Targeting USP7 identifies a metastasis-competent state within bone marrow-resident melanoma CTCs

Abstract

Systemic metastasis is the major cause of death from melanoma, the most lethal form of skin cancer. Although most melanoma patients exhibit a substantial gap between onset of primary and metastatic tumors, signaling mechanisms implicated in the period of metastatic latency remain unclear. We hypothesized that melanoma circulating tumor cells (CTCs) home to and reside in the bone marrow (BM) during the asymptomatic phase of disease progression. Using a strategy to deplete normal cell lineages (Lin-neg), we isolated CTC-enriched cell populations from blood of metastatic melanoma patients, verified by the presence of putative CTCs characterized by melanoma-specific biomarkers and upregulated gene transcripts involved in cell survival and pro-development functions. Implantation of Lin-neg population in NSG mice (CTC-derived xenografts, i.e. CDXs), and subsequent transcriptomic analysis of ex vivo bone marrow-resident tumor cells (BMRTC) vs. CTC identified protein ubiquitination as a significant regulatory pathway of BMRTC signaling. Selective inhibition of USP7, a key deubiquinating enzyme, arrested BMRTC in BM locales and decreased systemic micro-metastasis. This study provides first time evidence that the asymptomatic progression of metastatic melanoma can be recapitulated in vivo using patient-isolated CTC. Furthermore, these results suggest that USP7 inhibitors warrant further investigation as a strategy to prevent progression to overt clinical metastasis.

Introduction

Melanoma, the most aggressive skin cancer, frequently metastasizes to distant organs such as bone, lung, liver and brain (1). Unlike localized melanoma which can be surgically resected, metastatic melanoma with extra-nodal involvement is typically treated with systemic therapies including targeted and immune therapy (2). While outcomes are improving due to new therapies, the majority of patients with stage IV melanoma will die from their disease (1, 2).

Metastasis is a multi-step process that is initiated when tumor cells leave the primary tumor and traverse through the circulation to reach distinct organs (3, 4). While transiting through the peripheral blood, a subset of CTCs can home to the BM where they reside in a state of sustained quiescence (parallel progression model of metastatic melanoma) (5). The unique BM microenvironment exerts temporal and spatial selection pressures that are conducive for the survival of cell clones retaining long-term, self-renewal ability while acquiring the necessary traits for successful organ colonization, even in the absence of overt local bone invasion (3, 5, 6) . These BMRTCs are considered to be the “seeds” of future metastasis; targeting them for elimination or promoting the prolongation of their quiescent/growth-arrested state can therefore be a promising strategy to overcome or delay metastatic onset (5, 7). Accordingly, it is imperative to identify novel biomarkers for BMRTCs detection and to develop new therapeutic strategies that can eliminate BMRTCs before they progress to overt metastasis.

Successful surgical resection of primary melanoma (with clear margins) along with satisfactory lymph node dissection does not always prevent the incidence of late metastatic recurrence (5, 8). This suggests that metastatic dissemination is an early event wherein melanoma cells remain dormant but viable in distant organs while retaining abilities to generate overt metastasis at a later time (3, 8–11). For example, using the RET.AAD melanoma mouse models, Eyles et al. demonstrated that the process of tumor cell dissemination preceded the onset of symptomatic metastasis (4). Further, Ghossein et al. found that the presence of melanoma-specific transcripts in clinical blood and BM samples is associated with shorter median survival of patients (12). Despite these reports pointing to BM as the likely reservoir for disseminated CTCs, the precise role of BMRTCs in the pathogenesis of metastatic melanoma remains unclear and no significant progress has been made to target melanoma CTCs/BMRTCs for clinical metastasis prevention (4, 10–13).

We hypothesized that melanoma CTCs migrate to and reside in the BM during asymptomatic progression of the disease. We report here the isolation of CTC-enriched populations from the peripheral blood of metastatic melanoma patients, their expansion in immunocompromised mice, followed by their harvesting and subsequent immuno-phenotyping and molecular characterization. Using these approaches, we identified elevated USP7/PTEN expression as a distinct gene signature of BMRTCs. Further, we demonstrated that inhibition of USP7 reduces the metastatic potential of BMRTCs by prolonging their arrest in BM. Our investigations provide novel insights into the identity of BMRTCs that may be detectable in melanoma patients during asymptomatic metastasis, and present data to support the evaluation of USP7 inhibitors in melanoma patients at risk of developing metastasis.

Materials and Methods

Patient blood collection

Melanoma patients with stage III or stage IV disease were accrued according to protocols approved by the Institutional Ethical review boards at the University of Texas MD Anderson Cancer Center and Houston Methodist Research Institute (HMRI). All patient blood samples were collected after receiving informed written consent and according to the principles of Declaration of Helsinki. Clinical details and of each patient and PTEN H-score, included in the study are provided in supplementary table S1 and supplementary figure S1. Peripheral blood (18-20 ml) was obtained at the middle of vein puncture and was collected in CellSave^™ tubes (Menarini Silicon Biosystems, Inc.), or EDTA tubes under aseptic conditions. Samples were provided immediately to the laboratory for CTC isolation and analysis.

Antibodies and inhibitors used for the study

For multi-parametric flow cytometry and DEPArray, primary antibodies were obtained from the following sources: FITC-CD45 (#304054; 1:200), FITC-CD34 (#343504; 1:200), FITC-CD105 (#323204; 1:200), FITC-CD90 (#328108; 1:200), FITC-CD73 (#344016; 1:200), FITC HLA-A/B/C antibody (#311404; 1:200), PerCP/Cy5.5-CD146 (#342014; 1:100), PE-Human NG2/MCSP (#FAB2585P, 1:100), BV421-Ki67 (#350506; 1:100), were obtained from Biolegend, anti-Melan-A antibody (# AC12-0297-03; 1:200) from Abcore, and FITC-Anti-S100 (#ab76749; 1:50) was purchased from Abcam.

