRISPR knockout screening to identify KDM6A as an important regulator of both surface CD38 and daratumumab-mediated ADCC.54 Work by others also recently showed that KDM6A knockout can lead to CD38 transcript downregulation in myeloma models.55 In our studies, KDM6A was not a prominent hit (Figure 1B), possibly due to the differences between partial knockdown via CRISPRi and full knockout via CRISPR nuclease. Intriguingly, KDM4A, another histone demethylase, was one of the strongest hits in our screen that when knocked down led to surface CD38 increase (Figure 1B). These findings raise the possibility of an epigenetic interplay between these enzymes in the context of CD38 regulation.
Toward the goal of finding key regulators of CD38 that were not previously known, our epigenetic and machine learning approaches suggest that XBP1 is a critical regulator of plasma cell CD38. To our knowledge, there are not currently any known pharmacologic mechanisms to potentiate XBP1 activity. Given the important role of XBP1 splicing in myeloma plasma cells,56 future work will investigate the role spliced vs unspliced XBP1 in specifically regulating CD38, because this strategy may provide new avenues for CD38 manipulation. Future work will also investigate the role of XBP1 deletion in determining clinical response to daratumumab.43 The clinical study of Maura et al43 also found that genomic deletions of CYLD were strongly enriched in patient tumors that were nonresponsive to daratumumab. This gene, along with others such as TRAF3, are known to regulate NF-κB signaling in myeloma.57 However, we did not observe significant changes in surface CD38 after knockdown of CYLD or TRAF3 (supplemental Table 1), unlike NFKB1 and NFKB2. We cannot exclude insufficient knockdown of these genes in our assay, or cell-line–specific differences in NF-κB signaling, as explanations for why we found CRISPRi impacts of some, but not all, genes within the NF-κB pathway on surface CD38 (supplemental Figure 1C).
Given that plasma cells demonstrate frequent loss of surface CD38 after daratumumab treatment,5 a pressing question is whether CD38-low, daratumumab-resistant cells have novel immunotherapeutic vulnerabilities. However, our recently described strategy of “antigen escape profiling,”27 which involves CRISPRi knockdown followed by unbiased cell surface proteomics (Figure 4), suggests that other surface antigens on plasma cells do not exhibit consistent changes due to CD38 downregulation alone. That said, although our proteomic analysis after CD38 knockdown showed limited common features across 3 myeloma cell lines, we cannot rule out that major surfaceome remodeling could truly be present but with marked variability in responsive surface proteins from line to line. However, our currently favored explanation for this result is excess experimental noise in the surface proteomics quantification that leads to limited replication in the aggregated data across cell lines. We thus believe these findings support the notion that alterations in surface proteins found after mAb treatment on patent tumors, such as increases in CD55 and CD59,5 are caused by other therapy-induced selective pressure within the tumor microenvironment, not CD38 loss.
Although in vitro assays have suggested a strong relationship between CD38 antigen density and either NK-cell–58 or macrophage-mediated antibody-dependent cell killing,6 these experiments cannot readily take into account the critical role of the immune microenvironment in determining daratumamab response or resistance.8 Furthermore, even with the higher specificity of ATRA, there is the potential to alter CD38 expression on other hematopoietic cells, which may affect clinical responses to daratumumab.17 Notably, current clinical data are most consistent with pretreatment tumor CD38 antigen density positively correlating with daratumumab depth of response.5,11 Surprisingly, analysis of transcriptional data in CoMMpass demonstrate that increased tumor CD38 expression at diagnosis was not associated with improved outcomes in patients treated with daratumumab later in their clinical course (supplemental Figure 6). Pharmacologic manipulation of CD38 density on tumor cells may ultimately be most fruitful before treatment rather than in the context of daratumumab resistance. Similar strategies may also be most beneficial for other CD38-targeting immunotherapeutics.
Furthermore, directly related to mAb therapeutic effects, our unbiased phosphoproteomic results suggest that daratumumab binding to CD38 can directly decrease signaling along the MAP kinase and PI3K-AKT pathways. It remains to be investigated whether this inhibition of central proliferative signaling pathways plays a role in the antitumor effect of daratumumab in patients.
In terms of limitations of our work, the most prominent is that the many of our studies are derived from large-scale “omics” experiments in myeloma cell lines. There may be biological differences between our findings in vitro and primary tumors growing within the bone marrow microenvironment.
Taken together, our multiomic studies comprise a resource that reveals new insight into the genetic, epigenetic, and pharmacologic regulation of surface CD38 in myeloma plasma cells. We anticipate these findings will have utility in deriving new strategies to enhance CD38-targeting therapies in myeloma, including mAbs in current clinical practice as well as emerging antibody and cellular therapies. The technologies described here also comprise a blueprint to comprehensively assess determinants of surface antigen regulation and impacts of associated therapeutic manipulation, which could be applied across targets in hematologic malignancies.
Acknowledgments
The authors thank Ruilin Tian for assistance in applying MAGeCK scripts for the analysis here.
This work was supported by grants NIH K08 CA184116, NIH R01 CA226851, and the UCSF Stephen and Nancy Grand Multiple Myeloma Translational Initiative (A.P.W.), NCI P30 CA082103 supporting the Preclinical Therapeutics Core facility (managed by V. Steri and B.H.), and NIH K99/R00 CA181494 and a Stand Up to Cancer Innovative Research Grant (M.K.).
Authorship
Contribution: P.C., P.R., M.K., and A.P.W. conceived and designed the study; P.C., C.K., B.P.-E., A. Kang, A. Kishishita, S.R., J.C.P., O.G., L.S., Y.-H.T.L., P.R., N.P., and M.M. performed experiments; P.C., B.P.-E., N.P., Y.-H.T.L., B.G.B., and P.R. analyzed data; D.W., P.P., V. Steri, and B.H. performed murine studies; H.G. analyzed patient epigenetic data; V. Sarin performed machine learning analysis; and P.C. and A.P.W. wrote the manuscript with input from all authors.
Conflict-of-interest disclosure: P.C. is a shareholder of Genentech/Roche. P.R. is a shareholder of Senti Biosciences. A.P.W. is an equity holder and scientific advisory board member of Indapta Therapeutics, LLC and Protocol Intelligence, LLC. M.K. has filed a patent application related to CRISPRi screening (US patent number PCT/US15/40449); and serves on the scientific advisory boards of Engine Biosciences, Cajal Neuroscience, and Casma Therapeutics. The remaining authors declare no competing financial interests.
The current affiliation for P.C. is Genentech/Roche, South San Francisco, CA.
The current affiliation for P.R. is Senti Biosciences, South San Francisco, CA.
Correspondence: Arun P. Wiita, Department of Laboratory Medicine, University of California San Francisco, 185 Berry St, Ste. 290, San Francisco, CA 94107; email: arun.wiita@ucsf.edu.
References
Author notes
C.K. and B.P.-E. contributed equally to this study.
A. Kang, A. Kishishita, and S.R. contributed equally to this study.
RNA-sequencing (RNA-seq) and proteomics data have been deposited to public repositories. Proteomics data can be accessed at the Proteomics Identification Database repository (accession number PXD027594) and RNA-seq data at the Gene Expression Omnibus repository (accession number GSE181277). Additional processed data are included in supplemental Tables.
Other experimental data are available on request from the corresponding author, Arun P. Wiita (arun.wiita@ucsf.edu).
The full-text version of this article contains a data supplement.