Olaparib-Induced Senescence Is Bypassed through G₂–M Checkpoint Override in Olaparib-Resistant Prostate Cancer

Treatment of advanced prostate cancer remains challenging due to a paucity of durable therapeutic options. Inhibition of PARP has recently emerged as an exciting new treatment strategy for biomarker selected patients with cancer (1). PARP is responsible for mediating repair of single strand DNA breaks. Inhibition of PARP is thought to cause unrepaired single strand breaks to lead to increased double-strand breaks which must be repaired using other mechanisms such as homologous recombination. It has also been shown that PARP inhibition can cause PARP to become “trapped” at sites of DNA damage, ultimately resulting in DNA breaks (2). In those tumors which harbor defects in certain DNA repair genes, such as BRCA1 and BRCA2, PARP inhibition is hypothesized to cause undue cellular stress leading to cell death in a synthetic lethal manner. In 2005, both Farmer and colleagues and Bryant and colleagues independently determined that tumors deficient in homologous recombination are especially susceptible to PARP inhibition (3, 4). Since then, PARP inhibitors (PARPi) have been further studied and developed to translate this exciting and specific treatment strategy to the clinic.

Several trials have demonstrated the efficacy of PARPis for the treatment of metastatic castration-resistant prostate cancer (mCRPC). The TOPARP-A phase II clinical trial demonstrated that PARP inhibition using olaparib appeared efficacious in a subset of patients with mCRPC, the majority of which (14 of 16 responders) harbored mutations in DNA repair genes such as BRCA2 (5). These results were confirmed in the phase II TOPARP-B study which aimed to test olaparib in prospectively chosen patients with mCRPC harboring DNA damage response gene defects (6). The phase III PROfound trial demonstrated that patients harboring DNA damage repair defects (most notably BRCA1, BRCA2, and ATM) receiving olaparib fared better than those on investigators choice of either enzalutamide or abiraterone (7, 8). The TRITON2 study demonstrated that rucaparib has significant antitumor activity in patients with mCRPC with a BRCA deleterious alteration (9). In May 2020, both the PARPis rucaparib and olaparib were approved for the treatment of biomarker selected patients with mCRPC, representing the dawn of synthetic lethal, precision medicine approaches for the management of this disease. Additional PARPis in clinical development for prostate cancer include niraparib and talazoparib (10). However, it should be noted that biomarkers for optimal patient selection for treatment with a PARPi remain poorly understood.

Despite the promise of PARP inhibition for enhancing clinical management of patients with prostate cancer, several key questions remain to be answered to optimize their utility including (i) how specifically do PARPi-sensitive tumor cells respond to treatment, and (ii) how does PARPi resistance develop? It is thought that PARPis induce excessive DNA damage in tumor cells ill-equipped to repair double-strand breaks, leading to cell death. However, a full understanding of PARPi mechanism of action remains lacking. The signaling cascades which mediate response are not well understood and whether there are alternative responses to treatment is unknown. A better understanding of how PARPis function promises to enhance their clinical utility.

The development of therapeutic resistance appears inevitable with treatments used for late-stage prostate cancer. Already, PARPi resistance has been clinically documented in patients with prostate cancer with contributing mechanisms appearing to include (i) reversion mutations which replace lost function of defective DNA repair genes and possibly (ii) the emergence of a neuroendocrine phenotype (11–13). Basic research studies have also determined additional possible mechanisms of resistance which may include overexpression of the drug efflux transporter ABCB1 and overexpression of PLK1 (14–16). Understanding the development of clinical PARPi resistance will greatly augment the use of PARPi drugs through (i) prolonging their utility and (ii) helping to aid stratification of patients most likely to respond to treatment. Clinically relevant model systems for the study of PARPi response and resistance are needed to aid efforts to address critical, unanswered questions.

In the current study, we introduce two cell line models of olaparib resistance derived from LNCaP cells and its castration-resistant derivative cell line, C4-2B. Olaparib-resistant, OlapR, cell lines are shown to be highly insensitive to olaparib and other clinically relevant PARPis when compared with respective parental control cells. Interrogation of response to treatment in sensitive, parental cells shows that olaparib induces massive DNA damage which elicits a potent p21-mediated G₂–M arrest accompanied by both apoptosis and senescence. Interestingly, OlapR cells exhibit a blunted response to treatment which can be overcome through combination treatment with a CDK1 inhibitor. Our study establishes a cell line platform system to further study both response to treatment and the development of resistance.

