Targeting caseinolytic mitochondrial matrix peptidase, a novel contributor to the pathobiology of high-risk multiple myeloma Free

Regulated intracellular proteolysis through the ubiquitin-proteasome pathway (UPP) plays a key role in plasma cell (PC) biology. UPP client proteins are involved in downstream pathways critical to cellular homeostasis, including cell cycle progression, growth, survival, and drug resistance.^1,2 The UPP also degrades damaged or misfolded proteins and thus, along with protein refolding and sequestration, promotes cytosolic and nuclear protein quality control.³ Through this mechanism, the UPP minimizes endoplasmic reticulum stress and dependence on the unfolded protein response (UPR),^1,2 promoting PC survival. Notably, high UPP gene expression has been associated with a more advanced disease and a possibly increased risk of progression from precursors.^4-6 Proteasome inhibition has become a cornerstone of therapy in multiple settings and combinations,^1,2,7 and these agents work in part by disturbing the balance between proteasome load and capacity, triggering apoptosis.⁸ 

Mitochondria have distinct protein quality control pathways, which include Lon peptidase 1 (LONP1) and the caseinolytic protease (CLP) endopeptidase in the mitochondrial matrix, whereas 2 adenosine triphosphatases (ATPases) associated with diverse cellular activities (AAA), matrix-AAA and intermembrane space–AAA, are in the inner mitochondrial membrane.⁹ CLP is a serine protease,^9,10 whereas the 20S proteasome has active site threonines,¹¹ but CLP has a structure reminiscent of the proteasome. The caseinolytic protease proteolytic subunit (CLPP) forms 2 stacked heptameric rings and interacts with hexameric ATPases caseinolytic mitochondrial matrix peptidase chaperone subunit B (CLPB) and caseinolytic mitochondrial matrix peptidase chaperone subunit X (CLPX) that serve as disaggregases and chaperones, respectively, contributing to protein unwinding.^10,12 Once unfolded, targets are degraded in the catalytic chamber formed by CLPP-containing rings.^9,10,12 Key substrates include proteins involved in glycolysis; metabolism of branched chain amino acids, fatty acids, and pyruvate; the Krebs cycle; and respiratory electron transport.¹³ CLPP hyperactivation by activating mutations or imipridone compounds (ie, ONC201) was described as selectively degrading respiratory chain protein substrates and disrupting mitochondrial structure and function in leukemia.¹⁴ Interestingly, CLPP inhibition has also been noted in leukemia to produce accumulation of misfolded or degraded respiratory chain subunits and cause respiratory dysfunction.¹⁵ In our own group’s previous myeloma studies, performed before ONC201’s CLPP activating role was known, we found it induced apoptosis in part through Bim upregulation.¹⁶ However, the role of CLP in myeloma pathobiology has not been fully studied, and inhibition as an antimyeloma strategy has not been examined.

In the current studies, we revisited CLPP biology after finding that it was overexpressed in transformed vs normal and symptomatic myeloma vs precursor PCs and that its overexpression associated with an inferior prognosis. Genetic and pharmacologic CLPP inhibition reduced colony formation and adhesion to stroma while activating apoptosis and increasing autophagy. Mechanistic studies showed that inhibition reduced mitochondrial complex I and II subunit levels and adenosine triphosphate (ATP) production and inhibited the glycolytic, pentose phosphate, tricarboxylic acid (TCA), and oxidative respiratory pathways. In contrast, reactive oxygen species (ROS) and proteasome activity were increased, and CLPP inhibition showed synergy with proteasome inhibitors (PIs). Finally, CLPP inhibition showed efficacy against primary PCs and in a myeloma cell line–derived xenograft. These studies broaden our understanding of the role of mitochondrial CLP in myeloma pathobiology and support development and translation of strategies targeting CLP to the clinic.

The CLPP inhibitor A2-32-01 used at previously reported concentrations¹³ and the p300/CREB-binding protein (CBP) bromodomain inhibitor inobrodib were from MedChemExpress (Monmouth Junction, NJ), with additional details provided in the supplemental Methods, available on the Blood website.

