T-cell transcriptional profiling and immunophenotyping uncover LAG3 as a potential significant target of immune modulation in multiple myeloma

Abstract

Autologous Stem Cell Transplant (ASCT) is the standard of care for patients with multiple myeloma (MM). The clinical significance of peripheral blood T-lymphocyte (PBTL) immunologic changes associated with ASCT is poorly understood. Here we evaluated T-cell transcriptional mRNA profiles and immunophenotypes to correlate immunologic senescence, exhaustion, and anergy with clinical endpoints in a cohort of MM patients undergoing ASCT.

Results:

ASCT induced global transcriptional T-cell changes and altered molecular levels of markers of T-cell subtypes, T-cell activation and exhaustion. These included reduced CD4:CD8 ratio, skewing towards the Th-1 subset, reduced expression of co-stimulatory receptors CD27 and CD28, heightened T-cell activation, and increased expression of immune modulatory molecules LAG3 and PD1. Multicolor flow cytometry experiments confirmed altered circulating CD4 and CD8 subsets and skewing towards differentiated effector cells. Moreover, ASCT promoted an exhausted immunophenotype in CD3⁺CD4⁺ subsets and a senescent immunophenotype in CD3⁺CD8⁺ subsets. Subset-specific altered expression was also seen for surface molecules with immune modulatory function. ASCT affected soluble levels of molecules with immunomodulatory function by increasing plasma HVEM and TIM3. High molecular LAG3 level was associated with inferior EFS post-ASCT (HR=5.44, CI 1.92–15.46, p=0.001, adjusted p (controlling for false discovery rate) =0.038).

Conclusion:

Using a comprehensive evaluation of PBTL on a molecular and phenotypic level, we have identified that ASCT induces global T-cell alterations with CD4 and CD8 subset-specific changes. Moreover, LAG3 emerged as an early biomarker of adverse events post-ASCT. These findings will support the development of treatment strategies targeting immune defects in MM to augment or restore T-cell responses.

Introduction

Immunological aging is related to the development of cancer, but factors that impact this relationship are only partially understood. Several studies demonstrated that aging correlates with increased peripheral blood T lymphocyte (PBTL) expression of p16^INK4a [1, 2], a molecular marker associated with cellular senescence and irreversible cell cycle arrest [3]. PBTL p16^INK4a levels not only increase with chronological age, but are influenced by physiologic stress caused by smoking, chemotherapy, and physical inactivity [3]. Within the overall aging of the immune system, T cells also undergo immunological senescence, which is associated with shifts in T-cell production, subsets, and function, including poor antigen recall [4]. Exhaustion is another form of T-cell dysfunction resulting from chronic antigen stimulation provided by persistent infections, chronic inflammation, or tumors. Hallmarks of T-cell exhaustion include increased expression of immune checkpoint molecules such as PD1, CTLA4, TIM3 and LAG3, resulting in altered production of pro-inflammatory cytokines, and reduced proliferation in response to antigen [5, 6].

Multiple Myeloma (MM), a cancer that increases sharply in incidence with age, is also associated with severe and complex T-cell defects that resemble changes seen in chronologic aging [7, 8]. The relationship of PBTL homeostasis to biologic age in the context of MM therefore warrants further evaluation. PBTL changes in MM include inversions of the CD4:CD8 [9] and T-helper 1:T-helper 2 (Th1:Th2) ratios [10], as well as an expansion of regulatory T cell (T_REG) populations [11, 12] and a shift to highly activated and differentiated effector cells [13]. Of note, expanded CD8⁺ peripheral blood T-cell clones were recently shown to be associated with a favorable prognosis [14, 15]. This is particularly interesting as these clones exhibited impaired effector function and acquired an immune senescent effector phenotype. Findings in MM patients regarding the expression of immune checkpoint molecules are somewhat controversial. While clonal CD8⁺ T cells show reduced PD1 and CTLA4 expression, high immune checkpoint molecule expression overall was observed in CD8⁺ T cells in both blood and marrow [13, 15]. Moreover, various components of the myeloma microenvironment, such as bone marrow stromal cells, also utilize immune checkpoint signaling axes to provide tumor-promoting immune escape niches and to modify anti-tumor T-cell immune responses [8].

