An allogeneic multiple myeloma GM-CSF-secreting vaccine with lenalidomide induces long-term immunity and durable clinical responses in patients in near complete remission

Abstract

Purpose:

This proof-of-principle clinical trial evaluated whether an allogeneic multiple myeloma (MM) GM-CSF-secreting vaccine (MM-GVAX) in combination with lenalidomide could deepen the clinical response in patients with MM in sustained near complete remission (nCR).

Experimental Design:

Fifteen patients on lenalidomide were treated with MM-GVAX and pneumococcal conjugate vaccine (PCV) (Prevnar®) at 1, 2, 3 and 6 months.

Results:

Eight patients (53.3%) achieved a true CR. With a median follow-up of 5 years, the median progression-free survival had not been reached, and the median overall survival was 7.8 years from enrollment. MM-GVAX induced clonal T-cell expansion and measurable cytokine responses that persisted up to 7 years in all patients. At baseline, a higher minimal residual disease was predictive of early relapse. After vaccination, a lack of both CD27⁻DNAM1⁻CD8+ T cells and antigen-presenting cells was associated with disease progression.

Conclusions:

MM-GVAX, along with lenalidomide, effectively primed durable immunity and resulted long-term disease control, as suggested by the reappearance of a detectable, fluctuating M-spike without meeting the criteria for clinical relapse. For patients in a nCR, MM-GVAX administration was safe and resulted in prolonged clinical responses.

Introduction

Multiple myeloma (MM) is a neoplastic disorder of plasma cells characterized by clonal proliferation of malignant plasma cells in the bone marrow (BM) microenvironment and monoclonal protein in the blood and/or urine. The most common clinical manifestations include hypercalcemia, renal disease, anemia, lytic bone lesions and recurrent infections. MM arises from an asymptomatic malignant proliferation of monoclonal plasma cells (MGUS, monoclonal gammopathy of uncertain significance) that progresses to smoldering multiple myeloma (SMM) and finally to symptomatic MM (1). Immune dysfunction within the BM increases during this stepwise transition from MGUS to MM and underlies the loss of immune surveillance ultimately responsible for disease progression (2,3). The advent of novel therapeutic agents with fewer toxicities and greater tumor specificity have significantly improved clinical outcomes for patients with multiple myeloma (MM). Specifically, triple therapy with a proteasome inhibitor, an immunomodulatory derivative (IMiD) and steroids generate increased response rates and more durable remissions (4,5). This has translated into improved progression-free survival (PFS) and overall survival (OS) (6). However, a significant proportion of patients will eventually develop resistance to these agents and relapse. Approaches aimed at deepening and prolonging these responses have included long-term maintenance therapies with or without consolidation regimens (7,8). In particular, lenalidomide (Len) maintenance in both the transplant and non-transplant settings has shown the significant clinical benefit of such approaches (7,9). With the introduction of more effective treatments for MM, much of the overall improvement in clinical outcomes is likely due to the ability to achieve deeper responses. In fact, minimal residual disease (MRD) assessment by either flow cytometry or next generation sequencing can now detect as few as one in a million cells (10). A combination of more effective treatments coupled to deeper measurements of disease burden have enabled a better understanding of both the goals of therapy and what the relative significance of achieving such goals could have on overall disease outcomes. A growing body of literature strongly correlates the depth of response with improved clinical outcome to the point that MRD negativity is now being considered as a potential approvable endpoint for clinical trials (11). However, definitive clinical guidance and safe treatment options for the management of MRD-positive MM patients is currently lacking.

Immunotherapy exploits the capacity of the immune system to specifically recognize and eliminate cancer cells. In fact, immune checkpoint blockade (12) and genetically engineered T cells bearing chimeric antigen receptors (CAR-T) (13) have demonstrated clinical efficacy in hematologic malignancies and, on a more limited basis, in solid tumors. Conversely, cancer vaccines to date have not shown the same benefits (14). Although several factors, such as poor tumor antigen selection, choice of vaccine adjuvants and the absence of concomitant immunomodulatory therapy may account for the lack of clinical efficacy, the disease burden is likely to have a relevant impact on clinical outcomes (15). We have previously demonstrated the ability of Len to augment vaccine-specific cellular and humoral immunity (16). In the setting of MRD-positive MM, which implies a very low disease burden, we hypothesized that MM vaccination in combination with Len could represent a novel therapeutic approach to enhance treatment efficacy without additional toxicity.

In this context, the generation of productive vaccine-specific immune responses depends on the diversity and abundance of tumor-associated antigens, an effective adjuvant and concomitant immunostimulatory therapy (14). In cancer vaccine design, irradiated heterologous cell lines have been used as sources of numerous, diverse, and commonly shared, tumor-associated antigens to overcome the limitations of autologous tumor cell procurement (14). Moreover, established cell lines, such as K562, have been genetically modified to produce granulocyte-macrophage colony-stimulating factor (GM-CSF), a key immunostimulatory factor shown to improve efficient antigen presentation (17,18). Here we report the results of a proof-of-principle trial utilizing an allogeneic whole-cell GM-CSF-secreting MM vaccine administered in combination with Len in MM patients with a minimal residual disease burden defined as no detectable monoclonal spike but positive immunofixation electrophoresis (IFE). The primary endpoints were eradication of residual disease and conversion to complete remission (CR). Secondary endpoints included safety, time to response and immune monitoring of vaccine- and MM-specific T cell responses. To our knowledge, this is the first study attempting to treat patients with a minimal disease burden in an effort to further improve the disease response as well as to prevent disease progression.