Human-derived myeloma cell lines were sourced and validated as detailed in the supplemental Methods. For single-cell RNA sequencing, primary samples were collected from normal donors (n = 3) and patients with monoclonal gammopathy of undetermined significance (n = 13), smoldering myeloma (n = 22), newly diagnosed multiple myeloma (NDMM; n = 17), relapsed and/or refractory myeloma multiple (RRMM; n = 15), and PC leukemia (n = 2) using institutional review board–approved protocols after informed consent was obtained in accordance with the Declaration of Helsinki. Whole bone marrow tissues were collected and enriched using the EasySep CD138 Positive Selection Kit (STEMCELL Technologies, Vancouver, BC).

We evaluated the transcriptome and B-cell receptor variable (variable diversity joining [VDJ]) region from freshly enriched PCs using the 10x Genomics platform (Pleasanton, CA) as detailed previously.¹⁷ Briefly, Cell Ranger (v3.1.0) was used to align and count sample transcripts, whereas normalization, scaling, and linear dimension reduction were performed using Seurat (version 4.3.0).¹⁸ Shared, identical heavy and/or light chain VDJ sequences defined monoclonal PCs, whereas diverse VDJ sequences at low frequencies (<0.1%) defined polyclonal PCs. Differential expression analysis between monoclonal PCs and polyclonal PCs based on clinical features was conducted using FindAllMarkers with Seurat and a pseudobulk approach by DESeq2 (version 1.34).¹⁹ In the last one, unique molecular identifiers per gene were aggregated by PC population and not by sample. To identify cells with stemlike characteristics, we used CytoTRACE (Cellular [Cyto] Trajectory Reconstruction Analysis using gene Counts and Expression) in a downsampled Seurat object (maximum of 3000 cells per sample).²⁰ The primary data are available at the European Genome-phenome Archive (EGAC50000000271).

A lentiviral doxycycline-regulated system was used as described in the supplemental Methods.

Mass spectrometry analysis was performed using the established tandem mass spectrometry (MS/MS)–based method,²¹ and analyses were conducted as outlined previously²¹ and in the supplemental Methods.

Details are provided in the supplemental Methods.

These were performed as described previously^22,23 and in the supplemental Methods.

These were performed using established procedures^24,25 as described previously^26-28 and in the supplemental Methods.

These assays were performed as provided in previous studies^23,29 and the supplemental Methods.

Organoid cultures were conducted following established procedures³⁰ detailed in the supplemental Methods.

Assays were performed using the Seahorse XF Real-Time ATP Rate Assay Kit from Agilent as described previously³¹ and detailed in the supplemental Methods.

MM1.S-Luc cells were xenografted and treated as detailed in the supplemental Methods and monitored by bioluminescent imaging and lambda light chain enzyme-linked immunosorbent assays (Bethyl Laboratories, Montgomery, TX).³² 

CD138⁺ bone marrow aspirate–derived cells resuspended in RPMI 1640 medium with 10% fetal bovine serum were seeded into 24-well plates. A2-32-01 or vehicle was applied for 2 days and apoptotic cells were assessed using Pacific Blue annexin V staining with propidium iodide.

Data are represented as the mean plus standard deviation (for triplicate data from the same experiment) or standard error (for multiple independent experiments). The significance of drug-effect relationships was determined by 1-tailed unpaired t tests or analysis of variance using GraphPad Prism, with P < .05 considered significant.