MM-associated T-cell defects are further accelerated by autologous stem cell transplant (ASCT) [7]. We and others have demonstrated that T-cell p16^INK4a expression increases as a result of ASCT, suggesting that transplantation accelerates biological T-cell aging [16, 17]. Moreover, computational models based on RNA sequencing data suggest that ASCT alters the T-cell subset composition to a profile that usually occurs with chronological aging [17]. A direct link between ASCT, altered T-cell phenotype and CD8⁺ T-cell exhaustion was recently described in a murine transplantation model [18]. In this work, inhibitory receptor expression correlated strongly with myeloma progression, and antibody-mediated blockade of PD-1 and TIGIT signaling improved myeloma control [18].

The clinical significance of T-cell changes in MM patients and how these associate with aging are poorly understood. Furthermore, capitalizing on advancements in immune-modulatory therapies is increasingly relevant in MM treatment. Thus, an understanding of the full potential of T-cell related biomarkers in MM and the therapeutic targeting of T-cell function in the context of ASCT is essential. In this report, we characterized ASCT-related T-cell changes and immune checkpoint molecule expression in MM patient samples to identify biomarkers of clinical responses and inform the clinical development of T-cell modulatory therapies.

METHODS AND MATERIALS

Study Cohort

We performed a single-institution prospective study of 100 adults with plasma cell dyscrasia (OSU-13135) and analyzed samples from 80 patients who proceeded to ASCT. A detailed description of the study protocol was previously published[19]. The study was approved by the Institutional Review Board of The Ohio State University and written informed consent was obtained for all patients in accordance with the Declaration of Helsinki. Molecular T-cell changes were examined in 80 patients and 10 age- and sex- matched healthy controls using a custom NanoString Codeset (OSU_Senescence). Sequential pre-and 90 day post-transplant NanoString data were available for 31 patients. Using the same sample set, 20 representative samples and 10 healthy controls underwent further peripheral blood mononuclear cell (PBMC) immunophenotyping (Table 1). Plasma levels of soluble immune checkpoint molecules were examined in a subset of 16 patients and 7 controls. A schematic overview of patient samples examined for molecular and immunophenotypic characteristics of T-cell immune senescence and exhaustion (Supplemental Figure 1).

Table 1:

Pre-transplant clinical demographics in 20 patients whose samples were analyzed by flow cytometry. This group is representative of the overall cohort with NanoString data available before ASCT.

Pre-transplant clinical demographics (n=20)
Age, median (range)	59.5 (41–68)
Gender, no. (%)
Female	9 (45)
Male	11 (55)
Histology, no. (%)
MM	19 (95)
AL amyloidosis	1 (5)
International staging system (ISS)a, no. (%)
1	10 (56)
2	6 (33)
3	2 (11)
Pre-transplant disease status, no. (%)
Complete response (CR)	3 (16)
Very good partial response (VGPR)	4 (21)
Partial response (PR)	6 (32)
Stable disease (SD)	2 (11)
Progressive disease (PD)	3 (16)
Not evaluable (NE)	1 (5)
N/A or unknown	1
Number of regimens, median (range)	2 (1−4)
Days from diagnosis to transplant, median (range)	169 (range 97–1342)
Induction therapy, no. (%)
Cytoxan, lenalidomide, dexamethasone (CRd)	1 (5)
Cytoxan, bortezomib, dexamethasone (CyBorD)	6 (30)
Bortezomib, lenalidomide, dexamethasone (VRd)	5 (25)
Lenalidomide, dexamethasone (RD)	2 (10)
Bortezomib, dexamethasone (VD)	6 (30)
Lines of therapy, no. (%)
One line	15 (80)
More than one lineb	5 (20)
B2M (mg/L), median (range, unknown)	1.9 (1.3–5.2, 1)
Creatinine (mg/dL), median (range)	0.9 (0.6–2.0)
Albumin (g/dL), median (range)	4.2 (0.1–4.6)
WBC (K/µl), median (range)	5.1 (1.8–14.1)
Hemoglobin (g/dL), median (range)	12.1 (9–14.9)