For immunohistochemistry, anti-human, anti-Melan-A antibody (# ab51061; 1:100), anti-tenascin-C (# ab108930, 1:100) and anti-p21 (#ab188224, 1:100) were purchased from Abcam, HLA-ABC (#565292; 1:100) from BD Biosciences, USP7 (#GTX125894; 1:200) from Genetex, and PTEN antibody (# sc-7974; 1:20) was purchased from Santa Cruz Biotechnology, anti-p53 antibody (#SAB4503021, 1:100) was purchased from Sigma, CCP110 (#12780-I-AP, 1:100) antibody was obtained from Proteintech. Anti-mouse, USP7 (#26948-1-AP, 1:200) and PTEN (#603000-1-Ig, 1:200) antibodies were purchased from Proteintech, Alexafluor-conjugated anti-mouse, anti-rabbit secondary IgG antibodies used for immunofluorescence staining (1:500 dilution) were obtained from Cell Signaling Technology. USP7 inhibitors P5091 (#SML0770) and P22077 (#2301) were purchased from Biovision.

PBMC isolation and CTC enrichment by multi-parametric flow cytometry

PBMCs were isolated by established procedures (14). Briefly, whole blood was treated with red blood cell lysis buffer (154 mM NH4Cl, 10 mM KHCO3, 0.1 mM EDTA) at 1:25 ratio, followed by incubation at room temperature (25°C) for 5 min, then pelleting the remaining blood cells at 300 g for 10 min. Mononucleated cell pellet was then washed twice with 1 × PBS (with 5mM EDTA) and used for fluorescence labeling followed by multi-parametric flow sorting (FACSAria^™ II, BD Biosciences). Data recorded during cell sorting were analyzed by DIVA acquisition software version 8.0.1 (BD Biosciences). Antibodies and reagents described above were used. Data generated by FACS were analyzed by FlowJo V10.

Experimental xenografts

All animal study was approved by our Institutional Animal Care and Use Committee (IACUC) protocol. Immunodeficient animal experiments were performed using 4-8 weeks old NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (Jackson Labs, Bar Harbor, ME). Flow-sorted Lin-neg PBMCs (~50,000 cells) derived from patient blood (8 ml volume) were injected in anesthetized NSG mice through intracardiac injection under aseptic conditions. Previous reports shows that injection of melanoma cell lines (1×10⁶ cells) with different metastatic abilities in immunodeficient mice typically develop overt macro-metastases in 10-12 weeks (poorly metastatic) and 3-4 weeks (highly metastatic melanoma cell lines) (15). Based upon this reference time-frame and considering that melanoma CTCs are rare cells with low transcriptional activity (Supplementary figure S2a-b and supplementary table S2 and S3), xenograft mice were euthanized 6 months after injection with patient-isolated Lin-neg cells (13, 16–19) . This was based upon the reasoning that this period should suffice for in vivo selection and elimination of non-tumor cells and successful establishment of a niche for resident single cells in systemic viscera and BM. Peripheral blood, BM and organs were harvested for downstream analyses. Approximately 800-900 μl blood was collected in EDTA tubes by cardiac puncture of anesthetized mice. Animals were sacrificed soon afterwards and BM/organ tissues were harvested. In particular, BM was obtained from femur and tibia by flushing them out with 1x PBS with EDTA (5mM) using a 28G×½ needle, followed by centrifugation at 300g for 10 min. Blood and BM were processed immediately for PBMC isolation and FACS analyses.

To study the effect of USP7 inhibitors on BMRTCs/CTCs, two to three immunodeficient mice were injected with Lin-neg population (50,000 cells/mouse) selected from the same individual melanoma patient (16 ml blood volume). One animal of each patient group was then treated with USP7 inhibitor (either P5091 or P22077). CTC-derived xenografts (CDXs) were treated twice a week subcutaneously with USP7 inhibitors at a maximum tolerated dose [P22077 (15 mg/kg); P5091 (10 mg/kg)] (20). Control vehicle consisted of 1x PBS (100 μl) injected in untreated CDXs. After 11 weeks of treatment, mice were sacrificed and necropsy performed.

Eight to ten weeks old, syngeneic mice C57BL/6J (n=30) were purchased from Jackson Labs, Bar Harbor, ME. Xenografts were developed by intracardiac injection of aggressive B16F10 melanoma cells (ATCC) (50,000 cells/mouse) and treated with or without USP7 inhibitors (P5091 and P22077) at their maximum tolerated dose as previously described. After 3 weeks, each group of animals (n=10) was euthanized and organs (BM, lung, brain, liver and lymph node) were harvested to perform H & E staining analyses. All melanoma cells were obtained from American Tissue Culture Collection (ATCC), DNA fingerprinted and routinely validated for mycoplasma-free testing at Characterized Cell Line Core Facility, MD Anderson, Houston, TX.

Immunofluorescence and immunohistochemistry

FACS-isolated CTCs were subjected to quick air-dry on Millennia™ 2000 adhesive glass slides (StatLab), and fixed with 4% paraformaldehyde (21). Cells were permeabilized (0.05% Triton X-100 in 1x PBS) for 30 min, followed by 30 min incubation in blocking buffer (1% BSA + 1% normal goat serum in 1x PBS). Next, immunofluorescent cell staining was employed using selected primary and secondary antibodies. Magnified (100X) images were captured using Zeiss Axio Observer microscope Z1 (Carl Zeiss, Jena, Germany), and data were analyzed using Zeiss ZEN2 software. Harvested tissue was processed and stained for H&E and other immunohistochemistry markers by the research pathology core at HMRI. Images were captured by using EVOS XL Cell Imaging System (ThermoFisher Scientific). Immunohistochemistry images were taken and quantified by free-access ImageJ software. Reciprocal intensity (250-y) was measured by subtracting the mean intensity of the stained area (y) from the maximum (250) unstained white area intensity as described (22). Student t-test, type2, paired 2 were used to calculate the difference in reciprocal intensity between two groups (USP7 inhibitors treated vs. untreated) for each protein.