To identify novel genes contributing to myeloma pathobiology, we performed single-cell 5’-transcriptome sequencing on PCs from patients spanning the spectrum from normal to advanced disease. Monoclonal/neoplastic PCs were differentiated from polyclonal/normal PCs based on B-cell receptor/VDJ sequences, allowing each patient to serve as their own control.^33,CLPP expression was low in normal CD138⁺ cells, rose in precursors (monoclonal gammopathy of undetermined significance and smoldering myeloma), and was higher in NDMM and RRMM (Figure 1A). Overall, there was a significant difference between CLPP in asymptomatic and symptomatic states (P = .00755; Figure 1B), but not for CLPB or CLPX (supplemental Figure 1A-C). Similar associations with more advanced disease were seen in publicly available gene expression profiling (GEP) databases for CLPP but not CLPB or CLPX (supplemental Figure 1D). Increased CLPP levels were seen in myeloma vs other tumor lines (supplemental Figure 2A), and increased CLPP protein was seen in myeloma patient plasma (supplemental Figure 2B). Uniform manifold approximation and projection plots showed CLPP expression heterogeneity in individual patients (Figure 1C). CytoTRACE analysis²⁰ to examine differentiation states and stemness revealed a correlation between elevated CLPP and high CytoTRACE values (Figure 1D). Next, we examined outcomes for patients with NDMM in the Multiple Myeloma Research Foundation CoMMpass Study. High CLPP (highest quartile [Q4]) was associated with some markers of increased disease burden (Table 1) and an inferior progression-free survival (Q4 median, 2.11 years vs 3.28 for Q1 + Q2 + Q3) and overall survival (OS; Q4 median 5.58 years vs not reached for Q1 + Q2 + Q3; Figure 1E). Similarly, high CLPP associated with an inferior event-free survival (P = .0207) and OS (P = 2.48 × 10^–8) in the University of Arkansas for Medical Sciences database³⁴ (Figure 1F; median OS, 11.07 years for the low, 6.55 years for the high CLPP group, respectively). Within the GEP70 defined molecular subgroups, CLPP was highest in the high-risk proliferative (PR) subtype and lowest in the standard-risk CD2 subgroup (supplemental Figure 3A). More broadly, high CLPP levels were associated with a high-risk GEP70 score (supplemental Figure 3B) and with poor outcomes in the PR subgroup (supplemental Figure 3C), where CLPP expression was highest, and the MS subgroup defined by overexpression of FGFR3 and MMSET (supplemental Figure 3D), with average CLPP levels. Interestingly, high CLPP levels identified a group of patients with RRMM receiving bortezomib-based therapy with a shorter progression-free survival (Figure 1G). In addition, in patients with paired samples, relapse was associated with higher CLPP expression in the CoMMpass (supplemental Figure 3E; P < .05) and University of Arkansas for Medical Sciences (Figure 1H; P = 1.9 × 10^–5) databases, but not for CLPX (supplemental Figure 3E-F). Finally, patients with extramedullary disease at baseline or relapse³⁵ had higher CLPP levels than those without (Figure 1I), which again was not the case for CLPX (supplemental Figure 3G).

To examine the role of CLPP in pathobiology, protein levels were analyzed in cell lines (supplemental Figure 4A), and we generated MM1.S and H929 cells with a doxycycline-inducible CLPP-targeted short hairpin RNA (shRNA) and JJN-3 and H929 cells with inducible overexpression (supplemental Figure 4B). In an organoid assay, doxycycline did not affect growth in control shRNA-containing cells, but CLPP knockdown, which inhibited CLPP activity (supplemental Figure 4C), significantly reduced large (≥100 mm) colony formation (Figure 2A-B). In contrast, overexpression increased medium and large colonies and accelerated JJN-3 and H929 cell growth (Figure 2A-B; supplemental Figure 4D). Therefore, we examined the impact of CLPP inhibition with A2-32-01,³⁶ which reduced viability in a concentration-dependent fashion (Figure 2C) in myeloma cells representing different molecular subtypes with median inhibitory concentrations from 50 to 150 μM (supplemental Table 1). Although there was not a strong correlation between CLPP expression and inhibitor sensitivity across cell lines (data not shown), CLPP overexpression in JJN-3 and H929 cells enhanced A2-32-01 sensitivity (supplemental Table 2). Cell cycle analysis in MM1.S and H929 cells showed predominantly G1/S arrest (Figure 2D), and A2-32-01 induced a time- and concentration-dependent increase in apoptosis (Figure 2E). As confirmation that this impact was caused specifically by CLPP inhibition, we returned to our inducible knockdown models and found that 2 different shRNAs enhanced annexin V positivity in myeloma cells (Figure 2F) vs controls. Induction of apoptosis was also confirmed by western blotting for cleaved caspase 3 formation (supplemental Figure 5).