C4-2B cells were kindly provided and authenticated by Dr. Leland Chung (Cedars-Sinai Medical Center, Los Angeles, CA). LNCaP, Rv1, and PC3 cells were obtained from the American Type Culture Collection (ATCC). ATCC uses short tandem repeat profiling for testing and authentication of cell lines. All cell lines are routinely tested for mycoplasma using ABM mycoplasma PCR detection kit (Cat#: G238). All experiments with these cell lines and their derivatives were conducted within 6 months of receipt or resuscitation after cryopreservation. LNCaP, C4-2B, Rv1, and PC3 cells were maintained in RPMI 1640 media supplemented with 10% FBS, 100 IU penicillin, and 0.1 mg/mL streptomycin. Olaparib-resistant LN-OlapR and 2B-OlapR cells were derived from LNCaP and C4-2B cells respectively through chronic exposure to increasing doses of olaparib for over 1 year. Both OlapR cell lines are maintained in complete RPMI 1640 supplemented with 5 μmol/L olaparib. LNCaP and C4-2B cells were cultured alongside OlapR derivative cell lines as appropriate controls. All cells were maintained at 37ºC in a humidified incubator with 5% carbon dioxide. Olaparib (Cat#: S1060), niraparib (Cat#: S2741), talazoparib (Cat#: S7048), and BMS-265246 (Cat#: S2014) were purchased from Selleckchem (17). Rucaparib (Cat#: 6230/10) was purchased from R&D Systems. RNAi was performed using Dicer-Substrate siRNAs (DsiRNA) purchased from IDT. The following DsiRNAs were used: Negative Control (NC) = Cat#: 51-01-14-04, p21 = Cat#: hs.Ri.CDKN1A.13.1, CDK1 = Cat#: hs.Ri.CDK1.13.1. Transfection of DsiRNAs was done using Lipofectamine RNAiMAX (Cat#: 56532) purchased from Thermo Fisher Scientific according to the manufacturer's protocol with 25 or 50 nmol/L DsiRNA. Microscopy was performed using a Keyence BZ-X810 imaging system.

Cells were plated at a density of 10,000 to 40,000 cells/well in 24-well plates in complete media without any selection agent. After 24 hours, cells were subjected to indicated treatments. For experiments combining RNAi with drug, transfection of DsiRNA was performed 24 hours after plating and drug was administered as indicated the following day. Cell viability was assessed 72 to 120 hours after drug treatment using Cell Counting Kit-8 (Cat#: CK04-20) purchased from Dojindo Molecular Technologies, Inc. Alternatively, cells were counted via Coulter Particle Counter. Data are displayed as percent of control cell growth ± SD. All conditions were performed in triplicate or quadruplicate.

Cells were harvested, washed with PBS, and lysed in RIPA buffer supplemented with 5 mmol/L EDTA, 1 mmol/L NaV, 10 mmol/L NaF, and 1X Halt Protease Inhibitor Cocktail (Cat#: 78430) purchased from Thermo Fisher Scientific. Protein concentration was determined with Pierce Coomassie Plus (Bradford) Assay Kit (Cat#: 23236) purchased from Thermo Fisher Scientific.

Protein extracts were resolved by SDS-PAGE and indicated primary antibodies were used; PARP (Cat# 9532, rabbit-monoclonal, 1:1,000 dilution), yH2AX (Cat# 9718, rabbit-monoclonal, 1:1,000 dilution), phospho-p53 Ser15 (Cat# 9284, rabbit-polyclonal, 1:1,000 dilution), Bax (Cat# 2772, rabbit-polyclonal, 1:1,000 dilution), cleaved-PARP (Cat# 9541, rabbit-polyclonal, 1:1,000 dilution), and CDK1 (Cat# 9116, mouse-monoclonal, 1:1,000 dilution) were purchased from Cell Signaling Technology. PAR (Cat#: 4336-BPC-100, rabbit-polyclonal, 1:1,000 dilution) was purchased from Trevigen. p21 (Cat# SC-397, rabbit-polyclonal, 1:1,000 dilution) and Actin (Cat# SC-1616, goat-polyclonal, 1:1,000 dilution) were purchased from Santa Cruz Biotechnology. Tubulin (T5168, mouse-monoclonal, 1:6,000 dilution) was purchased from Sigma-Aldrich. Tubulin or Actin was used to monitor the amounts of samples applied. Proteins were visualized with a chemiluminescence detection system (Cat#: WBLUR0500) purchased from Millipore.