DEPArray^™ and CellSearch^™ CTC interrogation

Flow-sorted CDX-derived HLA+/Melan-A+ were fixed with 2% paraformaldehyde for 20 min at 25⁰C followed by washing with 1x PBS to visualize cells at single-cell level. CTCs were then stained with appropriate antibodies, re-washed with the SB115 buffer, and loaded onto the DEPArray^™ platform (Menarini Silicon Biosystems, Inc.). Cells were then detected and analyzed by using Cell Browser software v3.0 (Menarini Silicon Biosystems, Inc.), as previously reported (14, 21). For CellSearch^™ CTC enumeration, harvested murine blood was spiked with blood from healthy donors and was processed using CellTracks® circulating melanoma cell kit and CellSearch^™ platform (Menarini Silicon Biosystems, Inc.), following manufacturer’s guidelines.

RNA isolation and gene expression profiling

Total RNA isolation was performed using NucleoSpin® RNA XS isolation kit (Macherey-Nagel, Inc.), then immediately provided to the sequencing/non-coding RNA core facility (MD Anderson Cancer Center, Houston, TX). RNA and cDNA amplifications, quality controls and gene expression arrays were performed using the human microarray platform (Clariom D, Affymetrix, Inc.). BMRTC/CTC samples were RNA-normalized using Affymetrix Powertool 1.18.0, and annotations were taken from Affymetrix version 36. Gene expression data analysis from each of BMRTC/CTC subsets (derived from asymptomatic CDX mice, with absence of histopathological confirmed macro-metastasis) was performed using Transcriptome Analysis Console 3.0.0.466 (Affymetrix, Inc.). Two-way ANOVA were performed to calculate fold change and P-value. Pathway enrichment analyses were subsequently performed using the Ingenuity Pathway Analysis software version 01-07 (Qiagen, Inc.).

WTA amplification and quantitative real-time PCR

mRNA and cDNA amplification were carried out from isolated RNA using REPLI-g single-cell WTA amplification kit, according to manufacturer’s instructions (Qiagen, Inc.). Amplified cDNA was purified employing the ExoSAP method (Affymetrix; # 78202.4X.1.ML), and subjected to quantitative real-time PCR using the SensiFAST™ SYBR® Hi-ROX Kit (Bioline, #BIO-92020). CT values were normalized with housekeeping gene beta-actin and log2 (-ΔΔC_T) were calculated. Student t-test, type 2, paired 2 was performed to obtain the P-value of each gene. Primer sequences for each gene are shown in supplementary table S4.

Whole genome amplification and next generation sequencing

Whole genome amplification (WGA) was performed by REPLI-g single-cell WGA kit, according to manufacturer’s instructions (Qiagen, Inc.). Briefly, the pool of CTC subsets was lysed and denatured followed by amplification to obtain intact amplified DNA of 10kb. Ion Torrent^™ next generation sequencing (NGS) was performed by using Ion AmpliSeq™ Cancer Hotspot Panel v2 (Thermofisher Scientific, #4475346). Libraries were constructed using amplified DNA and loaded on Ion Torrent^™ Pmt sequencer. Sequences were assembled with human reference hg19 assembly and variant analyses were performed by Advaita iVariantGuide (http://www.advaitabio.com/ipathwayguide). Analyses were based on NGS for the detection of somatic mutations in the coding sequence of a minimum of 46 genes (range 46-128) which were performed on DNAs extracted from samples in the CLIA-certified molecular diagnostic laboratory (University of Texas MD Anderson Cancer Center).

Mitochondrial DNA assessment and cell surface and intra-nuclear flow cytometry

The ratio of mitochondrial DNA (mtDNA) to nuclear DNA (nDNA) copies was used to verify cell proliferative status (MTN index) (23). The mitochondrial target was strategically selected within the non-coding displacement loop (MTDL) of the mtGenome because of the rare occurrence of large-scale deletions that are common to other areas, e.g., the major arc55. Single-copy and low variability beta-2-microglobulin (B2M) gene was selected as nuclear target. Relative quantitative PCR (qPCR) was performed and MTN index was calculated [MTN index=2X2ΔCT, where ΔCT=(CTB2M−CTMTDL)].

We used the Permiflow^™ method (US patent #US7326577 B2) of cell fixation for concomitant cell surface (e.g., HLA+/Mel-A+), and intra-nuclear staining of target proteins (e.g., Ki67) in CDX-derived cell populations, as previously described (14). Briefly, xenograft-isolated BM and blood PBMCs were first stained with the live/dead fixable Zombie NIR dye (Biolegend, #423105, 1:100) for 15 min, followed by two sequential washes with 1x PBS. All samples were then fixed and permeabilized with the Permiflow^™ solution by incubating at 42°C for 1 hour, followed by washing with 1x PBS. Cells were then re-suspended in staining buffer and stained with appropriate antibodies for flow cytometric analyses.

Results

Isolation and validation of melanoma CTC-enriched cell populations isolated from patients’ blood