Methods

Eligibility criteria

Eligible patients were at least 18 years old with a diagnosis of multiple myeloma and an Eastern Cooperative Oncology Group (ECOG) performance status of 0–2 with adequate hematopoietic, hepatic and kidney function. Patients were eligible regardless of the number of prior lines of therapy. An autologous hematopoietic stem cell transplant could not have occurred within the past 12 months and prior allogeneic bone marrow transplant was not permitted. To be enrolled, patients had to maintain a sustained near complete remission (nCR, defined as absent measurable M-spike and positive serum/urine immunofixation) for an observation period of at least 4 months on a Len-containing regimen. If patients remained in a nCR, all MM therapy with the exception of Len was discontinued and patients were continued to be observed for 3 weeks prior to enrollment on single agent Len. Key exclusion criteria were disease progression after steroid discontinuation, defined as a detectable M-spike > 0.5 g/dL or conversion to a true complete remission (defined as absent M-spike and negative serum/urine immunofixation) during the observation period.

Study design and treatment

This single-center single-arm trial was designed to evaluate the safety and preliminary clinical efficacy of MM-GVAX in combination with Len. Upon enrollment, all myeloma therapy was discontinued with the exception of Len which was continued at the prior dose. Patients received four vaccinations on day 14 of the 1st, 2nd, 3rd and 6th month from enrollment (cycles 1, 2, 3, and 6, respectively), while continuing Len for at least a year at the dose administered prior to enrollment. Len was administered for 21 out of 28 days. On a single vaccination encounter, patients received intra-dermal MM-GVAX administered over three limbs in a total volume not greater than 1 mL and the pneumococcal conjugate vaccine 13 (PCV-13, Prevnar®) injected intramuscularly in one arm. Upon completion of the study, patients continued treatment with Len assuming evidence of clinical efficacy (Protocol in Supp. Info). All subjects underwent safety assessments, disease response determinations and sampling of BM and PB for correlative biomarker analysis just prior to treatment (baseline), at day 14 of cycle 3 (C3D14, prior to the third vaccine dose) and 1 year after enrollment. For some patients, additional follow-up BM and PB samples were collected past the 1-year timepoint and up to seven years after trial enrollment. Clinical data analysis was censored as of February 18^th, 2020. All investigators obtained written informed consent from the patients prior to participation in the study. This study was approved by institutional review boards prior to enrollment (NCT01349569) and conducted in accordance to the US Common Rule.

Vaccine formulation

The cell lines used for the vaccine were manufactured by the GMP-compliant Cell Processing and Gene Therapy Facility at the Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins.

U266 and H929 were originally acquired from ATCC. K562/GM-CSF was made as previously described (17). Equal numbers (50 million each) of the MM cell lines U266 and H929 were combined with 5 million cells of the bystander cell line K562/GM-CSF. Vaccine cells were irradiated prior to cryopreservation and stored in liquid nitrogen until the day of use. On the day of vaccination, the individual cells were thawed, mixed at the appropriate concentrations and drawn up into three syringes. The final vaccine syringes were kept on ice until administration that occurred within 60 minutes after thawing.

TCR/IGH sequencing and analysis

Immunosequencing of the CDR3 regions of human TCRβ and IGH and IGK/L chains from T cells and B cells, respectively, was performed on genomic DNA extracted from BM or PB samples using the immunoSEQ® Assay (Adaptive Biotechnologies, Seattle, WA). Extracted genomic DNA was amplified in a bias-controlled multiplex PCR, followed by high-throughput sequencing. Sequences were collapsed and filtered in order to identify and quantitate the absolute abundance of each unique CDR3 region for further analysis as previously described (55–59). Only productive rearrangements were used for repertoire analysis.

Minimal Residual Disease Testing

Dominant IGH and IGκ/λ cancer clones were identified from immunosequencing results in pre-treatment bone marrow using the following criteria: 1) The sequence must have frequency > 5% 2) The sequence must be present at > 0.1% of the total nucleated cells 3) The sequence must be discontinuously distributed (four or fewer sequences in the next decade of sequence frequencies) 4) The sample must have a template estimate of > 200. These identified dominant clones were tracked over time in bone marrow to determine the frequency of the cancer clone(s) at subsequent time points after treatment. To account for somatic hypermutation (SHM), IGH clones that had 2 or fewer mismatches with the dominant clone were also tracked in bone marrow over time. The MRD frequency in each sample was measured as the frequency of the cancer clones among all productive rearrangements of the locus being tested.

Statistical Analyses of TCR-β sequencing results

Clonality was defined as 1- Peilou’s evenness and was calculated on productive rearrangements by: $1+\frac{\sum _i^N{p}_i{\text{log}}_2\left(p_i\right)}{{\text{log}}_2\left(N\right)}$ where p_i is the proportional abundance of rearrangement i and N is the total number of rearrangements. Clonality values range from 0 to 1 and describe the shape of the frequency distribution: clonality values approaching 0 indicate a very even distribution of frequencies, whereas values approaching 1 indicate an increasingly asymmetric distribution in which a few clones are present at high frequencies.

Morisita Index was used to determine similarity between two TCR repertoires and was calculated on productive rearrangements by: $\frac{2*{\sum }_i^S{a}_i{b}_i}{{\sum }_i^S{a}_i^2+{\sum }_i^S{b}_i^2}$ where a_i is the frequency of clone i in sample a and b_i is the frequency of clone i in sample b. Morisita Index values range from 0 to 1, where values of 1 represent identical repertoires and values closer to 0 represent more disparate repertoires.

