VERSICAN PROTEOLYSIS PREDICTS IMMUNE EFFECTOR INFILTRATION AND POST-TRANSPLANT SURVIVAL IN MYELOMA

Autologous stem-cell transplantation (ASCT) for multiple myeloma continues to confer benefit in the “novel agent” era [1]. However, most patients relapse, risk stratification is inadequate and optimal post-ASCT maintenance therapy strategies remain unclear. Relapses are heralded by the emergence of IL-10-secreting regulatory myeloid cells (immunosuppressive macrophages and dendritic cells) and dysfunctional (exhausted/senescent) T cells expressing programmed death −1 (PD-1) and T-cell immunoglobulin and ITIM domain (TIGIT) receptors [2–5]. Moreover, IL-10-induced CD8+ regulatory T cells impede antigen-specific effector responses [6,7]. Infiltrating myeloid cells secrete versican (VCAN), a large matrix proteoglycan with immunosuppressive/ tolerogenic activities [8–10]. ADAMTS protease-mediated VCAN cleavage generates versikine, a bioactive proteolytic fragment (“matrikine”), that regulates BATF3-dendritic cell (BATF3-DC) differentiation [11]. BATF3-DC are a crucial DC subset that controls CD8+ abundance in the tumor microenvironment through enhanced CD8+ effector cross-priming (BATF3-DC are the most efficient DC subset in tumor antigen cross-presentation) as well as through chemokine-mediated CD8+ infiltration into tumor sites [12].

Immune reconstitution post-ASCT provides a favorable milieu for the institution of immunotherapies that aim to enhance effector function and control or eradicate minimal residual myeloma [4]. In this work, we sought to answer two questions: Firstly, does VCAN proteolysis correlate with CD8+ infiltration in the post-ASCT marrow microenvironment? Secondly, does VCAN proteolysis status carry prognostic significance in the post-ASCT setting?

Bone marrow core biopsies were collected with informed consent under a UW-Madison Institutional Review Board-approved protocol (HO07403). VCAN proteolysis status was determined through immunohistochemical staining for the neo-epitope DPEAAE that is exposed through ADAMTS-mediated proteolytic VCAN cleavage at the Glu⁴⁴¹- Ala⁴⁴² bond of the VCAN-V1 isoform [13]. DPEAAE constitutes the C-terminal end of the bioactive VCAN fragment, versikine [13]. Antigen retrieval was carried out in EDTA buffer (CD8 detection) or citrate (DPEAAE). Primary antibodies included anti-DPEAAE (PA1–1748A; Thermo Fisher, Waltham, MA) and anti-CD8 (c4–0085-80; eBioscience, San Diego, CA). Core biopsy anti-DPEAAE staining intensity was scored for each sample by at least three observers, including a pathologist (AMC), blinded to clinical parameters. Stained slides were examined using Olympus BX43 microscope with an attached Olympus DP73 digital camera (Olympus, Waltham, MA). Immunostaining for anti-DPEAAE was assessed by scoring staining intensity (1 for low/weak, 2 moderate and 3 for strong/intense staining). For CD8+ detection, the number of CD8+ T cells per hpf was calculated using a single area at 400X magnification (10X ocular with a 40X objective). Baseline and post-ASCT characteristics between the low and moderate/high VCAN proteolysis groups were compared using chi-square test, Fisher’s exact test or t-test as appropriate. Correlation of VCAN proteolysis status with CD8+ T cell count per hpf was performed using ANOVA and logistic regression. OS and PFS between the two groups were estimated using Kaplan-Meier method. Cox regression analyses were used to correlate factors with OS and PFS using VCAN proteolysis status as the main effect and hazard ratios (HR) and 95% confidence intervals (CI) were obtained. Logistic regression analyses were used to estimate predictors of 2-year OS and PFS and odds ratios (OR) with 95% CI were ascertained. Data were analyzed using SPSS version 21 (SPSS Inc, Chicago, IL), and p<0.05 was considered statistically significant.