Given the activation of apoptosis and caspases by CLPP inhibition, we next examined ROS, and in MM1.S and H929 cells, A2-32-01 produced a concentration- and time-dependent ROS increase (Figure 3A). Similarly, CLPP knockdown increased ROS significantly vs uninduced cells and compared with cells with nontargeted shRNAs (Figure 3B). Next, we studied the mitochondrial membrane potential (ΔΨM), which decreased in response to CLPP inhibition (Figure 3C) and knockdown (supplemental Figure 6A). Given that ROS can influence cancer cell migration and invasion,³⁷ we examined adhesiveness to HS-5 stromal cells and found that CLPP inhibition (Figure 3D) and knockdown (supplemental Figure 6B) reduced MM1.S and H929 adherence. ROS can also trigger UPP activation to eliminate oxidized and damaged proteins,³⁸ so we examined the impact of CLPP suppression on proteasome activity. After either pharmacologic inhibition (Figure 3E) or genetic knockdown (supplemental Figure 6C), myeloma cells exhibited enhanced proteasome activity. To determine whether ROS were key intermediates, we preincubated cells with N-acetyl cysteine (NAC) and found NAC reduced both the loss of adhesiveness (supplemental Figure 7A) and proteasome activation (supplemental Figure 7B).

Owing to the role of CLP in protein turnover, we next performed mass spectrometry studies on MM1.S cells exposed to A2-32-01 or vehicle and, in our inducible model, to broaden understanding of its contribution to myeloma biology. Data mapping using gene set enrichment analysis and the Kyoto Encyclopedia of Genes and Genomes and REACTOME Pathway databases identified early decreases in E2F targets, cell cycle, and G2/M checkpoint proteins (supplemental Figure 8) consistent with cell cycle arrest. Gene set enrichment analysis also identified disrupted cholesterol homeostasis (supplemental Figure 9) and upregulation of proteins in gene sets involved in lytic vacuole, membrane trafficking, and vesicle organization (Figure 4A) as prominent changes. Proteins of note (Figure 4B) included many critical for lysosomal integrity and degradation (ie, multiple cathepsins, galactosidases, and lysosomal-associated membrane proteins) and for mitophagy (ie, translocases of outer mitochondrial membrane and p62/sequestosome 1). To determine whether there was direct evidence of autophagy, cells were stained with a dye that is incorporated into preautophagosomes, autophagosomes, and autolysosomes. In both MM1.S and H929 cells, analysis showed a significant time- and concentration-dependent increase in these vacuoles with either pharmacologic CLPP inhibition (Figure 4C) or shRNA-mediated CLPP knockdown (supplemental Figure 10A). Mitophagy was detected using a dye that fluoresces when damaged mitochondria fuse to lysosomes and was also increased owing to inhibition (Figure 4D) or knockdown (supplemental Figure 10B). As another indicator of autophagy and mitophagy, we looked at levels of LONP1, which were significantly elevated after either pharmacologic (supplemental Figure 11A) or genetic CLP suppression (supplemental Figure 11B). Also of note, myeloma cells in which CLPP was inhibited (Figure 4E) or suppressed (Figure 4F) showed significant accumulation of microtubule-associated proteins 1A/1B light chain 3β-phosphatidylethanolamine conjugate (LC3-II) as a further confirmation of autophagy.

To begin to understand some of the mechanisms underlying these changes, we repeated selected studies in the presence of NAC, which, as expected, reduced ROS (supplemental Figure 12A). By scavenging ROS, NAC preserved or even enhanced the ΔΨM (supplemental Figure 12B), significantly reduced apoptosis (supplemental Figure 12C), and also inhibited but did not abrogate autophagy (supplemental Figure 12D). In addition, we performed similar studies with excess exogenous ATP to see the impact of blunting mitochondrial dysfunction. ATP reduced ROS production (supplemental Figure 12A) and ΔΨM loss (supplemental Figure 12B), but did not reduce apoptosis (supplemental Figure 12C) and only partially reduced autophagy in the presence of A2-32-01 vs controls (supplemental Figure 12D).