^a.2 patients unknown

^b.one patient with prior ASCT

Molecular T-Cell Analysis

CD3⁺ T cells were enriched from whole blood using RosetteSep negative isolation (StemCell Technologies) following the manufacturer’s recommendations. T-cell RNA was extracted with an RNeasy Plus Mini Kit (Qiagen) and analyzed using a custom NanoSstring Codeset OSU_Senescence platform [20]. Markers on the OSU_Senescence Codeset include standard housekeeping genes as well as indicators of T-cell subsets, cellular senescence, immune senescence, activation and exhaustion. cDNAs for CD244, CD274, CD276, IL-6 and IL-17a were pre-amplified (12 cycles) using custom primers, followed by standard nCounter analysis. Data normalization was performed using nSolver software. Molecular patterns were first assessed in all available samples before and after ASCT to avoid bias caused by differences in patients who did and did not undergo ASCT. Analyses seeking associations between mRNA levels and outcome were limited to patients with complete sequential pre-and post-ASCT data.

T-cell Immunophenotyping

Antibodies, clones, optimal concentrations and suppliers are listed in Suppl. Table 1. All flow cytometry panels were optimized on cryopreserved PBMCs isolated from a healthy donor. Patient and healthy control PBMCs were concurrently stained using antibody cocktails. After resuspending each sample in 100µl of fixable viability dye (eBioscience) diluted 1:1000 in PBS:Brilliant Buffer (BD) 1:1, antibody cocktail was added and the cells were incubated at 4°C for 30 min. Cells were then washed in PBS and fixed in 1.7% formaldehyde in PBS/1% BSA for 15 min at room temperature. After a second wash, the cells were stored overnight and analyzed on a BD Fortessa II. The sensitivity and linearity of the cytometer system were assessed before and after the acquisition of experimental and control samples using 6 peaks SPHERO™ Rainbow Calibration Particles (BD). An aliquot of PBMCs from the same donor and prepared under identical conditions served as the internal standard control across patient samples and experimental days. Fluorescence-minus-one (FMO) controls were prepared for every examined marker and served as gating controls. We determined surface expression of PD1, PDL1, LAG3 andCTLA4 by correcting the median fluorescence intensity (MFI) for each sample to the respective FMO control and then normalizing this value to the corrected MFI of the daily internal standard control. T-cell subsets were classified as naïve, central memory (CM), effector memory (EM), and effector memory RA (EMRA for CD8) or effectors (for CD4) using C-C chemokine receptor type 7 (CCR7) and CD45RA expression [21]. Senescent, exhausted, anergic and activated T-cells were characterized based on CD57, PD1 and CD28 expression [14]. All data were analyzed using KALUZA software (Becton Dickinson).

Soluble Plasma Immunomodulatory Molecules

Plasma from patients (n=16) and matched controls (n=7) was analyzed using a customized Milliplex MAP Human Immuno-Oncology Checkpoint Panel (Millipore) for soluble CD27, CD28, TIM3, HVEM, LAG3, PD1, CTLA4 and PDL1. Samples were prepared following the manufacturer’s recommendations and run in technical triplicates. Data was acquired and analyzed on a MagPix cytometer.