Statistical analysis was performed in R version 3.2.

Antigen-specificity assays and intracellular cytokine analysis of T cells

BM-derived mononuclear cells obtained at the indicated timepoints before and after vaccination were stimulated either in AIM-V medium with 2% human AB serum alone, with SW780 (bladder carcinoma cell line) lysate or with U266/H929 (MM-GVAX cell lines) lysates, respectively. After 5 days, cells were harvested and stained for flow cytometric analysis of intracellular cytokine production. Surface and intracellular staining cocktails are listed in Table S1. Staining for flow cytometry and data analysis are described in further detail in the following Methods section.

Polychromatic flow cytometry and high-dimensional data analysis

Bone marrow and peripheral blood samples were collected at the pre-established timepoints, enriched for mononuclear cells using Lymphoprep (STEMCELL Technologies®) gradient and cryopreserved in freezing media (50% complete AIM-V media, 40% human decomplemented AB serum and 10% DMSO). Samples were then thawed and washed twice with prewarmed (37C) AIM-V with 0.02 mg/mL DNase and phosphate buffered saline (PBS), respectively. Flow cytometry reagents were purchased from BioLegend, BD Biosciences and Invitrogen. A full list of the monoclonal antibodies (mAbs) used is provided in Table S2. mAbs were previously titrated to the optimal concentration. Surface staining was performed for 20 minutes at 37C, while intracellular detection of cytokines was performed following fixation of cells with CytoFix/CytoPerm kit (BD Biosciences) according to the manufacturer’s instructions and by incubating the cells with specific mAb cocktails for 20 minutes at room temperature. All data were acquired on a Gallios® Flow Cytometer (Beckmann-Coulter) equipped with three lasers (violet, 405nm; blue, 488 nm; red, 633nm) and capable of detecting 10 parameters. Flow cytometry data were compensated in FlowJo by using single stained cell controls and compensation beads (BioLegend). After pre-processing by biexponential transformation, standard gating to remove aggregates and dead cells, CD3⁺ CD8⁺ T cells were subsequently exported from FlowJo for further analysis in R (version 4.0.1) by a custom-made script that used Bioconductor libraries and R packages. Briefly, data were analyzed using the FlowSOM algorithm for unsupervised clustering and visualized with UMAP. Differential discovery analyses were performed on R using the diffcyt framework and the CATALYST workflow (60). Data were then reorganized as new files, one per each cluster and further analyzed in FlowJo to determine the frequency of positive cells for each marker and their mean fluorescent intensity (MFI).

Statistics

Statistical analysis was performed on GraphPad Prism v8, SAS v9.4 and R v4.0 (and v3.2) with a p-value of less than 0.05 as a threshold for statistical significance.

Code and data availability

Materials generated by this study are available upon request, custom-made R scripts are available on CodeOcean (https://codeocean.com/capsule/6580382/tree/v2). Figures 1A and 1B were created with BioRender.com.

Figure 1.

MM-GVAX induces clinical responses in patients with low disease burden multiple myeloma. (A) Scheme of clinical trial. Patients received four doses of vaccine at the indicated timepoints (arrows) while on Len maintenance. (B) Flow diagram of patients enrolled on the clinical study. (C) Kaplan-Meier plot of OS probability of all enrolled patients (n = 15). (D) Kaplan-Meier of PFS of all enrolled patients (E) MRD levels as determined by NGS at the specified timepoints. (E) Change over time of the serum M-spike (upper panel) or light chains, either kappa or lambda (LC, lower panel) when appropriate. * indicates biochemical relapse, ** indicates relapse due to appearance of a new osteolytic lesion, *** indicates relapse due to appearance of a previously absent M-spike in a patient with light chain only disease. (F) OS, overall survival. PFS, progression free survival. MRD, minimal residual disease. NGS, next generation sequencing.

Results

Patients characteristics

Fifteen patients with multiple myeloma (MM) on a Len-containing regimen that achieved a stable near complete remission (nCR) for at least 4 months (observation period) were enrolled in the trial. A nCR was defined as a non-detectable M-spike and a positive serum and/or urine immunofixation (IFE) (19). Upon trial enrollment, all anti-MM therapy was discontinued with the exception of Len that was continued until relapse. The enrolled patients received four MM-GVAX vaccinations at 1, 2, 3, and 6 months in combination with Len at pre-enrollment dose. (Fig. 1A) Patients who achieved a nCR but subsequently either showed the reappearance of an M-spike or achieved a true CR during the observation period were not enrolled in the study (Fig. 1B). Baseline characteristics of the patients are summarized in Table 1. The median dose of Len was 15 mg (range 2.5–25mg). The median age at enrollment was 69 years (range 45–81 years). At diagnosis, 53% of the enrolled patients had stage I (International Staging System, ISS) disease (20). Notably, none of the enrolled patient had high-risk MM features as defined by the IMWG (International Myeloma Working Group) cytogenetic criteria (21) and five patients (33%) had undergone high-dose chemotherapy with autologous stem cell transplantation at a median time of 2.93 years (range 1.44 – 5.26 years) prior to vaccination. The median number of prior lines of therapy was 1.7 (range 1–4). Overall, the median time from diagnosis to trial enrollment was 2.1 years (range 1–8 years), while the median time from achievement of nCR to vaccine administration was 7.1 months (range 4–9 months). Only 53% (n = 8) of patients were on a regimen of single-agent Len, the remainder were on a two- or three-drug regimen. In order to compare MM-GVAX efficacy with a standard of care approach, we defined an observation group that includes all the patients that were screened for enrollment but experienced an improvement in their disease status (n = 6) or reappearance of a detectable M-spike, not consistent with disease relapse (n = 8, total n = 14, Fig. 1B). The median age at the end of the screening period was 65 years (range 40–83 years, p = 0.89 when compared to treatment group). All patients in both groups were previously exposed to immunomodulatory drugs (thalidomide and lenalidomide). Additional details regarding previous therapies, including ASCT, are presented in Table 1.