Mitophagy activation supported the appearance of compromised mitochondrial function and cellular energy homeostasis, so we next looked at this consequence of CLPP inhibition. In our mass spectrometry database, subunit abundance levels of 5 electron transport chain complexes and mitochondrial translation (supplemental Table 3) declined upon CLPP inhibition or knockdown in MM1.S cells, with complex I and IV most affected (Figure 4G). Along with these changes, total cellular ATP declined in MM1.S cells exposed to A2-32-01, especially at higher inhibitor concentrations (Figure 4H), and with CLPP knockdown (Figure 4I). In addition, we leveraged the Seahorse assay and found that CLPP inhibition affected mitochondrial ATP production in a manner dependent on time and A2-32-01 concentration in MM1.S (Figure 4I) and H929 cells (supplemental Figure 13).

Given that impaired mitochondrial function could have broader-reaching effects than just on ATP production, we next performed metabolomic analyses. Ion chromatography–coupled high-resolution mass spectrometry of 301 polar metabolites revealed significant changes in MM1.S cells exposed to A2-32-01 and after induction of CLPP-targeted shRNAs (Figure 5A). A total of 16 and 44 metabolites were upregulated during at least 2 time points at 50 or 250 mM, respectively (Figure 5B), whereas 15 and 42 were downregulated during at least 2 time points (Figure 5C). Unsupervised clustering analysis revealed that A2-32-01 depleted metabolites representative of glycolytic activity such as fructose-1-phosphate and glycerates, especially at higher drug concentrations (Figure 5D, top and middle panels). Similarly, CLPP suppression also downregulated glycolytic pathway metabolites (Figure 5D, bottom panel). Consistent with its impact on ATP generation, both pharmacologic (Figure 5E, left panels) and genetic CLPP inhibition (Figure 5E, right panel) suppressed TCA cycle function and redox metabolism metabolites. Another dysregulated downstream target was the pentose phosphate pathway (PPP; Figure 5F) as indicated by reduced ribose-5-phosphate levels, a key DNA and RNA precursor. In contrast, proteomic data showed an increase in abundance of proteins involved in glycolysis, the TCA cycle, and fatty acid metabolism (supplemental Figure 14A), as well as metabolism of branched chain and other amino acids (supplemental Figure 14B).

To assess the potential that pharmacologic CLPP inhibition could be considered for clinical application, we studied an MM1.S-derived xenograft (Figure 6A). Compared with vehicle-treated mice, A2-32-01 significantly hindered tumor growth as determined by whole animal bioluminescent imaging (Figure 6B) and by a decline in human lambda light chain levels (Figure 6C). Given that novel drugs are typically first tested in the RRMM setting, we examined A2-32-01’s activity against a panel of cell lines with resistance to conventional and novel agents. Notably, the median inhibitory concentration of A2-32-01 was comparable in resistant and wild-type counterparts across models of alkylating agent, immunomodulatory drug, and PI resistance (supplemental Table 4). Moreover, A2-32-01 significantly increased annexin V positivity in CD138⁺ PCs from patients with NDMM or RRMM (Figure 6D-E). Interestingly, when we looked at the impact of CLPP overexpression, we found that this consistently conferred resistance to melphalan, but not PIs (supplemental Figure 15).