Statistical Analyses

Linear mixed models with unstructured covariance matrices were used to test for changes in gene expression following transplant adjusting for patient age. The Kenward-Roger method was used to compute degrees of freedom for these tests [22]. Cox proportional hazards models were used to measure the association between gene expression and event-free survival defined as time until disease recurrence or death, whichever occurred first. In each analysis, the Benjamini-Hochberg method was used to control the false discovery rate across all genes tested [23]. Immunophenotypic analyses were considered as validations of molecular patterns. Datasets were compared with an unpaired t test (controls vs. patients) or a Wilcoxon matched pairs signed rank test (patients pre vs. post ASCT).

RESULTS

ASCT induces global transcriptional T-cell changes

Between April 2014 and October 2015, PBTLs were collected before and after ASCT from 80 patients with plasma cell dyscrasia treated at OSU. Characteristics of the study population and protocol are previously described [20]. The mean and standard error (SE) of RNA expression of markers of T-cell subtypes, T-cell activation and exhaustion were determined before and after transplant and adjusted for age. ASCT led to a significant reduction in CD3 (age-adjusted mean change −0.13, SE 0.05, adjusted p-value 0.038) and CD4 (−0.44, SE 0.12, p=0.003) and induced increased CD8 (1.22, SE 0.11, p<0.0001) mRNA levels, resulting in a reduction of the CD4:CD8 ratio (−1.71, SE 0.19, p<0.0001, Figure 1A). TBCX21 (T-bet) mRNA levels increased significantly after ASCT (1.7, SE 0.14, p<0.0001, Figure 1A), suggesting skewing towards the Th-1 subset. Similarly, ASCT led to increased mRNA levels of IFN-y (0.47, SE 0.13, p=0.004, Figure 1B), which is a key cytokine driving Th1 differentiation [24]. ASCT also induced a significant reduction of CD27 (−0.32, SE 0.11, p=0.015) and CD28 (−0.46, SE 0.10, p<0.0001) mRNA levels, indicating reduced expression of co-stimulatory receptors and heightened T-cell activation (Figure 1B). Increased T-cell activation was also reflected in significantly increased expression of immune modulatory molecules such as LAG3 (1.08, SE 0.14, p<0.0001) and PDCD1 (PD1, 1.08, SE 0.12, p<0.0001, Figure 1C), and in the terminal differentiation marker KLRG-1 (1.24, SE 0.14, p<0.0001). mRNA encoding B3GAT1, which is associated with clonal exhaustion and replicative senescence, was also significantly increased following ASCT, although this comparison was based on a smaller number of pre- (n=13) and post- (n=19) ASCT samples (0.53, SE 0.2, p=0.038). Interestingly, BTLA (−0.53, SE 0.1, p<0.0001) and CD274 (PDL1, −0.61, SE 0.25, p=0.038, Figure 1C) mRNA decreased after ASCT, suggesting selective usage of immune checkpoint pathways in this setting.

Figure 1: Changes in T-cell mRNA markers before and after ASCT:

RNA levels of established markers of T-cell subsets, stimulation, activation and exhaustion were normalized to a set of standard housekeeping genes using a custom NanoString marker codeset and assessed in all available samples before and after ASCT. Linear mixed models with unstructured covariance matrices were used to test for changes in gene expression following transplant adjusting for patient age. All graphs show mean±SEM (standard error of the mean). RNA levels measured in healthy age-matched controls are included for descriptive purposes. (A) Changes in mRNA levels of CD3, CD4, CD8A and TBX21 (Tbet) as markers of T-cell subsets. (B) Changes in mRNA levels of CD27, CD28 and IFN-γ as markers of T-cell stimulation and activation. (C) Changes in mRNA levels of BTLA, CD274 (PDL1), LAG3, PDCD1 (PD1), B3GAT1 and KLRG1.