Table 1.

Baseline Characteristics.

	GVAX group	Obs group
n	15	14
Age at diagnosis (median)	66 (45–81)	65 (40–83)
MM type
IgG λ	2	3
IgG k	6	8
IgA λ	1	1
IgA k	3	2
K light chain	1	0
λ light chain	1	0
IgD	1	0
ISS at diagnosis
I	8	6
II	4	6
III	2	2
n/a	1	0
Response at enrollment
CR (IFE−)	0	6
nCR (IFE+)	15	0
Less than nCR (< VGPR)	0	8
FISH
high risk	0	0
low risk	15	14
prior Treatments, mean (range)	1.7 (1–4)	1.8 (1–4)
PI	12 (80%)	10 (71%)
IMiD (Thalidomide, Lenalidomide)	15 (100%)	14 (100%)
Alkylating agent (Melphalan, Cyclophosphamide)	6 (40%)	5 (36%)
Others (including trials)*	1 (6%)	1 (7%)
prior ASCT	5 (33%)	3 (22%)
median time from Dx to enrollment (years)	2.1	3.1
median time of observation (months)	7.1	6.5
median Len dose at enrollment	15	n/a

^*One patient enrolled in the MM-GVAX trial received experimental therapy with anti-PD1 antibody and lenalidomide with no clinical benefit more than 2 years prior to enrollment, while one patient in the observation group had a history of allogeneic bone marrow transplant more than 10 years before enrollment.

Safety

No treatment-related dose-limiting toxicities were observed and no grade 3 drug-related adverse events (AEs) were reported. The most common vaccine-related AEs were grade 1 injection site reactions of burning and pruritus reported by 80% of the patients (n = 12/15, Table S1).

Efficacy

All 15 patients were available for clinical response evaluation at 1 year after administration of the first MM-GVAX dose. Eight of 15 patients (53.3%) deepened their disease response consistent with true complete remission (CR), defined as an absent M-spike and negative IFE, with a median time to CR of 11.6 months (range 1.4–13.9 months) from enrollment. This rate of conversion to CR of 53.3% (95% CI: 26.6 – 78.7%) was significantly higher than a null hypothesis of 25% conversion to CR rate defined by design (p = 0.011) thereby indicating clinical activity and meeting the primary endpoint of this trial. Strikingly, only 6 of the 15 enrolled patients experienced disease progression, defined as the appearance of a M-spike of at least 0.5 g/dL and/or an increase in the involved free light chain of more than 100 mg/dL confirmed with a repeat measurement (22). At the time of the analysis, the estimated median overall survival (mOS) was 11.5 years from the MM diagnosis (95% CI: 5.9-n/a years) and 7.8 years from enrollment (95% CI: 4.2–7.8 years, n = 6/15, 40%, Fig. 1C). With a median follow-up of 6.9 years (range 4.4 – 15.2) from disease diagnosis and of 5.13 years (range 3.15–8.43 years) from enrollment in this vaccine trial, the median progression free survival (mPFS) could not be estimated because of the low number of events and their distribution over time (Fig. 1D). Specifically, only three patients (Pt 6, Pt 7 and Pt 9) experienced early disease relapse within the first year of trial enrollment. Interestingly, seven patients (46.7%) subsequently developed a detectable M-spike that did not meet the criteria for disease relapse which variably persisted over time but did not require any change in treatment (Fig. 1E). In contrast to previously published data, a subgroup analysis of patients achieving a CR showed no statistically significant difference in terms of PFS and OS compared to patients who maintained a stable nCR or those that subsequently developed a measurable M-spike (Fig. S1A)(23). This finding suggests that vaccination can likely result in long-term disease control even without the complete eradication of malignant MM plasma cell clones. Interestingly, the PFS of patients who were enrolled was significantly longer than the PFS of patients who continued with standard of care because they achieved a true CR or developed a detectable M-spike and therefore did not receive any vaccination (Fig. 2A).

Figure 2.

Clinical effect of MM-GVAX and in vivo persistence of vaccine-induced T cell clonotypes. (A) Kaplan-Meier plot of PFS probability of patients enrolled on this trial (GVAX) compared to screened patients who did not meet the eligibility criteria. (** indicates a p-value < 0.01). (B) Kaplan-Meier of the PFS probability of patients separated by MRD level (p = 0.0101) The dotted lines represent the 95% confidence intervals. (* indicates a p-value < 0.05). (C) Representative pairwise scatterplots of two patients showing clonal expansion of pre-existing T-cell clones after vaccination as well as the recruitment of novel clonotypes previously absent in either PB or BM. (D) The frequency of T-cell clones expanded at C3D14 was tracked over time in blood and bone marrow in all patients. Expanded T-cell clones were detected for up to 7 years in both tissue compartments.