A2-32-01 is a useful tool but there is not yet an available CLPP inhibitor that can be translated to the clinic. Using the Ominer tool (http://www.signalingpathways.org/ominer/query), we identified multiple E1A binding protein p300/CBP consensus sites in the CLPP, CLPB, and CLPX promoters (supplemental Figure 16). Using the EP300/CBP bromodomain inhibitor CCS-1477 (inobrodib),³⁹ we found that it reduced CLPP (Figure 6F) and CLPX expression (Figure 6G) and binding of EP300/CBP to the CLPX promoter by chromatin immunoprecipitation (Figure 6H). Inobrodib reduced myeloma cell viability (supplemental Figure 17), and interestingly, although CCS-1477 did not reduce CLPP expression, it did significantly reduce CLPX levels (Figure 6I), perhaps caused in part by the latter’s shorter half-life (supplemental Figure 18). Given that CLPX is needed to unfold target proteins and translocate them in an ATP-dependent fashion into the CLPP catalytic chamber, CLPX reduction could be sufficient to inhibit the activity of the overall complex. Consistent with this possibility, inobrodib enhanced the accumulation of mitochondrial phosphoenolpyruvate carboxykinase 2 (Figure 6I), a known CLPP substrate.¹³ 

Given that we had previously linked CLPP inhibition to autophagy and the latter can have a cytoprotective function, we examined the impact of chloroquine, which inhibits autophagic flux. A2-32-01 and chloroquine together showed enhanced antiproliferative effects (Figure 7A), and combination index analysis revealed these 2 were additive to synergistic under many conditions (Figure 7B). Next, owing to the impact of CLPP inhibition on intermediary metabolism, we looked at A2-32-01 with the glycolysis/PPP inhibitor 6-aminonicatinamide. These together also induced enhanced effects (Figure 7C) that showed largely synergistic interactions (Figure 7D). Finally, given that antimyeloma agents are often used as part of rationally designed combinations, we examined the possibility that, because CLPP inhibition activated the UPP, blockade of both pathways could be attractive. Consistent with this possibility, compared with bortezomib alone (supplemental Figure 19A), A2-32-01 with bortezomib produced enhanced antiproliferative effects (Figure 7E). When carfilzomib was used as the PI, similarly enhanced efficacy was seen (Figure 7F) compared with carfilzomib alone (supplemental Figure 19B). In addition, combination index analysis revealed evidence for synergy between A2-32-01 and bortezomib (Figure 7G) or carfilzomib (Figure 7H) for many of the conditions. This provides a rationale for translation to the clinic of dual approaches targeting both of these key proteolytic pathways for therapeutic benefit against myeloma.

Antibody production and secretion by normal and neoplastic PCs are intensive processes that require glucose and fatty acids to generate energy, and amino acids and glucose for immunoglobulin synthesis, glycosylation, and proper folding.⁴⁰ These processes generate misfolded proteins and ROS, with the latter further contributing to oxidative stress and protein damage, inducing greater dependence on mechanisms to support cell survival including autophagy, metabolic/mitochondrial reprogramming, and the UPR.^41-43 Proteolysis through the CLP endopeptidase has in other models been shown to contribute to these processes through multiple mechanisms, starting with the UPR given that CLPP is part of the mitochondrial protein quality control machinery⁴⁴ and may work cooperatively with LONP1.⁴⁵ CLPP with LONP1 also degrades the complex I ROS-generating domain, thereby reducing potentially toxic ROS levels.⁴⁶ In plant models, conditional CLPP depletion resulted in protein quality control gene activation and an autophagy-like response.⁴⁷ Finally, CLPP function promotes oxidative phosphorylation and ATP production,⁴⁸ contributing to metabolic adaptations to hypoxia. This background provides strong support for the possibility that CLP would also be a critical contributor to myeloma pathobiology, where its role has not been well studied.