Cell-cycle distribution was determined by propidium iodide staining analysis using a GUAVA MUSE cell analyzer flow cytometer (Luminex). Propidium iodide stain (Cat#: ab139418) was purchased from Abcam and staining was done following manufacturers protocol. Briefly, cells were treated as indicated for 5 days followed by harvest, wash, and fixed overnight in 66% EtOH at 4°C. Cells were washed the following day and stained with propidium iodide staining solution containing 100 μg/mL RNase A and subsequently run and analyzed via GUAVA MUSE. Data are displayed as percent of cells in each phase ± SD. All conditions were performed in triplicate.

Beta-galactosidase activity was assessed using staining kit (Cat#: 9860S) purchased from Cell Signaling Technology using manufacturers protocol. Briefly, cells were treated as indicated for 5 days, washed, fixed, and stained overnight in a dry, 37°C incubator. Stained cells were subsequently imaged the following day and images were used for quantification via percent of positively stained cells in several fields per image. Data are displayed as an average percent of positively stained cells per field ± SD.

Significance was assessed using a two tailed two sample equal variance Student t test. A P value ≤ 0.05 was accepted as significant. All experiments were performed at least twice.

The data generated in this study are available within the article.

We began our study by determining response to olaparib in various cell line models of prostate cancer. We found that the castration-sensitive cell line, LNCaP, and its castration-resistant cell line derivative, C4-2B, exhibited a stronger response to olaparib than other tested cells (Fig. 1A). A previous study showed that LNCaP harbors a BRCA2 mutation likely to confer sensitivity to PARP inhibition making it and its derivative, C4-2B, attractive models for the study of PARPi sensitivity (18). To study olaparib resistance, we subjected both LNCaP and C4-2B cells to chronic exposure of increasing doses of olaparib over an approximately 1-year period (Fig. 1B). This resulted in the development of two, novel olaparib-resistant cell line derivatives, LN-OlapR from LNCaP cells and 2B-OlapR from C4-2B cells. Cell growth assays demonstrate that each OlapR cell line exhibits robust resistance to olaparib versus parental, olaparib-sensitive cell lines (Fig. 1C). Respective IC₅₀ values for each are as follows: LNCaP = 6 μmol/L, LN-OlapR = 18 μmol/L, C4-2B = 3 μmol/L, 2B-OlapR = 106 μmol/L. Imaging of each cell line reveals clear morphological differences in OlapR cells versus parental cells indicating robust alterations to cellular phenotype (Fig. 1D). Notably, both OlapR cell lines appear more rounded than corresponding parental cell lines.

As previously noted, additional PARPis are either approved or are undergoing clinical testing for the treatment of prostate cancer, including rucaparib, niraparib, and talazoparib. However, each PARPi has a unique structure and possesses varying abilities to both inhibit PARP catalytic activity and trap PARP onto DNA (2, 19). Whether resistance to olaparib induces PARPi cross-resistance is not known but may have important research and clinical implications. Interestingly, cell growth assays demonstrate that each OlapR cell line exhibits cross-resistance to other clinically relevant PARPis, including rucaparib, niraparib, and talazoparib (Fig. 2). These data suggest that OlapR cells represent useful models of pan-PARPi resistance and that it is likely that mechanisms which confer resistance to olaparib may confer resistance to other PARPis.

We sought to better characterize resistance in OlapR cells. PARP1 is the most abundant member of the PARP family and accounts for approximately 90% of PARylation activity (20, 21). We first asked whether altered PARP1 expression may be involved in the development of resistance. Western blots revealed no difference in PARP1 expression in OlapR cells versus control cells (Fig. 3A). We next tested whether altered PARP activity plays a role. We performed Western blots for PAR to assess olaparib's ability to inhibit PARP catalytic function in OlapR and parental cells. Western blots show that after 5-day olaparib treatment, PARylation is significantly inhibited in both OlapR and parental cells (Fig. 3B), suggesting that resistance is not dependent upon alteration to olaparib's ability to inhibit PARylation.

It is thought that PARPis induce DNA damage and subsequent death in cells harboring DNA repair defects. Western blots reveal a dramatic increase in yH2AX levels, a marker for double-strand DNA breaks, in olaparib-treated LNCaP and C4-2B cells (Fig. 3B). This is consistent with the currently accepted mechanism of action of olaparib. However, in OlapR cells, there is no increase in yH2AX levels when treated with olaparib (Fig. 3B). Interestingly, we do note that untreated, both OlapR cell lines possess increased yH2AX expression relative to parental cells. These data suggest that OlapR cells do not accumulate DNA damage in response to olaparib treatment, unlike sensitive parental LNCaP and C4-2B cells. However, PARPi-resistant cells appear to retain and manage a high level of residual DNA damage from chronic treatment.