CTCs derived from patients display significant heterogeneity in cell surface biomarkers, epitomized by the co-existence of tumor-initiating and stem cell populations within a single blood draw (16, 17). A consequence of this diversity is that multiple marker-based platforms (including CellSearch^™, the only FDA-cleared platform for clinical CTC testing), and cell size-based microfiltration devices are unable to capture the entire spectrum of CTC subtypes that may be biologically and/or clinically relevant in cancer progression (18, 24, 25). Because there is no universal melanoma CTC marker capable of identifying the heterogeneous CTC subgroups – particularly clones that may transit to- and -from the BM, we used immunocompromised mice for the in vivo selection of cells capable of BM homing and residence. To implement this strategy, we first employed multi-parametric flow cytometry to deplete normal cell lineages present in the peripheral blood of melanoma patients (patients enrolled in this study are listed in supplementary table S1), thereby enriching cell populations containing putative as well as uncharacterized abnormal cell populations (here referred as Lin-neg cells). Flow cytometric gating consisted of doublet discrimination and dead cell (DAPI+) elimination, followed by depletion of leukocytes (CD45+), circulating endothelial cells (CD34+), and mesenchymal stem cells (CD73+/CD90+/CD105+) (Figure 1a). Using this strategy, we successfully captured putative melanoma CTCs identified by the presence of established melanoma markers (Melan-A and S100) from the peripheral blood of eight melanoma patients validated through immunofluorescence (Figure 1b) (16). Second, we performed Ion Torrent^™ next-generation DNA sequencing and analyzed the genetic profiles of 409 oncogenes and tumor suppressor genes of CTCs (Lin-neg and Melan-A+ cells) derived from two independent melanoma patients (patients #9 and 10). We identified 176 and 231 clinico-pathological mutations in respective patients (#9 and 10), present on NF1, KRAS, TP53, RB1, ATM and EGFR genes (Figure 1c; supplementary tables S5a and b), confirming the neoplastic identity of these cells. Of note, CTCs from one of the analyzed patients (patient #10) harbored a melanoma-associated mutation in CDKN2A (p.Ala148Thr) (COSM3736958) ((https://www.ncbi.nlm.nih.gov/clinvar/variation). We also confirmed the presence of skin cancer-associated mutations in XPA (c.-4A>G), ERCC1 (p.Asn118Asn), and ERCC5 (p.Gly1534Arg) (https://www.ncbi.nlm.nih.gov/clinvar/variation). Third, to validate that melanoma patient-isolated Lin-neg cells (patients #2, 3, 4 and 7) were distinct from normal peripheral blood mononuclear cells (PBMCs), we performed whole genome microarray analyses on these cells (n=4), and compared them with the gene expression profiles of PBMCs derived from healthy donors (n=9). PBMC dataset (GSE100054) was specifically chosen because it was obtained using the same human microarray platform (Clariom D, Affymetrix^™, Inc.). A total of 15,056 genes were upregulated and 23,430 genes were downregulated (fold-change <−2 or >2, P-value=0.05) in Lin-neg cell populations compared to healthy donor PBMCs, suggesting low transcriptional activity (Figures 1d, 1e and supplementary table S2). Unsupervised hierarchical clustering showed distinct gene expression patterns in Lin-neg and PBMC cell populations. (Figure 1f). Lin-neg populations possessed significantly higher expression of BAGE, MAGEA1, B4GALNT1, DCD and S100A3 confirming the presence of specific melanoma-associated markers within this population (Supplementary table S3) (16, 19, 26). Downstream pathway-enrichment analyses in Lin-neg cell pools predicted increased activation of nuclear receptor and oncogenic signaling, along with inhibition of immune regulation and pro-growth signaling pathways (supplementary figure S3 and supplementary table S6 and S7). These findings were reflected in cellular functions, e.g., upregulated transcripts implicated in cell survival and pro-development functions with a concomitant decrease in pathways involved in cell proliferation and inflammation. Collectively, these results asserted the presence of putative and uncharacterized melanoma CTCs in Lin-neg cell populations isolated by multi-parametric flow cytometry.

Figure 1. Isolation, validation and molecular characterization of de novo Lin-neg cell population.

(a) Enrichment of viable Lin-neg cell population (CD45-/CD34-/CD90-/CD73-/CD105- cells) from blood of eight melanoma patients through multi-parametric flow sorting (patients #1-8); (b) Validation of FACS procedures by melanoma markers (S100 and Melan-A) expression in Lin-neg cell population through immunofluorescence staining. Representative images are shown. Scale bar=10 μM; (c) Genetic mutational profiling of Lin-neg/Melan-A+ cell populations derived from two representative melanoma patients (patients #9 and 10) by AmpliSeq Ion Torrent^™ Cancer Hotspot Panel V2. Number of missense, nonsense, and silent mutations of each patient is indicated; (d) Differential gene expression profiling of Lin-neg cell population from melanoma patients (n=4; patients #2, 3, 4 and 7), compared to dataset of PBMCs isolated from blood of normal healthy donors (GSE100054) (n=9). Two-way ANOVA were performed to calculate fold change and P-value; (e) Scatter plots showing global gene expression changes of Lin-neg cell population; (f) Hierarchical clustering showing distinct transcriptomic signatures of Lin-neg cell population vs. PBMCs.

Expansion of melanoma patient-isolated Lin-neg cells in vivo

To evaluate whether Lin-neg cells isolated from melanoma patients could recapitulate the metastatic cascade, we first injected Lin-neg cell populations (n=8) into the left ventricle of immunodeficient mice (patients #1-8). Mice were euthanized at the pre-designated study end-point (6 months), after collecting blood by cardiac puncture. BM from long bones (femur, tibia and humerus), and organs typically affected by metastatic melanoma (lung, brain and liver) were collected for downstream analyses. Interestingly, while routine histopathological inspection did not identify any macro-metastatic disease in these organs, extensive immunohistochemical analyses employing human (HLA-ABC) and melanoma (Melan-A) biomarkers detected discrete resident human melanoma cells in multiple murine organs (CTC-derived xenografts, i.e. CDXs) (Supplementary figures S4a and b) (27). Second, to assess the presence of human melanoma CTCs in blood of CDX models (derived from patient #4), we spiked CDX blood with healthy donor blood, and performed analyses using the CellSearch^™ platform to visualize and enumerate CTCs. We identified 8 DAPI+/CD45-/Melan-A+ cells from 500 μl of murine blood by CellSearch^™ (Figure 2a). Next, we harvested blood and BM for PBMCs isolation followed by multi-parametric flow sorting, and isolated HLA+/Melan-A+ ex vivo CTCs and BMRTCs (Figure 2b). We validated the veracity of flow procedures by immunofluorescence analyses, detecting the presence of human (HLA) and melanoma (Melan-A) markers in HLA+/Melan-A+ FACS-sorted populations (Figure 2c). Third, we confirmed the heterogeneity of a pool of melanoma markers (Melan-A/CD146/NG2/S100) in CDX-derived BMRTC/CTC populations (patient #3) by the DEPArray^™ (Figure 2d) (16, 17). The DEPArray^™ is an antigen-agnostic CTC capturing technology that separates viable rare cells at single-cell level. An additional important aspect of this system is the absence of shear force stress on cells during recovery (14). We identified, analyzed and quantitated heterogeneous melanoma cell populations displaying specific melanoma markers in FACS-sorted HLA+/Melan-A+ cell population derived from BMRTC/CTC of the same CDX mice: Melan-A+/S100+/CD146+/NG2+ (BMRTCs=27; CTCs=8), Melan-A+/S100-/CD146+/NG2- (BMRTCs=11; CTCs=10), Melan-A+/S100+/CD146+/NG2- (BMRTCs=11; CTCs=14) and Melan-A+/S100-/CD146-/NG2- (BMRTCs=0; CTCs=14) (Figure 2d). Fourth, we performed genetic profiling for a family of 50 key cancer genes by Ion Torrent^™ next-generation sequencing on ex vivo BMRTCs and CTCs (patients #3 and 4). We discovered novel mutations in PIK3CA, NRAS, KRAS, ERBB4, PTEN and APC genes along with two hotspot cosmic mutations (COSM21451 & COSM25085) in the PIK3CA gene using hg19 human genome assembly as reference (Figure 2b, and supplementary table S8). Lastly, we performed differential gene expression analysis of ex vivo vs. de novo CTCs (Lin-neg/CTC-enriched population) from paired patient samples (patients #2, 3, 4 and 7) by human microarray profiling (Clariom D, Affymetrix^™, Inc.). We observed 3,777 significantly up-regulated and 3,879 significantly down-regulated transcripts (FC=2.5; P-value=0.05) when CTCs vs. Lin-neg/CTC-enriched population-derived microarray data were compared. We also observed 12,431 significantly overlapping genes between arrays of these population (P-value=0.05) (Supplementary figure S5). Collectively, these results demonstrate that CTC-enriched cell populations isolated from melanoma patients and expanded in CDX models could be successfully harvested from blood and BM of these mice.