ASCT alters circulating CD4 and CD8 subsets

We next examined whether transcriptional changes discovered using the OSU_Senescence platform were reflected in the immunophenotype of PBTL. ASCT led to a reduction of the percentage of CD3⁺ T cells (p=0.0349, Figure 2A) accompanied by an inversion of the CD4:CD8 ratio (p<0.0001, Figure 2B, C). Percentages of CD25⁺CD127⁻ Tregs were significantly higher in pre-transplant MM patients compared to healthy controls (p=0.0197) and post-transplant patients (p<0.0001, Figure 2D). Among CD3⁺CD4⁺ T cells, ASCT resulted in a reduction of the percentage of CCR7⁺CD45RA⁺ naïve cells (p=0.0054), and an enrichment of CCR7⁺CD45RA⁻ central memory (CM) (p=0.0497) and CCR7⁻CD45RA⁻ effector memory (EM)(p=0.0004) cells (Figure 2E). Among CD3⁺CD8⁺ cells, this skewing was even more pronounced, as ASCT induced a loss of CCR7⁺CD45RA⁺ naïve cells (p<0.0001) and an expansion of CCR7⁺CD45RA⁻ CM (p<0.0001), CCR7⁻CD45RA⁻ EM (p=0.0054) and CCR7⁻CD45RA⁺ TEMRA (p=0.003) cells (Figure 2F). Altogether, these data show that T-cell subset changes and skewing towards differentiated effector cells indicated by our OSU-Senescence platform are recapitulated in the immunophenotype of PBTL from study participants.

Figure 2: Immunophenotypic characterization of T-cell subset changes:

Cryopreserved peripheral blood mononuclear cells (PBMCs) were analyzed by multicolor flow cytometry and compared between healthy controls and MM patients before and after ASCT. Datasets were compared with an unpaired t test (controls vs. patients) or a Wilcoxon matched pairs signed rank test (patients pre vs. post ASCT). All graphs show mean±SEM, with the exception of (E) and (F) where only the mean is shown. (A) Changes in relative frequencies of overall CD3⁺ T cells, (B) CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells, (C) CD4:CD8 ratio, (D) regulatory T cells (Tregs), (E) CD4 CD45RA⁺CCR7⁺ naïve, CD45RA⁻CCR7⁺ central memory (CM), CD45RA⁻CCR7⁻ effector memory (EM) and CD45RA⁺CCR7⁻ effector subsets, (F) CD8 CD45RA⁺CCR7⁺ naïve, CD45RA⁻CCR7⁺ CM, CD45RA⁻CCR7⁻ EM and CD45RA⁺CCR7⁻ terminally differentiated T cells (TEMRA), with representative flow plots demonstrating CCR7 and CD45RA expression.

ASCT results in CD4- and CD8-specific exhaustion, senescence and anergic phenotypes

We next examined co-expression patterns of CD57 and PD1 to further determine how ASCT affects exhausted, senescent, or anergic CD4 and CD8 subsets in peripheral blood. The gating strategy and definition of specific subsets are summarized in Figure 3A and Suppl. Table 2. Among CD3⁺CD4⁺ cells, ASCT led to aberrant CD57 and PD1 expression in naïve, CM and EM subsets (Figure 3A). After ASCT, naïve, CM and EM CD3⁺CD4⁺ subsets predominantly expanded PD1⁺CD57⁻ exhausted cells over PD1⁺CD57⁺ anergic and PD1⁻CD57⁺ senescent cells (Figure 3B). Among CD3⁺CD8⁺ cells, ASCT led to aberrant CD57, CD28, and PD1 expression in all subsets (i.e. naïve, CM, EM and effector memory RA (TEMRA) cells, Figure 3A). In contrast to what was observed in the CD3⁺CD4⁺ subset, ASCT also significantly reduced frequencies of naïve and CM CD3⁺CD8⁺ cells, with an enrichment of PD1⁻CD57⁺ senescent and PD1⁺CD57⁺ anergic cells and only minimal changes in the frequencies of PD1⁺CD57⁻ exhausted CD8 T cells after ASCT (Suppl. Figure 2). The enrichment of senescent cells was especially pronounced in EM and EMRA CD8 subsets (Figure 3C). Together, these data show that ASCT promotes an exhausted immunophenotype in CD3⁺CD4⁺ subsets and a senescent immunophenotype in CD3⁺CD8⁺ subsets, providing a more detailed T-cell characterization than the previously described profiles in CD3⁺ T cells overall.