Our data suggest that resistance to olaparib is not dependent on altered PARP expression or on inhibition of PARPi function, but rather downstream response to treatment. Our data further support that olaparib induces DNA damage in sensitive cells, but not OlapR cells, although OlapR cells appear to manage high levels of residual DNA damage accumulated during resistance acquisition. To further characterize response to PARP inhibition, we performed flow cytometry to assess cell-cycle distribution in response to olaparib. We found that treatment induces a robust G₂–M arrest in both LNCaP and C4-2B cells, suggestive of G₂–M cell-cycle checkpoint activation in response to induced DNA damage (Fig. 3C). However, in both OlapR models, we found no G₂–M arrest in response to treatment, suggesting resolution of G₂–M checkpoint activation in these cells. Altogether, our data suggest that olaparib induces a robust G₂–M arrest in sensitive LNCaP and C4-2B cells, but that OlapR cells evade PARPi response via override of G₂–M checkpoint activation, allowing for cell survival and management of high levels of residual DNA damage.

The G₂–M checkpoint can be activated and mediated by several well characterized cascades which may include activation of p53 and downstream signaling. Western blots show that olaparib-treated LNCaP and C4-2B cells display increased levels of Ser15 phosphorylated p53, indicating activation (Fig. 3D). Furthermore, we found increased levels of the pro-apoptotic, p53-target BAX and cleaved-PARP, suggesting induction of cell death (Fig. 3D; refs. 22–24). We found comparatively less p53 activation in OlapR cells and an absent or blunted apoptotic response. These data suggest that p53 signaling and upstream signal activation is blunted in OlapR cells in response to PARPi-induced DNA damage.

Close visual observation of olaparib-treated LNCaP and C4-2B cells suggested that although some cells undergo apoptosis, a significant population of cells appear to display a cytostatic response with an altered morphology (Fig. 4A). In contrast, we did not observe significantly altered morphology in olaparib-treated OlapR cells. We hypothesized that PARPi treatment may induce G₂–M arrested senescence, which could have important clinical implications. To characterize this phenotype, we performed beta-galactosidase activity assays (Fig. 4B). Staining for increased beta-galactosidase activity, a hallmark of senescent cells, indicates that olaparib treatment induces senescence in LNCaP and C4-2B cells, but to a much lesser extent in OlapR cells. Together, our results imply that (i) PARP inhibition induces G₂–M checkpoint activation resulting in increased p53 signaling leading to both cell death and senescence and (ii) PARPi-acquired resistance is mediated by G₂–M checkpoint override.

Treatment-induced senescence is thought to have important clinical implications. Thus, understanding treatment-induced senescence and the mechanisms responsible for this phenotype is key in better utilizing certain drugs. Senescence is thought to be maintained largely through the activity of cyclin-dependent kinase inhibitors such as p21 (25, 26). p21 is a critical target of p53, prompting us to first assess the possibility that olaparib-induced senescence is mediated by p21. Indeed, we found increased p21 expression in olaparib-treated LNCaP and C4-2B cells, but not in olaparib-treated OlapR cells (Fig. 5A).

Senescence may provide some cells an escape from PARPi cytotoxicity through cell-cycle arrest. These cells may then be a repository of surviving cells which may later promote treatment resistance (27). We hypothesized that inhibition of p21 would block entry into senescence and promote increased cell death in response to olaparib due to increased DNA damage during cellular proliferation. Western blots show that p21 inhibition causes increased DNA damage in combination with olaparib and marked increases in cleaved-PARP, suggesting robust apoptosis (Fig. 5B). RNAi-mediated inhibition of p21 expression led to enhanced efficacy of olaparib in reducing cellular viability (Fig. 5C). Furthermore, imaging of treated cells suggests increased cell death in lieu of senescence when p21 inhibition is combined with olaparib treatment (Fig. 5D). Our data support that senescence associated with p21 activity is an alternative PARPi response which may provide a repository of surviving cells which escape PARPi cytotoxicity and may give rise to resistant cell populations.