The Minimal Residual Disease burden is predictive of sustained responses to the vaccine

To quantify and track the MM disease burden, minimal residual disease (MRD) testing by next generation sequencing was performed on 7 (46.7%) patients which were the only ones for whom baseline MM DNA was available (24). Since the tested samples in all but one subject had an MRD frequency greater than 10⁻⁶ (equivalent to 10⁻⁴ % when expressed as a percentage of productive rearrangements) in the B-cell repertoire, we identified an arbitrary threshold of 10⁻³ (10⁻¹ %) that allowed us to evaluate the clinical significance of a disease burden above (high-level MRD⁺) and below (low-level MRD⁺) this threshold in the bone marrow (Fig. S1B). Interestingly, all high-level MRD⁺ patients experienced disease relapse within a year from enrollment (median = 4.8 months, range: 2.8 – 9.5 months), while low-level MRD⁺ patients had a significantly longer PFS (median = 84.15 months, range: 51.9 – 97.3 months, p = 0.01, Fig. 2B). Detection of high-level MRD compared with low-level MRD was associated with an increased likelihood of clinical relapse (hazard ratio, HR = 25.79, 95% CI: 2.17 – 306.4). Notably, Pt 2 and Pt 4 had a prolonged clinical response despite developing a variably detectable M-spike that never met the criteria for disease progression. In contrast, Pt 12 and Pt 13 achieved a true CR and maintained their response over time. These findings suggest that the MRD burden at the time of vaccination may be predictive of long-term clinical outcomes.

MM-GVAX vaccination induces systemic myeloma immunity

To determine the impact of MM-GVAX vaccination on the global T cell receptor (TCR) repertoire, we performed deep sequencing analysis of the TCR chain Vβ (TRBV) on matched peripheral blood (PB) and bone marrow (BM) samples from all patients at baseline, prior to the third MM-GVAX dose (C3D14) and at 1 year. Productive clonality varied greatly among different patients and over time, but no vaccine-related pattern could be identified (Fig. S2). Considering that minimal TCR repertoire skewing was observed with vaccination, we examined the changes in clonal abundance pre- and post- vaccination by comparing the frequencies of each clone. Despite the absence of vaccine-associated changes in overall clonality, there is clear evidence of clonal expansion and contraction in both PB and BM samples. Interestingly, many of the significantly expanded clones after vaccination were not observed prior to vaccination suggesting either the expansion of clones previously present below the threshold of detection or the recruitment of novel clonotypes upon vaccination (Fig 2C). Notably, significantly expanded T-cell clones in both the PB and BM persisted relatively stably for up to 7 years post vaccination (Fig. 2D) Collectively, these results show that MM-GVAX administration induced a broad and durable T-cell response with measurable clonal expansion in both BM and PB for several years after vaccination.

Vaccination induces MM-specific polyfunctional T cell responses in the bone marrow

We also sought to functionally characterize T cell responses to both vaccine-related and unrelated MM antigens in the BM. Considering that this study aimed to examine the ability of vaccines to prime MM-specific responses in the context of Len, patients needed to be in a sustained nCR for at least 4 months on a Len-containing regimen and to then discontinue all other MM-drugs except Len for at least three weeks prior to receiving their first vaccine. The baseline BM biopsy was obtained at the end of these wash-out period, before the first vaccine dose was administered. As such, the baseline timepoint is de facto examining the single effect of Len on MM-specific immunity and thus indirectly serves as a biological negative control to determine the efficacy of MM-GVAX in eliciting such responses. BM mononuclear cell samples from all patients and timepoints were co-cultured in vitro with whole-cell protein lysates from either the MM-GVAX cell lines (U266 and H929) or unrelated MM cell lines (KMS-11 and KMS-12) and analyzed for intracellular cytokine production. Due to the technical difficulty of using autologous MM cells in patients with very low disease burden, we and others (25,26) have used whole-cell lysate from allogeneic cell lines to pulse the autologous antigen-presenting cells in the BM samples and therefore stimulate the T cells in an HLA-dependent manner, while avoiding alloreactivity. To specifically exclude non-vaccine-related immune responses, we also stimulated the BM samples with SW780 lysate, an unrelated bladder cancer cell line. While responses to SW780 were absent at all the tested timepoints (Supplementary Fig. S1C), MM-GVAX-specific interferon-γ (IFNγ) and TNFα responses were mostly undetectable before vaccination at baseline and markedly increased upon vaccination in all patients and in both CD8⁺ and CD4⁺ T cell subsets at C3D14 and 1 year (Fig. 3A). The frequency of CD4⁺ or CD8⁺ T cells producing either IFNγ and/or TNFα in response to vaccine-related and unrelated MM-antigens significantly increased with only two vaccinations and remained persistently elevated for up to 4 years or more (p < 0.0001, Fig. 3B). Notably, MM-GVAX significantly increased the frequency of polyfunctional CD4⁺ and CD8⁺ T cells, defined as co-producing IFNγ and TNFα, as well as the fraction of single-cytokine producing T cells, albeit to a lower extent. Interestingly, the most profound changes appeared to be the generation of CD8⁺T cells producing either TNFα or IFNγ/TNFα (Fig. 3C–D). Pt 6, Pt 7 and Pt 9, who relapsed early after vaccination, developed vaccine-specific T cell responses comparable to patients that achieved long-term disease remission. These results demonstrate that Len alone fails to generate spontaneous MM-specific immunity and that the MM-GVAX-induced immune response is polyfunctional, directed towards a broad range of commonly shared MM-associated antigens in both CD8⁺and CD4⁺ T cells. These findings support data from other studies showing that effective vaccine-induced anti-tumor immunity requires both CD8⁺ and CD4⁺ tumor-specific T cells (27,28).

Figure 3.