Figure 2. Isolation and validation of melanoma patient CTCs/BMRTCs from CDXs.

(a) Capture, visualization and enumeration of human melanoma CTCs by CellSearch^™ (CD45-/CD34-/DAPI+/Melan-A+ cells) (patient #4); (b) Multi-parametric flow cytometry gating used to isolate human melanoma cells from BM and blood of CDXs models (DAPI-neg/HLA+/Melan-A+ cells); Table showing number of genomic mutations (50 cancer gene sequencing panel) analyzed by AmpliSeq Ion Torrent^™ in representative patients (see “Materials Methods” for details). (c) Immunofluorescence visualization of patient-derived CTCs and BMRTCs from CDXs; scale bar=10 μM; (d) Representative images of ex vivo CTCs and BMRTCs isolated by the DEPArray (patient #3).

Divergent transcriptomic signatures between ex vivo BMRTCs and CTCs

To determine functional differences between BMRTCs and CTCs, we performed transcriptome profiling of HLA+/Melan-A+ cell populations obtained from CDX (derived from patients #1-8) BM and blood. We used paired BMRTC/CTC populations derived from 8 CDX models (showing absence of macro-metastatic disease) generated by injecting Lin-neg cell population from individual metastatic melanoma patients (n=8), and performed human microarray profiling (Clariom D, Affymetrix^™, Inc.). Analyses identified 3,645 differentially expressed genes; 2,439 genes were upregulated while 1,206 genes were downregulated (FC=2.5, P-value=0.05) in BMRTCs vs. CTCs (Figures 3a and b). Divergent transcriptomic signatures of BMRTCs/CTCs in CDX models reflected the differential behavior of Lin-neg cells owing to their spatial niche (Figures 3a, 3b and supplementary figure S6). Subsequent pathway enrichment analyses showed that the top five upregulated canonical pathways were EIF2 signaling, actin cytoskeleton, systemic lupus erythematosus, hypoxia signaling and, notably, the protein ubiquitination pathway (Figure 3b and supplementary table S9). Molecular and cellular functional annotations also indicated that BMRTCs upregulated genes involved in gene expression, RNA post-translational modifications, protein synthesis, cell cycle and cell survival with a concomitant decrease in cell death and apoptosis. Second, because the distinct gene expression pattern of ex vivo BMRTCs may have been conferred by the osteogenic niche during their residence in BM and could be a reflection of their dependence on these unique signaling mechanisms, we focused on individual genes in the protein ubiquitination pathway that were most strikingly elevated in BMRTCs. We found that Phosphatase and Tensin Homolog (PTEN), a known tumor suppressor gene and Ubiquitin Specific Peptidase 7 (USP7), a deubiquitinating enzyme involved in post-translational modifications, were significantly upregulated (PTEN, FC=219, P-value=0.0028; USP7, FC=5, P-value=0.0072) in ex vivo BMRTCs (Figure 3b) (28, 29). To confirm these findings, we assessed USP7 and PTEN expression on FACS-sorted HLA+/Melan-A+ ex vivo BMRTCs/CTCs by immunofluorescence, and observed elevated expression of USP7 and PTEN in BMRTCs (Figure 3c). Next, we performed survival analyses of USP7/PTEN gene expression in patients (n=114) with cutaneous melanoma whose data were obtained from TCGA database by employing Oncolnc software tool (http://www.oncolnc.org/) (30). Kaplan-Meier survival analyses show that high expression of USP7 correlated with decreased patient survival, whereas PTEN expression was not significantly associated (Log-rank P-values of 0.0167 vs. 0.112; Hazard ratio 1.598 vs. 0.7937 with 95% confidence interval) (Figure 3d). Additionally, we didn’t observe any correlation between patient’s PTEN mutational status and its expression on CDX-derived BMRTCs population. Collectively, these findings revealed that USP7 is uniquely upregulated in ex vivo BMRTCs and correlates with shorter patient survival.

Figure 3. Transcriptional profiling of melanoma patients BMRTCs vs. CTCs in CDXs.

(a) Hierarchical clustering (FC=2.5, P-value=0.05); (b) Volcano plot showing differentially expressed genes in ex vivo BMRTCs vs. CTCs by human array (Clariom D, Affymetrix, Inc.), two-way ANOVA were performed to calculate fold change and P-value [patients #1-8; BMRTCs=2 and CTCs=1 (shown below)];Table displaying significantly altered top canonical pathways in ex vivo BMRTCs vs. CTCs population; (c) Increased expression of USP7-PTEN axis in BMRTs vs. CTCs by immunofluorescence analyses (scale bar=10 μm). Graph shows the number of USP7 and PTEN-positive cells in BMRTC/CTC populations. One sample t-test was performed to compare the P-value; (d) Kaplan-plot analyses of USP7 and PTEN performed using cutaneous melanoma patients TCGA database (n=114) (http://www.oncolnc.org/). Blue/red color lines indicate USP7 and PTEN low/high expression, respectively. Log rank P-value and hazard ratio (USP7=1.599 and PTEN=0.7937) were calculated by GraphPad Prism ver7 (95% confidence interval).