CLPP first came to our attention during studies of publicly available databases and our own single-cell transcriptomic studies seeking to identify genes that could be differentially expressed in advanced myeloma vs precursors. This led us to examine the impact of CLPP on patient outcomes, and we found that high expression was associated with inferior outcomes in the newly diagnosed and RRMM settings. Moreover, although especially high CLPP levels were seen in the poor-risk PR molecular subtype of the GEP70 classification system, high expression in other subtypes was also associated with inferior outcomes. Considering the role of CLP in supporting multiple pathways that enhance PC fitness, including the metabolic reprogramming that is key to myeloma survival,⁴⁹ this is perhaps not surprising. In addition, given that metabolic reprogramming of tumor cells and their microenvironment has been linked to drug resistance,⁵⁰ this is likely another mechanism through which enhanced CLPP function confers a worse prognosis. Indeed, in serial samples, CLPP was expressed at higher levels than at baseline, further suggesting a contribution to drug resistance. This also is supported by our finding that CLPP overexpression conferred melphalan resistance, which may be because resistance to this alkylating agent has been linked to the Warburg effect and an elevated oxidative stress response,⁵¹ both of which would be promoted by CLP activity. Notably, because some patients even with “good risk” myeloma subtypes may have inferior outcomes, high CLPP expression could be worthy of further study as a candidate biomarker to stratify those with better or worse prognoses. Moreover, given that studies in AML^14,15 and in our models identified higher CLPP expression as predicting for greater sensitivity to CLPP inhibition, there could be a path forward for clinically relevant approaches that target CLPP in this higher risk setting.

Mechanistically, CLPP inhibition was associated with ROS increases and decreased subunit abundance and function of multiple complex I components. ROS is likely an early messenger considering the ability of NAC as a scavenger to inhibit downstream changes in ΔΨM and apoptosis, although ATP supplementation only inhibited the latter weakly. This could be a consequence of the fact that, even without ROS, CLP endopeptidase inhibition would still affect mitochondrial protein quality control and activate the mitochondrial UPR. These ROS-related findings are consistent with the known role of the CLP endopeptidase in mitoribosomal assembly,⁵² which is necessary for translation of 7 mitochondrial genes that contribute proteins to complex I.⁵³ Moreover, as previously noted, CLPP also plays a direct role in reducing ROS formation,⁴⁶ which would be lost with CLPP inhibition. Activation of autophagy and mitophagy was seen, which occurs likely as an initial cellular survival response to remove dysfunctional mitochondria and is supported by data showing chloroquine with A2-32-01 enhances cell death. However, autophagy then likely further contributes to the loss of complex I protein levels through mitochondrial degradation, and we also noted increased LONP1 levels, which may degrade complex I subunits that become more misfolded and/or oxidized due to increased ROS. Loss of the mitochondrial transmembrane potential compromises ATP production and induces cell cycle arrest and apoptosis while forcing the cell to try to adapt by shifting to alternative energy sources, probably in part mediated by increasing ROS levels.⁵⁴ Among these are lipid and fatty acid pathways, the TCA cycle, and amino acid metabolism, leading to the depletion of metabolic intermediates involved in glycolysis, the PPP, and oxidation/reduction. Ultimately, given that these adaptations produce less energy and are unsuccessful in restoring redox balance, myeloma cells then undergo terminal apoptosis.

Inhibition of CLPP proved to be effective in inducing cell death in both drug-naïve and drug-resistant models and both in vitro and in vivo. Although clinically relevant CLP endopeptidase inhibitors are not yet available, there is interest in such agents as antitumor drugs^14,15 and antibacterials,⁵⁵ indicating that future translation to the clinic could be possible. In the meantime, we noted that inobrodib reduced CLPP and CLPX messenger RNA levels and reduced CLP endopeptidase activity, likely in part through decreased CLPX levels given that the latter is needed to unfold proteins before degradation by CLPP. Because this drug is already in the clinic and has shown preliminary evidence of activity against both AML and RRMM,³⁹ our data suggest the possibility that downregulation of CLPP function may be one of this agent’s mechanisms of action. Interestingly, inobrodib was found to have additive to synergistic activity with the PI bortezomib preclinically,³⁹ which we also found in our studies of bortezomib or carfilzomib with a direct CLPP inhibitor. Given the strong role of PIs in our antimyeloma chemotherapeutic armamentarium, these findings support the possible use of a dual strategy combining PIs with either direct or indirect CLPP inhibitors, especially in patients with poor prognosis and a CLPP-overexpressing disease.