Vaccination induces durable, polyfunctional MM-specific T cell responses in the bone marrow. (A) Representative plots showing IFNγ and TNFα production before, during (C3D14) and after vaccination in both CD8⁺ and CD4⁺ T cell compartments. (B) Cytokine production increased after vaccination in all patients and was maintained for more than 4 years (p < 0.0001 for both CD8⁺ and CD4⁺ compartments). (C) UMAP plots divided by timepoint showing the different proportions of cytokine-producing T cell clusters. (D) Boxplots showing frequencies of each individual cluster across patients and timepoints. (E) Representative pairwise scatterplots showing persistence, in both BM and PB, at 1 year and 5 years post-vaccination of T-cell clones that were expanded at C3D14. (F) T cell clones expanded post-vaccination could be tracked in both PB and BM up to 7 years after MM-GVAX administration. (G) Representative plots showing IFNγ and TNFα production upon in vitro antigen-stimulation of BM from vaccinated patients at the indicated, long-term follow-up timepoints. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001.

Vaccine-induced MM-specific polyfunctional T cell responses persist for several years after vaccination

Important hallmarks of adaptive immunity are its persistence over time and its capacity to mount an effective response upon antigen re-encounter. As such, we sought to detect and characterize MM-GVAX-specific functional responses in available samples collected several years after vaccination. Surprisingly, intracellular cytokine staining of these BM samples stimulated with both MM-GVAX-related and unrelated MM-antigens showed a persistent polyfunctional CD4⁺ and CD8⁺ T cell response up to 7 years post-vaccination (Fig. 3E). In summary, vaccine-specific clones in PB and BM (Fig. 2D) as well as MM-specific polyfunctional T cells persist for several years after vaccination even in the presence of detectable disease and may be responsible for the long-term disease control observed in this study.

Immune correlates of clinical outcome after MM-GVAX vaccination

To gain insight into the immunophenotypes of BM memory CD8⁺ T cells post-vaccination and their association with clinical outcome, we analyzed C3D14 BM memory CD8⁺ T cells with FlowSOM, an unsupervised clustering algorithm, and used dimensionality reduction approaches, such as Uniform Manifold Approximation and Projection (UMAP), to simplify the visualization of different T cell clusters. Hierarchical metaclustering of FlowSOM clusters grouped CD8⁺ T cell subpopulations with similar immunophenotypes and identified two subsets of CD8⁺ T cells, i.e. C1 and C2, that were enriched in patients with durable disease control (responder group) compared to those showing early progressive disease (relapse group) within the first year after vaccination (Fig. 4A). Remarkably, clusters C1 and C2 were defined by low-to-absent DNAM1 expression and lack of CD27, and comprised relatively heterogeneous subpopulations including senescent, effector and exhausted CD8⁺ T cells (Fig. 4B). Further characterization of cluster C2 identified a CD69⁺ CD57⁻ subpopulation with intermediate PD1 expression, suggesting that these CD8⁺ T cells enriched in the responder group are BM-resident and likely mediate long-term MM control. This cluster lacks the hallmarks of terminal exhaustion, such as high PD1 expression and increased levels of TIM3 (29). Interestingly, cluster C1 exhibits increased CD57 expression, suggesting that senescent cells may still be capable of effector-like function, despite their lack of proliferative potential. Overall, the absence of this memory CD8+ T cell population is associated with early relapse upon vaccination (Fig. 4C, S3A).

Figure 4.

Low frequency of DNAM1⁻ CD27⁻ CD8⁺ T cells in the BM is associated with early relapse after vaccination. (A) Boxplots representing relative abundance of the 8 FlowSOM metaclusters in the two groups (relapse and responder). (B) Heatmap showing the MFI of specific markers in discrete FlowSOM clusters that are hierarchically metaclustered to group subpopulations with similar immunophenotypes. (C) UMAP plots of concatenated CD8⁺ T cells separated by clinical outcome showing that clusters C1 and C2 are markedly under-represented in the relapse group. *, p < 0.05, **, p < 0.01.

Increased frequencies of BM-resident memory CD8+ T cells and HLA-DR+ antigen-presenting cells are associated with disease control

We next sought to further characterize the bone marrow microenvironment and to deepen our phenotypic characterization of memory CD8+ T cells. Firstly, we reproduced the results obtained in our exploratory analysis with FlowSOM using a 25-color spectral flow cytometry panel (Fig. 5A). Interestingly, the BM microenvironment in patients with long-term controlled disease had significantly more HLA-DR+ antigen-presenting cells than patients whose MM escaped immune control and rapidly progressed (Fig. 5B). Patients with early progressive disease had lower frequencies of both HLA-DR+ cells and DNAM1− CD27− memory CD8+ T cells, suggesting that for MM to evade control it has to disrupt the BM microenvironment, especially its immune components. Notably, phenotypic characterization of the four subsets defined by DNAM1 and CD27 revealed peculiar features of the DNAM1/CD27 double negative population that we found to be associated with long-term MM control. The lack of CD27 has been associated with increased effector-like function, especially in MM (30), whereas DNAM1 maintains T cell functionality in the tumor microenvironment when professional antigen-presenting cells are lacking (31,32). Here, we describe a subset of memory CD8+ T cells with BM-resident features, such as increased CD69 and Eomes expression (33,34), and intermediate Tbet expression, suggesting retained effector potential (Fig. 5C). These findings support the hypothesis that immune control of MM in the BM relies on both tissue-resident memory CD8+ T cell subpopulations and antigen-presenting cells.

Figure 5.