Figure 3: Comparison of exhausted, senescent and anergic CD4 and CD8 T-cell subsets:

Cryopreserved peripheral blood mononuclear cells (PBMCs) were analyzed by multicolor flow cytometry and compared between healthy controls and MM patients before and after ASCT. Datasets were compared with an unpaired t test (controls vs. patients) or a Wilcoxon matched pairs signed rank test (patients pre vs. post ASCT). All graphs show mean. (A) Gating strategy and definition of PD1⁻CD57⁺ senescent, PD1⁺CD57⁺ anergic and PD1⁺CD57⁻ exhausted subsets of CD4 naïve, CM, EM and CD8 naïve, CM, EM, EMRA and effector cells. CD4 effector cells were not further analyzed due to overall very low frequencies of these cells. Surface PD1 and CD57 were used to identify PD1⁻CD57⁺ senescent, PD1⁺CD57⁺ anergic and PD1⁺CD57⁻ exhausted cells. (B) ASCT-induced exhausted, anergic and senescent CD4⁺ T-cell subset changes. (C) ASCT-induced exhausted, anergic and senescent CD8⁺ T-cell changes.

ASCT alters expression of surface antigens with immunomodulatory function

We next used flow cytometry to characterize the expression of T-cell surface antigens with known immunomodulatory functions (i.e. co-stimulatory and inhibitory receptors). ASCT significantly increased the median fluorescence intensity (MFI) of CD28 (broadly indicative of increased expression) on overall CD3⁺CD4⁺ T cells (pre- vs. post-ASCT p=0.0018) and especially on CM cells (p<0.0001, Figure 4A). A trend towards decreased CD28 expression was seen in CD3⁺CD8⁺ T cells from MM patients, which was significantly further reduced after ASCT (p=0.0002), especially on EM (p=0.0001) and EMRA (p=0.0016) cells (Figure 4B). As CD28 is a commonly used marker to differentiate senescent, anergic and exhausted cells (Suppl. Table 2), similar to PD1, these observations are in line with an enrichment of CD28⁺ exhausted and anergic CD4⁺ subsets and CD28⁻ senescent CD8⁺ subsets (Suppl. Figure 3A+B). In line with the increase in PD1 expressing T-cell subsets described above, surface PD-1 was generally significantly increased after ASCT not only in overall CD3⁺CD4⁺ (p=0.01) and CD3⁺CD8⁺ (p=0.0327) cells (Suppl. Figure 4A), but also in CD25⁺CD127⁻ Tregs (Figure 4C). LAG3 expression was generally low, but increased significantly after ASCT on overall CD3⁺CD4⁺ (p=0.0474) cells, mainly via an upregulation on naïve (p=0.0052) and CM (p=0.0113) subsets (Figure 4D). ASCT did not have an effect on LAG3 expression on overall CD3⁺CD8⁺ cells, as LAG3 increased significantly on naïve CD3⁺CD8⁺ (p=0.0015) but decreased significantly on EMRA CD3⁺CD8⁺ cells (p=0.0329, Figure 4E). No significant overall changes were observed for PD-L1 and CTLA4 expression, although ASCT increased the expression on a proportion of naïve CD4 and CD8 cells (Suppl. Figure 4B, C). Together, these data indicate that ASCT modulates the expression of immunomodulatory surface molecules in a subset-specific manner.