Although parental LNCaP and C4-2B cells undergo G₂–M checkpoint activation in response to olaparib, resistant OlapR models do not. Interestingly, despite override of the G₂–M checkpoint, both LN-OlapR and 2B-OlapR cells appear to have much higher, intrinsic DNA damage versus parental cells, even in the absence of treatment (Fig. 3B). These data suggest that olaparib-resistant cells (i) devise means to survive persistent DNA damage and/or (ii) upregulate DNA damage repair mechanisms. A previous report demonstrated that in the context of BRCA-wild-type tumor cells (presumably homologous recombination proficient), CDK1 inhibition could impair DNA damage repair and sensitize cells to PARP inhibition, while not sensitizing nontransformed cells (28). Although the mechanism which promotes survival despite persistent DNA damage remains to be elucidated, we hypothesized that CDK1 inhibition would impair DNA repair, further augment DNA damage, and sensitize OlapR cells to treatment. Indeed, we found that siRNA-mediated downregulation of CDK1 drastically increased the DNA damage marker yH2AX, as well as cleaved-PARP indicating induction of apoptosis (Fig. 6A). Furthermore, we found that CDK1 knockdown decreased OlapR cell viability and enhanced olaparib efficacy (Fig. 6B). Like RNAi, a CDK1 small-molecule inhibitor, BMS-265246 (BMS), also decreased OlapR cell viability, increased DNA damage, induced apoptosis, and sensitized to olaparib (Fig. 6C–E; ref. 17). Altogether, our data suggest that inhibition of CDK1 can impair DNA damage repair in olaparib-resistant cells, reduce cellular viability, and enhance olaparib efficacy.

PARP inhibition represents the dawn of precision of medicine for prostate cancer and ushers in an era of new strategies aimed at exploiting synthetic lethality in tumors. The year 2020 saw two PARPis, olaparib and rucaparib, approved for prostate cancer treatment, whereas additional PARPis remain in clinical development. This exciting new approach promises to significantly enhance our ability to prolong the lives of men suffering from this disease, but several questions remain to be answered regarding their use, including (i) how specifically PARP inhibitors function, (ii) how do PARPi-sensitive tumor cells respond to treatment, and (iii) how does PARPi resistance develop? In the present study, we characterize PARPi-sensitive prostate cancer cell lines and developed olaparib-resistant derivative cells (OlapR). OlapR cells are shown to be highly resistant to olaparib and to exhibit marked cross-resistance to other clinically relevant PARPis, including the approved rucaparib. Interrogation of response to treatment shows that olaparib induces massive DNA damage in sensitive cells which elicits a potent p21-mediated G₂–M arrest response accompanied by both apoptosis and senescence. Interestingly, OlapR cells exhibit a blunted response to treatment which can be overcome through combination treatment with a CDK1 inhibitor. It should be noted that a previous study presented data suggesting Rv1 may also harbor deleterious BRCA2 mutation, but not PC3 (18). We hypothesize that uncharacterized mechanisms of resistance render Rv1 less sensitive to treatment. Why LNCaP and C4-2B cells are so much more sensitive is not completely understood, but clearly, additional study is needed to fully elucidate markers of resistance and sensitivity to PARP inhibitors. Our study establishes a cellular platform system to further study both response to treatment and the development of resistance.

Our study showed that OlapR cells are robustly cross-resistant to other PARPis in prostate cancer clinical development. Thus, OlapR cells can be thought of as pan-PARPi resistant, which could enhance their utility as a research tool. These data suggest that mechanisms of resistance to one PARPi will often induce resistance to the drug-class, which indicates that sequencing additional PARPis post initial treatment is likely a poor management strategy. Despite this finding, it is known that different PARPis have distinct structures and varying strengths and properties (2, 19). Most notably, it is known that PARPis both inhibit PARP catalytic activity and trap PARP at sites of DNA damage. Trapping is thought to be more indicative of anti-tumor activity and different PARPis exhibit varying abilities to trap PARP, with talazoparib being the strongest (19). Whether different drug properties will eventually lead to an optimal clinical choice remains unknown.