Long-term clinical remission is supported by DNAM1−/CD27− CD8+ T cells and antigen presenting cells. (A) Representative dot plots showing manual gating analysis of DNAM1^−/low CD27⁻ CD8⁺ T cells (left) and summary of the frequency of this CD8⁺ T cell subset in both groups. (B) HLA-DR+ antigen-presenting cells are present at significantly increased frequency in long-term responders compared to patients who experienced early MM relapse. (C) DNAM1−/CD27− cytotoxic T cells display an effector-memory, tissue-resident phenotype as demonstrated by CD69 expression, lower TCF1 levels and intermediate Tbet/Eomes expression. *, p < 0.05; **, p < 0.01; ***, p < 0.001; ****, p < 0.0001.

Discussion

The introduction of novel agents such as immunomodulatory drugs (thalidomide, lenalidomide and pomalidomide) and proteasome inhibitors (bortezomib, carlfilzomib and ixazomib) has led to major improvements in clinical outcomes for patients with multiple myeloma. Highly effective drug combinations have dramatically increased CR rates (35) and next generation sequencing (NGS) techniques have enabled the quantification of the “minimal residual disease” (MRD) burden to deeper levels. MRD negativity is now recognized as a major prognostic factor in MM (8, 26) and is increasingly becoming a measurable endpoint for both clinical studies and standard treatment as it is increasingly achievable. However, guidance on how to optimally treat MM patients in CR, but still with MRD positivity is currently lacking. In this setting, vaccines represent an attractive approach to eradicating a small residual disease burden which could translate into clinically meaningful outcomes.

This single-arm proof-of-principle trial examined an allogeneic GM-CSF-producing MM vaccine (MM-GVAX) administered to patients that achieved a stable nCR, defined as an absent M-spike and a positive IFE in either serum or urine, for at least 4 months. The objectives in this study were to demonstrate the feasibility, toxicity, evidence of priming a myeloma-specific immunity as well as deepening of the clinical response. In the current trial, the rate of conversion from nCR to true CR was 53.3% with 8 patients improving their clinical response within a median time of 11.6 months from enrollment. Strikingly, the median PFS could not be estimated because only 6 patients, including the three early non-responders, experienced disease relapse over a median follow-up of 5 years. Despite the requirement for a nCR at the time of enrollment, three patients who relapsed within 1 year of enrollment showed a disease burden greater than 10⁻⁴ (10⁻¹%)by MRD testing at baseline suggesting that the amount of MRD prior to vaccination could be a major determinant of the clinical responses to anti-cancer vaccines (37). This would be consistent with the hypothesis that effective cancer vaccines work primarily in the setting of a low disease burden to effectively eradicate the minimal disease burden present after primary treatment (38). Several factors are required to generate active, tumor-specific T cells. These include the presence of appropriate antigens, effective antigen presentation, inhibition of the suppressive tumor microenvironment and expression of the appropriate chemokines. To facilitate efficient T cell trafficking to the tumor site, broad antigen recognition and efficient tumor priming, different vaccine strategies have been developed (15,39). Examples of polyvalent vaccines include autologous-dendritic cell (DC) fusion vaccines (40,41) and allogeneic GM-CSF based vaccines. The benefit of autologous DC-fusion vaccines is a broader antigenic repertoire that is patient specific. In contrast, allogeneic MM-GVAX relies only on shared tumor-associated antigens that are patient-independent but allow for development of an off-the-shelf product that can easily be utilized in the setting of MRD MM. Our vaccine design takes into account several of these key components.

First, the source of tumor-associated antigens (TAA) is composed of two established heterologous MM cell lines, H929 and U266, that display a diverse pattern of somatic mutations frequently associated with high risk and relapsed MM (42). Specifically, H929 harbors a t(4;14) translocation and a mutated NRAS, while U266 has several mutations involving BRAF and TP53 pathways (43). Considering the emerging evidence that disease relapse occurs as a result of clonal evolution leading to the acquisition of more aggressive genetic mutations (42), priming the immune system to several of these putative high-risk antigens prior to their appearance during clonal evolution could significantly impact the timing and/or aggressiveness of disease relapse.

Second, in addition to the unmodified MM cell lines, this MM-GVAX also include the genetically modified bystander GM-CSF-secreting cell line, K562/GM-CSF (17). GM-CSF is a key immune adjuvant used in several clinical cancer vaccine studies believed to be enhance antigen presentation (14, 15, 30). Importantly, the high amounts of GM-CSF produced by this bystander cell line allowed us to lower the cell dose and still deliver the optimal amount of GM-CSF that was neither insufficient nor supratherapeutic as to reduce its efficacy through the induction of myeloid derived suppressor cells while maximizing antigen delivery (45). Furthermore, irradiation inhibited proliferation of these tumor cell lines and induced immunogenic cell death to improve antigen delivery (14).

Third, immunomodulatory drugs (IMiDs), such as lenalidomide, markedly improve T cell responses in cancer patients and enhance vaccine efficacy which could serve as an important vaccine adjuvant in this setting (16,46). Overall, these unique features of MM-GVAX, in combination with continuous lenalidomide administration and a low tumor burden resulted in effective, long-lasting anti-MM immunity.