Bone marrow biopsies from 35 myeloma patients collected at Day 90–100 post-ASCT were analyzed. Table 1 provides patient baseline characteristics. The median age was 61 years (range 24–74) and 63% were males. Twenty four percent had high-risk cytogenetics, and 18% had very good partial response or better (≥VGPR) at the immediate pre-transplant evaluation. All patients received high dose melphalan at 200 mg/m² for conditioning. Disease evaluation was done at day 90–100 which showed that 73% of patients were in ≥VGPR. Most patients received maintenance therapy (63%) with no significant differences between VCAN-proteolysis classification groups (p= 0.455, Table 1). Anti-DPEAAE staining was robust despite prior decalcification of samples, a condition known to adversely impact detection of other immune biomarkers, such as PD-L1, by immunohistochemistry. VCAN proteolysis-low intensity (1+) was present in 18 (51%) patients, while VCAN proteolysis-high intensity (2+ and 3+) was present in 17 (49%) patients (2+ intensity= 11 patients and 3+ intensity= 6 patients). Representative staining patterns are given in Fig. 1A. Core biopsies were stained for the effector T cell marker CD8 and CD8+ counts per high-power field (hpf) were correlated with the VCAN proteolysis status (Fig. 1A). There was a statistically significant correlation between VCAN-proteolysis intensity and CD8+ T cell infiltration (VCAN proteolysis intensity 1+: CD8+ count per hpf mean, SD= 17.7 (13.1); VCAN proteolysis intensity 2+: CD8+ count per hpf mean, SD= 27 (14.8); VCAN proteolysis intensity 3+: CD8+ count per hpf mean, SD= 67.3 (43.1); p<0.001) (Fig. 1B). On univariate logistic regression, intense VCAN proteolysis (2–3+) was associated with higher CD8+ T cell count (OR, 1.06, 95% CI 1.01–1.12; p=0.027).

Table 1:

Baseline and post-transplant characteristics of patients.

Characteristics	Total (n=35)	VCAN proteolysis		p value
		Low (1+) (n=18)	Moderate/High (2–3+) (n=17)
Age in years, median (range)	61 (24–74)	60 (24–74)	61 (46–73)	0.369
Male gender, n (%)	22 (63)	13 (72)	9 (53)	0.305
Caucasian race, n (%)	33 (94)	17 (94)	16 (94)	0.367
Isotype, n (%)
IgG lambda	4 (11)	3 (17)	1 (6)	0.285
IgA lambda	2 (6)	1 (6)	1 (6)
IgG kappa	16 (46)	9 (50)	7 (6)
IgA kappa	4 (11)	0	4 (23.5)
Lambda	3 (9)	1 (6)	2 (12)
Kappa	6 (17)	4 (22)	2 (12)
ISS stage, n (%)
I	9 (26)	6 (33)	3 (18)	0.177
II	11 (31)	7 (39)	4 (23.5)
III	15 (43)	5 (28)	10 (59)
Cytogenetics, n (%)
Standard risk	26 (76.5)	15 (88)	11 (65)	0.225
High risk	8 (23.5)	2 (12)	6 (35)
Induction regimen, n (%)
VRD	12 (34)	7 (39)	5 (29)	0.311
CyBorD	4 (11)	2 (11)	2 (11)
Rd	8 (23)	4 (22)	4 (23.5)
Vd	5 (14)	4 (22)	1 (6)
Others	6 (17)	1 (6)	5 (29)
Pre-transplant response, n (%)
sCR/CR	3 (9)	1 (6)	2 (12)	0.907
nCR/VGPR	3 (9)	2 (11)	1 (6)
PR	22 (63)	12 (67)	10 (59)
SD	2 (6)	1 (6)	1 (6)
PD	5 (14)	2 (11)	3 (18)
Post-transplant response, n (%)
sCR/CR	16 (46)	11 (61)	5 (29)	0.071
nCR/VGPR	10 (27)	6 (33)	4 (23.5)
PR	3 (9)	0	3 (18)
SD	2 (6)	0	2 (12)
PD	4 (11)	1 (6)	3 (18)
Maintenance therapy, n (%)
Lenalidomide	11 (31)	7 (39)	4 (23.5)	0.455
Others	11 (31)	6 (33)	5 (29)
None	13 (37)	5 (28)	8 (47)
Follow-up in months, median (range)	32 (7–116)	36 (8–116)	13 (7–80)	0.088

Abbreviations: VRD (Bortezomib, Lenalidomide, Dexamethasone), CyBorD (Cyclophosphamide, Bortezomib, Dexamethasone), Rd (Lenalidomide, Dexamethasone), Vd (Bortezomib, Dexamethasone), sCR (stringent complete remission), CR (complete remission), nCR (near complete remission), VGPR (very good partial response), PR (partial remission), SD (stable disease), PD (progressive disease).

Figure 1.