USP7 inhibition modulates metastatic competence and proliferative properties of CDX-isolated BMRTCs/CTCs

We investigated the effects of USP7 inhibitors on metastatic competency of BMRTCs as melanoma-derived, BM-associated CTC population. We performed drug susceptibility assays in CDXs using two clinically-relevant USP7 inhibitors: P5091 and P22077. We injected Lin-neg cells (50,000 cells/mouse) from 6 melanoma patients (patients #11-16) simultaneously into 13 NSG mice and treated mice in parallel with: P5091, P22077 or vehicle (PBS). Mice were euthanized after 11 weeks of treatment and visceral organs and CTCs from blood and BM were harvested for downstream analysis. After euthanization, vehicle and USP7 inhibitor-treated CDX mice (11 weeks of USP7 inhibitor treatment) injected with Lin-neg cell population from the same individual patient, we performed immunohistochemistry to assess USP7 and PTEN expression in BM and lung as most common metastatic melanoma sites (1). We found a significant reduction of USP7, PTEN and Melan-A+ cells in BM and lung tissues (micro-metastatic disease) in USP7 inhibitor-treated vs. PBS-treated CDXs, revealing that USP7 targeting leads to a significant reduction of metastatic competency in patient-derived, BM-associated CTCs (Figure 4a-e and supplementary figure S7a-e).

Figure 4. USP7 inhibitors affect the metastatic potency of BMRTCs population in CDXs.

(a-e) Histopathological Melan-A evaluation with co-localization of USP7 and PTEN staining in lung and BM tissues from USP7 inhibitors-treated vs. untreated CDXs (immunodeficient mice) (patients #11-15). Black arrows indicate positive staining for respective markers. HLA and HE staining of the same region are shown in extended figure 7 of “Supplementary information” (Supplementary extended figure S7a-e). Graph showing semi-quantitative estimation of USP7/PTEN staining by using ImageJ software. Student t-test, type 2, paired 2 were performed to calculate P-value. Scale bar=100 μm.

We also investigated the effects of the USP7 inhibitors on metastatic competency in a syngeneic mice model. We injected metastasis-competent B16F10 melanoma cells through intra-cardiac procedure in syngeneic C57BL/6J mice. We grouped mice into three categories (n=10 for each group): (i) PBS-treated control group and groups treated with USP7 inhibitors (ii) P5091 (iii) P22077, respectively. Mice were euthanized after 21 days, and BM, lung, liver, lymph node and brain tissues were harvested for analysis. Histopathological evaluation showed a significant reduction of melanoma metastasis in with both USP7 inhibitors vs. untreated control mice groups (Figure 5a and 5b). Third, we investigated the effect of USP7 inhibitors on nucleus-cytoplasmic localization of PTEN (supplementary figure S8) (28). We observed significant reduction of nuclear:cytoplasmic PTEN ratio on BMRTCs population derived from USP7 inhibitor treated vs. untreated CDXs (supplementary figure S8a and b). However, we observed restoration of nuclear PTEN in lung mets of mice treated with USP7 inhibitors vs. untreated (supplementary figure S8b and c). This suggests that distinct microenvironment properties influence the USP7 regulation of PTEN expression.

Figure 5. USP7 inhibitors affect the metastatic potency of BMRTCs population in syngeneic mice.

(a) Representative H&E staining of lung, BM, liver, brain and lymph node tissue section derived from syngeneic mice treated vs. untreated with USP7 inhibitors. (b) Five to eight images of each mouse were quantified for melanoma cells in distant tissue organs. Student-t test, type2, paired 2 was performed to calculate P-value. Scale bar=100 μm.

Fourth, we investigated the effects of USP7 inhibitors on the metastatic competency of ex vivo BMRTCs, considering that roles of USP7 in the pathogenesis of metastatic melanoma remain unclear. We assessed USP7 inhibitors effect on the proliferative capacity of ex vivo BMRTCs and CTCs (derived from patients #13 and 14) by performing flow cytometric analyses for Ki67– a marker of cellular proliferation (14). We detected greater percentage of Ki67-high cells in CTCs vs. BMRTCs irrespective of USP7 inhibitor treatment [P5091: 89.1% (CTCs) vs. 68.4% (BMRTCs); P22077: 91.5% (CTCs) vs. 42.4% (BMRTCs)] (Figure 6a). We observed increased percentage of Ki67high among ex vivo BMRTCs isolated from USP7 inhibitor-treated vs. untreated CDXs. Conversely, we did not observe any difference in Ki67 status among ex vivo CTCs isolated from USP7 inhibitors-treated vs. untreated CDXs. Further, because mitochondrial DNA (mtDNA) copy number is an indicator of mitochondrial biogenesis and proliferative potency (31), we compared mtDNA vs. nuclear DNA content ratios in ex vivo BMRTCs vs. CTCs of 6 CDX mice (derived from patients #12, 13 and 14). Quantitative PCR showed significantly higher MTN index in BMRTCs derived from USP7 inhibitor-treated vs. untreated CDX mice (Figure 6b). Additionally, CellSearch^™ analyses of ex vivo BMRTCs/CTCs from CDXs (derived from patient #12) treated with USP7 inhibitors showed higher number of melanoma CTCs (CD45-/CD34-/DAPI+/Melan-A+ cells) in BMRTCs vs. corresponding CTCs (Figure 6c). Cumulatively, these findings suggest that USP7 inhibition causes a significant reduction of metastatic competency of BMRTCs by restricting melanoma cell population within BM locales (Figure 4a-e, 5a-b, 6a-c and supplementary figure S7a-e).

Figure 6. Effects of USP7 inhibitors on the proliferative ability of patient BMRTCs/CTCs and BMRTC gene expression isolated from CDXs.