Taken together, we have developed an “off-the-shelf” vaccine strategy capable of priming long-lasting myeloma immunity and imparting long-term progression free survival. This is especially evidence when comparing the results of our vaccinated patients with those from our initial observation cohort that went on to achieve a true CR (IFE negative) and were thus ineligible. Their median PFS was 24 months compared to the vaccine group which has not been reached at approximately 96 months (Fig 2A). The significant difference in vaccine responses based on MRD disease burden confirms that the best vaccine efficacy occurs in conditions of a low disease burden (Fig 2B). Specfically, effective vaccination requires the ideal tumor antigens, an effective immune adjuvant and a low disease burden. As MRD becomes increasingly more achievable and correlates with improved clinical outcomes, it also underscores how long-term disease control in MM still remains elusive. Most of these will still relapse with their disease. This is most clearly demonstrated in the recent BCMA CAR-T cell data showing that patients achieving a CR ultimately relapsed with disease (47) with earlier data from this same trial showing only a delayed relapse, but not durable remission, in patients that achieved MRD negativity. Taken together, we have now reached a point in MM therapy in which we can efficiently reduce the disease burden but are not capable of eradicating it. Our vaccine approach generates long-lasting MM-specific immunity that appears to significantly delay disease progression.

In addition to the rapid induction of antigen-specific T cells and their proliferation upon antigen re-encounter, persistence over time in the setting of minimal or undetectable antigen levels is a cardinal feature of T cell memory. In this trial, we detected T cell clonotypes that expanded after vaccination with a polyfunctional cytokine T cell phenotype that persisted up to seven years post vaccination. Long-term persistence of polyfunctional vaccine-specific responses were seen both in patients who achieved a complete remission as well as in patients who unable to eradicate the disease but showed long-term disease control. Such data supports a model of dynamic MM-immune equilibrium resulting in effective cancer surveillance (48). This model of a myeloma-MGUS (Monoclonal Gammopathy of Uncertain Significance, the premalignant state of myeloma) equilibrium that is held in check by continued activity of T cell-mediated immunity is further supported by the identification of a BM-resident memory CD8⁺ T cell population that is associated with long-term disease control. Recent studies identified T_RM cells as important drivers of immune equilibrium in solid tumors (49) and also support their role in enhancing cancer vaccine efficacy (50). Moreover, loss of the stem-like/resident T cell population in the BM of patients with MGUS has been associated with progression to MM (2). We provide evidence of the existence of a BM (tissue)- resident, quiescent T cell population lacking the hallmarks of exhaustion and senescence and displaying high PD1 levels and an effector memory-like phenotype. These observations support the concept of the BM as a reservoir for antigen-experienced memory T cells and provide evidence for the putative mechanism whereby this occurs (38, 39). Recent studies have demonstrated that T_RM cells retain a degree of developmental plasticity as compared to T_CM cells despite possessing a more effector-like transcriptional profile (53). Specifically, these quiescent, tissue-resident progenitor T cells are responsible for maintaining a pool of tumor-specific effector T cells (33).

Flow cytometric analysis of BM samples obtained after only two vaccinations (C3D14) allowed us to examine how the immune microenvironment differs between patients with controlled disease compared to those who experienced early relapse. Unsupervised clustering of the post-vaccination BM memory CD8⁺ T cells identified a subpopulation associated with long-term disease remission. These DNAM1^low CD27⁻ memory CD8⁺ T cells were virtually absent in patients who experienced early post-vaccine relapse. Accordingly, CD27⁻ CD8⁺ T cells with a heterogeneous, partially dysfunctional phenotype, defined by the combined expression of both exhaustion and activation markers, have been recently identified as a potential source of MM-reactive lymphocytes and associated with positive outcomes in newly diagnosed MM (30). On the basis of the MM-MGUS model of persistent immune surveillance, the loss of tumor-reactive CD8⁺ T cell subpopulations would significantly contribute to immune escape and clinically meaningful disease progression. In support of this hypothesis, we show that the lack of BM-resident, effector-like memory CD8+ T cells associated with a significantly reduced frequency of MHC-class II+ cells preceded clinically evident disease relapse (Fig. 5).

Emerging evidence highlights the direct correlation between the depth of MRD response and clinical outcomes (54). Our baseline data certainly supports these findings. However, the stable reappearance of the monoclonal protein without meeting the criteria of disease progression in our study provides evidence that MM-GVAX can impart effective MM control by both deepening the clinical response and/or re-establishing the MM-MGUS equilibrium thereby delaying disease progression when administered in a low disease burden state.

Limitations of our study are its single-arm design and the limited sample size. Since the treatment was a combination of Len and MM-GVAX and there was no Len monotherapy comparator, the deepening of clinical responses, prolonged progression free survival and epitope spreading cannot be definitively linked to the vaccine. Moreover, due to limited sample availability, the precise mechanisms that orchestrate the immune niche in the BM could not be unequivocally identified and dissected. However, our clinical results would clearly suggest both clinical and biologic activity. This has resulted in the design of a double-blinded, randomized phase II trial to definitively prove clinical efficacy (NCT03376477). In summary, we have demonstrated the safety and feasibility of this MM-GVAX vaccine in combination with lenalidomide and shown the development of long-term MM-specific immune responses in patients with low disease burden. We also observed durable clinical responses and identified immune subsets that are associated with prolonged disease control.

Current MM therapies have dramatically increased the depth of clinical responses. However, effective therapies for the treatment of MRD-positive MM are currently lacking and no consensus on the management of this ever-growing patient population exists. Here, we suggest that MM vaccination in combination with lenalidomide might be an effective treatment to prevent disease relapse in patients who did not achieve complete eradication of MM clones or are known to have short-lived response durations even after achieving MRD negativity(13).

Statement of significance.

MM-GVAX vaccine in combination with lenalidomide induces durable immunity and long-term disease control in a setting of a low disease burden. To our knowledge, this represents the first clinical trial evaluating a novel paradigm for the treatment of multiple myeloma with minimal disease burden. A randomized phase II clinical trial is currently ongoing (NCT03376477).