(A), Representative photomicrographs of post-ASCT bone marrow biopsies stained with an antibody detecting VCAN proteolysis (anti-DPEAAE neoepitope) (top) and an anti-CD8 antibody (bottom). hpf= high-power field. (B) Box-and-whisker plots demonstrating the distribution of CD8+ count per hpf according to VCAN proteolysis intensity. (C) Kaplan-Meier curves demonstrating overall survival (OS, p= 0.003) (left) and progression-free survival (PFS, p=0.054) (right) according to VCAN proteolysis intensity: low (1+) and moderate/high (2+ and 3+ intensity combined).

Outcomes were compared between patients with low VCAN proteolysis intensity (1+) and a combined group comprising of patients with moderate and high VCAN proteolysis (2+ and 3+ combined). Overall response rate (ORR) was 83% and 46% patients achieved complete remission. ORR and the depth of response, while superior in patients with low VCAN proteolysis intensity, were not statistically significant (p=0.071, Table 1). Low VCAN proteolysis compared to moderate/high VCAN proteolysis was associated with better OS (median not reached vs. 13 months, 95% CI 10–16, p=0.003) and a trend for better PFS (median 56 months; 95% CI 26–87 vs. 10 months, 95% CI 1–18; p=0.054) (Fig. 1C). In Cox regression analyses adjusted for age and gender, moderate/high VCAN proteolysis compared to low VCAN proteolysis predicted poor OS (HR 4.87, 95% CI 1.33–17.82; p=0.017) and a similar trend was seen for PFS although it was not statistically significant (HR 2.29, 95% CI 0.92–5.68; p=0.074). Higher CD8+ count as a continuous variable was associated with poor OS (HR 1.03, 95% CI 1.01–1.05; P=0.002). In multivariate Cox regression model including VCAN proteolysis status, CD8+ count and cytogenetic risk group, moderate/high VCAN proteolysis compared to low VCAN proteolysis remained independent predictor of poor OS (HR 5.07, 95% CI 1.04–24.76; p=0.045). Patients with low VCAN proteolysis compared to moderate/high VCAN proteolysis had better 2-year PFS (72% vs. 29%, p=0.018) and 2-year OS (83% vs. 35%, p=0.006). In logistic regression analyses adjusted for age and gender, low VCAN proteolysis compared to moderate/high VCAN proteolysis predicted higher 2-year PFS (OR 6.80, 95% CI 1.41–32.76; p=0.017) and 2-year OS (OR 7.64, 95% CI 1.50–39.07; p=0.015). In age- and gender-adjusted multivariate logistic regression model including VCAN proteolysis status, CD8+ count and cytogenetic risk group, low VCAN proteolysis compared to moderate/high VCAN proteolysis independently predicted better 2-year OS (OR 7.77, 95% CI 1.06–56.71; p=0.043) and 2-year PFS (OR 8.48, 95% CI 1.09–66.09; p=0.041).

We present the first set of data ascribing prognostic significance to the VCAN proteolysis immunoregulatory pathway in myeloma. We observed the somewhat paradoxical association between intense VCAN proteolysis and high CD8+ T cell infiltration with dismal post-ASCT survival. These observations support a model in which immunosuppressive myeloid cells (macrophages and regulatory DC) secrete tolerogenic VCAN. Subsequent proteolysis of accumulated VCAN releases versikine. Versikine in turn promotes CD8+ infiltration (likely through BATF3-DC chemokine networks [12]). Intramedullary CD8+ effectors become dysfunctional and/or frankly immunoregulatory at the tumor site [2–7], thus perpetuating immunosuppressive loops and sustaining conditions favorable for relapse. Multiparametric analyses suggested negative effects of VCAN proteolysis independent of CD8+ T cell status, alluding to yet unexplored interactions between intrinsically high-risk disease and tumor matrix remodeling.

The detection of VCAN accumulation and turnover thus indicates profound immunosuppression in the post-ASCT marrow microenvironment and pinpoints those patients at risk for relapse and early death. The dismal outcomes associated with VCAN biomarker positivity may be rationally overcome through cutting-edge immunotherapies such as anti-TIGIT checkpoint inhibition [2], dendritic cell vaccines [14] or myeloid reprogramming therapies [3,15]. At the time of writing, VCAN proteolysis detection has been incorporated as an ancillary study in a Phase II autologous DC/myeloma cell fusion vaccination protocol (NCT02728102).