In line with current hypotheses, our data show that olaparib abolishes PARP catalytic activity and induces massive amounts of DNA double-strand breaks. Furthermore, we find that olaparib treatment induces a robust G₂–M arrest characterized by activation of p53 signaling. Although this ultimately leads to increased cell death as expected, we also found a significant proportion of cells evade death and undergo senescence associated with p21 activity. Persistent, PARPi-induced senescent cells could have profound clinical implications. Senescent cells, although no longer able to divide, are known to be highly active and to elicit profound effects on their microenvironments, largely through their acquisition of the senescence-associated secretory phenotype (29, 30). These properties are thought to be possibly linked to the promotion of tumor progression and the development of treatment resistance (31). Furthermore, the potential for re-emergence from senescence could allow persistent cells to form a repository of surviving cells which may then go on to proliferate and give rise to recurrent, resistant disease (27). A recent study found that PARPi treatment also induces both cell death and senescence in the context of ovarian and breast cancer (26). This study found that PARP inhibition could induce a p21-dependent senescence which was reversible possibly providing a pathway to recurrent disease. However, their work demonstrated that senescence was independent of p53 and rather relied upon Chk2 activation, highlighting the diverse mechanisms which can be activated in response to these drugs. Fully understanding response to PARP inhibition in each tumor context is imperative toward maximizing PARPis as a treatment strategy. Inputs leading to activation of p53 in our study remain unknown, but their elucidation could provide novel therapeutic targets for combination with PARP inhibition. It is thought that targeting senescence using senolytic drugs could prove to be an effective form of precision medicine by clearing treatment-induced senescent cells, thus enhancing treatment efficacy (32). Senescent cells rely upon activation of various pro-survival pathways. Furthermore, senescent cells are highly metabolically active. These specific pathways may provide vulnerabilities which can be exploited for the specific clearing of persistent, senescent cells, thus providing a means to greatly increase the utility of a treatment which induces senescence, like olaparib.

Interestingly, there is increased yH2AX in OlapR cells versus control cells, which suggests that although olaparib does not increase DNA damage in resistant cells, OlapR models deal with constant and persistent DNA damage acquired during progression to a resistant state. Despite this drastic increase in persistent damage, OlapR models continue to proliferate and do not exhibit G₂–M arrest. These findings imply that PARPi-resistant cells must employ mechanisms to manage persistent DNA damage and/or mitigate the DNA damage response while increasing survival pathways. Although these specific resistance mechanisms remain to be defined, we sought a means to sensitize resistant cells to treatment. A previous study demonstrated that targeting CDK1 could render BRCA-proficient cells sensitive to PARP inhibition through impairment of homologous recombination potential (28). In line with these findings, our data show that inhibition of CDK1 induces DNA damage and cell death in OlapR cell line models. Furthermore, CDK1 inhibition reduces cellular viability and sensitizes to olaparib treatment. These data provide proof of concept that CDK1 targeting could be developed to overcome acquired resistance to PARP inhibition.

In conclusion, our findings demonstrate that response to olaparib is characterized by increased DNA damage which elicits p53 activation and G₂–M arrest. This response leads to both cell death and senescence associated with p21 activity which may be targetable to further augment DNA damage and promote cell death over senescence. PARPi-resistant cells appear to override the G₂–M checkpoint in response to treatment but resistance may be overcome through further targeting DNA repair via inhibition of CDK1. Data presented here represent a platform for the study of PARP inhibition in prostate cancer which promises to lead to critical insights into the development and mitigation of resistance to this exciting new treatment strategy.

A.P. Lombard reports grants from NIH during the conduct of the study. C.P. Evans reports grants and personal fees from Astellas and Janssen and personal fees from Myovant Sciences and Pfizer outside the submitted work. M. Dall'Era reports grants from Janssen and Tempus outside the submitted work. No disclosures were reported by the other authors.

A.P. Lombard: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. C.M. Armstrong: Data curation, investigation, methodology. L.S. D'Abronzo: Data curation, investigation, methodology. S. Ning: Data curation, methodology. A.R. Leslie: Data curation, investigation. M. Sharifi: Investigation. W. Lou: Investigation, methodology. C.P. Evans: Investigation, writing–review and editing. M. Dall'Era: Writing–review and editing. H.-W. Chen: Resources. X. Chen: Resources. A.C. Gao: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

This work was supported in part by grants CA179970 (A.C. Gao), CA 225836 (A.C. Gao), and DOD PC180180 (A.C. Gao); the U.S. Department of Veterans Affairs, Office of Research & Development BL&D grant number I01BX004036 (A.C. Gao); and BLR&D Research Career Scientist Award IK6BX005222 (A.C. Gao). A.C. Gao is also a senior research career scientist at VA Northern California Health Care System, Mather, California. This study was also supported by NCI K01 1K01CA262351-01 (A.P. Lombard).

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.