Figure 4: Changes in surface and circulating molecules with immunomodulatory function:

(A-E) Cryopreserved peripheral blood mononuclear cells (PBMCs) were analyzed by multicolor flow cytometry and compared between healthy controls and MM patients before and after ASCT. Datasets were compared with an unpaired t test (controls vs. patients) or a Wilcoxon matched pairs signed rank test (patients pre vs. post ASCT). All graphs show mean±SEM. Median fluorescence intensities (MFI) were corrected against daily fluorescence-minus-one (FMO) control and normalized to MFI of internal standards. (A) CD28 MFI changes on overall CD3⁺CD4⁺ T cells (top panel) and representative histograms showing CD4 naïve, CM, EM and effector subsets. (B) CD28 MFI changes on overall CD3⁺CD8⁺ T cells (top panel) and representative histograms showing CD8 naïve, CM, EM and EMRA subsets. (C) PD1 MFI changes in regulatory T cells. (D) LAG3 MFI changes on CD3⁺CD4⁺ and (E) CD3⁺CD8⁺ subsets. (F) Soluble immune modulatory molecules CD27, CD28, HVEM, TIM3, PDL1, PD1, CTLA4 and LAG3 in plasma.

ASCT significantly increases levels of soluble HVEM and TIM3

Soluble components with immunomodulatory function are increasingly recognized as important mediators of tumor immunoediting [25]. In MM, high levels of soluble PD-L1 have been associated with poor outcome [26, 27]. We therefore examined the effect of ASCT on soluble immunomodulating molecules CD27, CD28, HVEM, TIM3, PDL1, PD1, CTLA4 and LAG3 (Figure 4F). Herpes virus entry mediator (HVEM) is a binding partner of CD160 and BTLA4. Plasma HVEM levels were significantly higher in MM patients compared to healthy controls (p=0.0263), and increased further after ASCT (p=0.0233). ASCT also significantly increased soluble TIM3 levels in MM patients (p=0.0085). No significant changes were seen for PD1, PDL1, LAG3, CTLA4, CD27, and CD28. Together, these results demonstrate upregulation of soluble TIM3 and HVEM post-ASCT in MM patients, suggesting a role of secreted immunomodulatory molecules in mediating T-cell responses in this setting.

Molecular post-ASCT LAG3 level is associated with inferior EFS

The association between p16^INK4a as a molecular marker of T-cell senescence and clinical outcomes such as fatigue has been reported [20]. To assess whether global T-cell changes could serve as predictors of clinical outcome, we next examined the associations between PBTL immunomodulatory marker transcript expression among a subset of patients (n=31) evaluated 90 days post-ASCT, with event-free survival (EFS), defined as relapse or death. LAG3 was the only molecular PBTL transcript profile with a significant association with clinical endpoints. The median EFS for patients with high LAG3 expression was 641 days, while the median EFS was not reached in patients with low LAG3 transcript expression (HR=5.44, CI 1.92–15.46,) p=0.001 for one unit increase in expression, adjusted p (controlling for false discovery rate) = 0.038). The association between LAG3 transcript expression and EFS adjusted for age was aHR=5.66, CI 1.83–17.47, p=0.003, adjusted p (controlling for false discovery rate) = 0.056 (Figure 5). LAG-3 RNA transcript expression was measured serially for a subset of patients (n=9) at Visit 1 (V1) prior to transplant, ~ 90 day post-transplant at Visit 2 (V2) and ~ 1 year following transplant Visit 3 (V3). Patients had PBTL LAG-3 transcript expression at V1 and V3 only (n=6) and serial visits at V1, V2, V3 (n=3). The mean LAG-3 RNA transcript at was significantly increased at 1-year (V3) in contrast to baseline (V1) (V1 mean=6.89 (range 5.65 – 8.33), V3 mean=9.10 (range 8.29–9.96); p=<0.05). (Figure 6). These data reveal that high LAG3 transcript expression has a significant association between relapse and death among post-transplant MM patients and furthermore that LAG-3 transcript expression increases are sustainable. No associations were found with other molecular or immunophenotypic markers of T-cell dysfunction.

Figure 5: Prognostic value of LAG3 mRNA expression:

The median EFS for patients with high LAG3 mRNA after ASCT (defined as expression > median LAG3) was 641 days, and not reached in patients with low LAG mRNA after ASCT.