(a) Effects of USP7 inhibitors on Ki67 HLA+/Melan-A+ BMRTC/CTC populations. Mean fluorescence intensity (MFI) of Ki-67 via intracellular multi-parametric flow cytometry (patients #13 and 14). Details of procedures and reagents used are described in the “Materials and Methods” section; (b) Mitochondrial DNA (MTN) index of FACS-sorted HLA+/Melan-A+ population derived from CDXs (n=3) paired BM and blood (patients #12, 13 and 14). Total gDNA was extracted and purified, then subjected to MTN determination using qPCR; (c) Capture, visualization and enumeration of CTCs by CellSearch^™ analyses of melanoma patient BMRTCs/CTCs isolated from CDXs treated with or without USP7 inhibitors (patient #12). 500 μl of CDX blood was collected and analyzed according to CellSearch^™ specifications (CD45-/CD34-/DAPI+/Melan-A+ cells). Displayed are BMRTC/CTC numbers/500 μl blood per treatment condition; (d) Differential gene expression profiling of BMRTCs population isolated from CDXs treated with and without USP7 inhibitors; (e) Scatter plot showing differentially expressed genes in BMRTCs by human array (Clariom D, Affymetrix, Inc.). Two-way ANOVA were performed to calculate fold change and P-value (FC=2, P-value=0.05); (f) Hierarchical clustering (FC=2.5, P-value=0.05) showing differential expressed genes in BMTRCs isolated from CDXs treated vs untreated USP7 inhibitors P5091 (patients #11 and 12) and P22077 (patients #12 and 16). Upregulated/downregulated genes are highlighted in red/black color, respectively; (g) Validation of microarray data by q-PCR analyses. Relative mRNA expression and standard deviation were calculated for each gene. Student t-test, type 2, paired 2 were performed to calculate respective P-values.

Effects of USP7 inhibitors on ex vivo BMRTCs/CTCs

To delineate genes involved in the metastatic potency of BMRTCs and affected by USP7 inhibition, we performed comparative transcriptional profiling of ex vivo BMRTCs from paired vehicle vs. USP7 inhibitor-treated CDX mice (derived from patients #11, 12 and 16). Differential gene expression profiling revealed 321 significantly altered genes (138 upregulated, 183 downregulated) in BMRTCs treated with USP7 inhibitors (Figures 6d and e). CDX-derived BMRTCs clustered together in two groups when hierarchical clustering was performed (FC=2, P-value=0.05), suggesting that USP7 affected BMRTCs by a distinct gene signature (Figure 6f). We also evaluated USP7-altered, metastasis competent targets in distinct CDX-derived BMRTCs and found significantly 4 upregulated (POLE, CCP110, SMAD13 and MTND4LP12) and 9 downregulated (TNC, FBXO25, RGS22, PDE1A, GPCPD1, ABSB4, SERINC3, EIF3F, and UBXN4) genes associated with metastatic colonization. Further, miRNA analyses revealed upregulation of MIR885-5p, MIR9-1, MIR6870 and MIR3657 and downregulation of MIR4719, MIR 548A2, MIR548AE1, MIR548AJ1 and MIR548AX2 transcripts in USP7-modulated BMRTCs. Quantitative PCR revealed significant downregulation of USP7, PTEN, TNC and GPCPD1 transcripts in HLA+/Melan-A+ ex vivo BMRTCs derived from CDXs treated with USP7 inhibitors (P5091 and P22077) vs. vehicle (Figure 6g). RGS22 transcripts were significantly downregulated in CDX BMRTCs when treated with inhibitor P22077 only, but not with inhibitor P5091 vs. vehicle (Figure 6g). Further, expression of genes involved in cell invasion (TNC, SERINC3, ABSB4 and SMAD13), cell migration (TNC, ASB4, RGS22 and GPCPD1), cell proliferation (MIR885-p, and CCP110), and cell apoptosis (TNC, FBXO25, PDE1A, EIF3F) were decreased by USP7 inhibitor treatment, possibly highlighting regulatory roles of USP7 at various metastatic steps during melanoma progression (Figure 7) (32–39). Next, to determine the role of USP7 inhibition in BMRTC-driven lung colonization, we performed immunohistochemistry for two USP7-modulated markers (TNC and CCP110) in lung tissue of CDX mice treated vs. untreated with USP7 inhibitors (derived from patients #11, 15 and 16). Immunohistochemistry quantification shows significant upregulation of CCP110 and downregulation of TNC in CDX-derived lung tissue treated vs. untreated with USP7 inhibitors (P-value=0.05) (supplementary figure S9a-d). Of relevance, USP7 inhibition has been shown to play a role in MDM2 degradation which leads to reactivation of tumor suppressor genes p53 and p21 resulting tumor growth inhibition (40). Accordingly, we quantified p53 and p21 expression in lung and BM tissues (derived from patients #13 and 15) of CDXs treated vs. untreated with USP7 inhibitors (Supplementary figure S10a-d). Immunohistochemistry quantification shows significantly elevated expression of p53 and p21 on lung and BM tissue derived from CDX mice treated with vs. without USP7 inhibitors (P-value=0.05).

Figure 7.

Model representing the asymptomatic progression of melanoma in patient-derived xenografts and effects of USP7 inhibition, either as a potential direct action on cells already disseminated to distant organs or as a result of USP7 targeting BMRTCs.

Discussion

This study provides first time evidence that asymptomatic progression of metastatic melanoma can be recapitulated in vivo using patient-isolated CTCs, and demonstrates that melanoma CTCs at advanced disease stages contain heterogeneous cell pools bearing distinct characteristics associated with BM vs. blood locales. Additionally, we show that USP7 plays critical roles in BMRTCs metastatic potential by regulating genes related to cellular proliferation, apoptosis and cell migration. Complementing previous investigations that have characterized the genetic landscape of primary and metastatic melanoma lesions, we demonstrate that transcriptional subtyping of melanoma CTCs provides key insights into the molecular mechanisms that regulate metastatic potency by proposing a novel rationale-based approach for targeting BMRTCs (5, 7, 8, 10, 11, 13, 16, 17). In contrast to the majority of CTC studies that have largely elucidated the spatial attributes of peripheral blood-based CTCs in the context of metastasis, we provide insights into the primordial stages of CTC dissemination and survival in BM without progression to overt metastasis.