The authors thank the MD Anderson Flow Cytometry and Animal Imaging Core Facility and the Characterized Cell Line Core Facility, which are supported by the Cancer Center Support Grant (P30 CA16672). Proteomics were performed by the Baylor Mass Spectrometry Proteomics Core, which is supported by the Dan L Duncan Comprehensive Cancer Center Support Grant (P30 CA125123), Core Facility Awards Cancer Prevention and Research Institute of Texas (RP170005 and RP210227), Intellectual and Developmental Disabilities Research Center Award (P50 HD103555), and National Institutes of Health High End Instrument Award (S10 OD026804, Orbitrap Exploris 480).

R.Z.O. acknowledges support from the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation, the Riney Family Multiple Myeloma Research Fund at MD Anderson from the Paula and Rodger Riney Foundation, the Leukemia & Lymphoma Society (SCOR-12206-17), the MD Anderson Cancer Center High Risk Multiple Myeloma Moon Shot, and the Brock Family Myeloma Research Fund. Additional support came from the Myeloma Solutions Fund, the Jake and Nina Kamin Endowment for Multiple Myeloma Research, the Alexanian Fellowship, and the James B. & Marie R. Baker and David & Sara Anne Baker Hopkins Research Endowment. L.Y.M.R. acknowledges support from a Multiple Myeloma Research Foundation 2022 Research Fellow Award. H.C.L. acknowledges support from the Leukemia & Lymphoma Society (SCOR-12206-17) and the Riney Family Multiple Myeloma Research Fund at MD Anderson from the Paula and Rodger Riney Foundation. K.K.P. acknowledges support from the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation and the Riney Family Multiple Myeloma Research Fund at MD Anderson from the Paula and Rodger Riney Foundation.

Contribution: L.Q. performed most of the studies described herein with contributions from L.Y.M.R., U.R., H.W., and I.K.; L.Q. and R.Z.O. drafted the manuscript; Q.Y. and D.E.S. provided key input; U.R., I.M., L.T., S.L., H.C.L., D.E.S., F.Z., and J.D.S. provided helpful input; L.Q. and R.Z.O. finalized the work; L.Y.M.R., S.L., and H.L. provided bioinformatic and statistical analyses; E.E.M., H.C.L., and K.K.P. provided patient samples; M.J.A.C. processed the samples; I.M., L.T., and P.L.L. performed metabolomic studies and analyses; and D.E.M., F.Z., and J.D.S. performed gene expression profiling analyses of the University of Arkansas for Medical Sciences data.

Conflict-of-interest disclosure: H.C.L. has provided consultancy services to Amgen, Inc, Celgene, a wholly owned subsidiary of Bristol Myers Squibb, GlaxoSmithKline (GSK), Janssen Pharmaceutical, Sanofi-Aventis, and Takeda Pharmaceutical and has received research funding from Amgen, Inc, Celgene, a wholly owned subsidiary of Bristol Myers Squibb, Daiichi Sankyo, GSK, Janssen Pharmaceutical, and Takeda Pharmaceuticals. K.K.P. reports research support from Celgene, a wholly owned subsidiary of Bristol Myers Squibb. R.Z.O. declares research funding unrelated to this work from Heidelberg Pharma AG, Asylia Therapeutics, and Biotheryx; has served on advisory boards for Amgen, Inc, Bristol Myers Squibb, Celgene, EcoR1 Capital LLC, Forma Therapeutics, Genzyme, GSK Biologicals, Ionis Pharmaceuticals, Inc, Janssen Biotech, Juno Therapeutics, Kite Pharma, Legend Biotech USA, Molecular Partners, Sanofi-Aventis, Servier, and Takeda Pharmaceuticals North America, Inc; and is a founder of Asylia Therapeutics, Inc, with an equity interest. The remaining authors declare no competing financial interests.

Correspondence: Robert Z. Orlowski, Department of Lymphoma and Myeloma, The University of Texas MD Anderson Cancer Center, 1515 Holcombe Blvd, Unit 429, Houston, TX 77030-4009; email: rorlowski@mdanderson.org.