Employing a negative selection strategy to deplete normal cell populations of blood, we implanted patient Lin-neg cells in immunodeficient mice to generate CDX models that can faithfully recapitulate the early events of melanoma metastasis. Lin-neg cell populations not only had over-expression of melanoma markers, but also harbored mutations on multiple melanoma oncogenes, demonstrating their atypical nature (Figure 1a-f and supplementary table S5) (14, 41). Further, combining the use of an antigen-agnostic platform (DEPArray^™) with genomic sequencing, we show that melanoma CTCs home to and reside in BM of CDX mice when injected with patient-isolated Lin-neg cell population (Figures 2a-d and 4a-e). We consider this approach useful for establishing PDX models particularly when patient tumor tissue may not be available.

Whole-genome transcriptomic arrays and qPCR for individual target genes identified distinct transcriptional signatures of ex vivo melanoma BMRTCs, potentially reflecting effects of BM microenvironment on BMRTC gene expression. Additionally, the proportion of Ki67-low cells was ~50% higher in ex vivo BMRTCs vs. CTCs, suggesting that BMRTCs contain a distinct cell sub-population that may be in a state of subdued proliferation (Figure 6a). However, we did not observe any correlation between high Ki67 index present in USP7 inhibitors-treated BMRTCs population and metastatic burden in visceral organs or bone. Of note, Ki67 as a good indicator of proliferation and present in G1/M and S phases of cell cycle while absent in the G0 phase (42). Therefore, this is likely due to the effect of USP7 inhibitors which arrests the Ki67+ BMRTCs population at G1/S phase of cell-cycle and inhibits their metastatic potential (43, 44). This suggests that USP7 inhibitors while augmenting the proliferative potency of BMRTCs, they do not augment their dissemination from BM to generate lethal metastasis. Interestingly, we detected BMRTC-specific mutations in SMAD4 and APC genes, suggesting that besides affecting gene expression, BM microenvironment may impart temporal selective pressures on BMRTCs, resulting in their acquiring unique mutations (Figure 2b and supplementary table S8) (5).

The deubiquitinase enzyme USP7 is involved in the post-translational modification of PTEN altering its tumor suppressive activity. Loss of PTEN function in melanoma patients has been linked to significantly faster progression to overt brain metastasis along with poorer overall survival (45). In our CDX models, we found significantly higher USP7 expression that co-localized with PTEN in BMRTCs when compared to ex vivo CTCs (Figures 3a-d and 4a-e). It is likely that the high expression of USP7 transcripts detected in BMRTCs reflects the intrinsic behavior of heterogeneous CTC pool as well as properties of the BM microenvironment. This is because: (1) patient-derived CTCs contain a heterogeneous pool of melanoma cells and CTC subsets have the tendency to seed the BM and (2) USP7-modulates the osteogenic differentiation of mesenchymal stem cells in bone marrow (5, 46). Kaplan-Meier survival analyses indicated that high USP7 (but not PTEN) expression is significantly associated with decreased overall survival of melanoma patients (Figure 3d); while targeting USP7 with specific inhibitors reduced ex vivo CTC numbers at metastatic organ sites along with a concomitant decrease in USP7 and PTEN expression (Figures 4a-e, 5a and 5b). Our immuno-histopathological evaluation shows a significant up-regulation of p53 and p21 expression in lung and BM tissues derived from CDX mice treated with USP7 inhibitors vs. PBS-treated group (Supplementary figure S10a-d), however, we did not observe a significant difference in p53 transcript expression between ex vivo BMRTCs vs. CTCs (28, 40). However, gene expression microarray analysis of USP7 inhibitor treated vs untreated ex vivo BMRTCs revealed several candidate genes significantly affected by USP7 inhibition (Figure 6d-f). Among them, TNC and FBXO25 are genes implicated in the development of metastatic melanoma, breast and non-small cell lung cancers (32, 33, 47). Further, downregulation of genes involved in cell migration (RGS22, GPCPD1 and ASB4) and invasion (TNC, SERINC3, ABSB4 and SMAD13) was also detected, suggesting that USP7 inhibition may impede BMRTC re-shedding from BM, arresting them in BM locales (35–37, 48). USP7 inhibition also decreased the expression of proliferation-related biomarkers (Ki67 and MTN indices) (Figure 6a-f), possibly by affecting genes (e.g., CCP110, POLE, p21 and p53) (Supplementary figure S9a-d and S10a-d) that arrest BMRTCs cell-cycle progression and may induce senescent phenotype (38, 39). Collectively, these findings suggest that USP7 plays a central role in mediating melanoma CTC residence in BM (Figure 6a-g and 7).

Our study has some limitations. First, analyses were performed on a low number of patients, therefore we cannot conclude that all melanoma patients follow these models and pathways. Second, we profiled a limited number of cells in each patient which although done stochastically, may have some inherent sampling bias. Third, although we suggest that USP7 inhibition-mediated reduction of micro-metastasis is not a direct effect on cells already disseminated to distant organs, we cannot exclude a direct effect of USP7 inhibition on these cells. However, our results demonstrate that melanoma patient-isolated CTCs home to and reside in BM, and emphasize the relevance of determining the prognostic value of BM-associated, CTC-derived cell populations. The systematic interrogation of these BMRTCs identified USP7 as an important mediator of BMRTC-specific signaling, and provided evidence that applying USP7 inhibitors in clinical settings can be relevant to eliminate residual melanoma cells in BM and thereby prevent further metastatic spread.

Significance.

Findings provides insights into mechanism of melanoma recurrence and propose a novel approach to inhibit systematic metastatic disease by targeting bone marrow-resident tumor cells through pharmacological inhibition of USP7.

Abbreviations

CTC

Circulating tumor cell

BM

Bone marrow

BMRTC

Bone marrow resident tumor cell

PBMC

Peripheral blood mononuclear cell

PTEN

Phosphatase and tensin homolog

USP7

Ubiquitin specific peptidase 7

FC

Fold change

CDX

CTC-derived xenografts

PDX

Patient-derived xenografts