Figure 6.

RNA transcript expression was measured serially for a subset of patients (n=9) at Visit 1 (V1) prior to transplant), ~ 90 day post-transplant at Visit 2 (V2) and ~ 1 year following transplant Visit 3 (V3). Patients had PBTL LAG-3 transcript expression at V1 and V3 only (n=6) and serial visits at V1, V2, V3 (n=3). The mean LAG-3 RNA transcript at was significantly increased at 1-year (V3) in contrast to baseline (V1) (V1 mean=6.89 (range 5.65 – 8.33), V3 mean=9.10 (range 8.29–9.96); p=<0.05).

DISCUSSION

The composition and dynamics of the adaptive immune system in MM, both before and after transplant, are critical determinants of disease control and patient survival. Understanding these aspects will also be necessary to inform future treatment strategies including immune therapies. Here, we studied the transcriptional and phenotypic composition of PBTL in MM patients before and after ASCT to provide new information regarding the T-cell defects in this disease. In particular, an understanding of T-cell anergy, exhaustion, and senescence will ultimately allow us to capitalize on reversible functions in eradicating malignant plasma cells via anti-tumor immunity and enhanced cytotoxic capacity.

T-cells are termed ‘exhausted’ when cells lose cytotoxic capacity, show sustained expression of inhibitory checkpoint molecules such as PD-1, LAG3, CD160 or 2B4, and have distinct altered transcription profiles in CD4 and CD8 subsets [6]. Senescent PBTL are irreversibly in cell cycle arrest with downregulated CD28 and upregulated CD56, yet continue to produce pro-inflammatory cytokines. Anergic PBTL are in a hypo-responsive state, showing defective production of IL-2, IFN-γ, and TNF, typically accompanied by an increase in Foxp3⁺ regulatory T cells [28, 29]. PBTL of MM patients are known to be dysfunctional with overlapping phenotypes of exhaustion, anergy and senescence [30–32]. Here we demonstrate that ASCT further alters the T-cell phenotype in MM, including inversion of the CD4:CD8 ratio, upregulation of CD25⁺CD127⁻ Tregs, and a skewing toward differentiated effector cells. We then explored the molecular transcript profile and phenotype of PBTL to define early markers of clinical relapse that may inform the design of future therapeutic strategies.

ASCT resulted in expansion of CD4⁺ naïve, CM and exhausted PD1⁺CD57⁻ cells and a reduction of PD1⁺CD57⁺ anergic and PD1⁻CD57⁺ senescent cells. In contrast, CD8⁺ PD1⁻CD57⁺ senescent and PD1⁺CD57⁺ anergic cells were increased, and only minimal changes were observed in the frequencies of CD8⁺ PD1⁺CD57⁻ exhausted cells. Thus, our data show that ASCT promotes an exhausted immunophenotype in CD4⁺ cells and a senescent/anergic immunophenotype in CD8⁺ subsets. This PBTL profile is in contrast to that reported by Suen et al. [14], in which the authors examine a large cohort of MM patients (on treatment, not on treatment, and after ASCT) and describe the PBTL populations as senescent and neither anergic nor exhausted. These differences in conclusions likely reflect analysis of cells from patients at various treatment stages and not further separated by CD4 vs. CD8 status.

We also observed differential expression of CD28 on PBTL following ASCT, with significantly increased expression on CD4⁺ and decreased expression on CD8⁺ PBTLs. Other changes in inhibitory receptor expression included increased PD1 on Tregs and upregulation of LAG3 on CD4⁺/CD8⁺ naïve and CD4⁺ central memory subsets. Importantly, increased expression of LAG3 in PBTL at 90 days post-ASCT was found to be an early indicator of relapse or death. These findings relating to increased LAG3 expression post-ASCT, are in contrast to a publication by Chung et al, who reported no significant increase in the expression of inhibitory receptors at similar timepoints [33]. This difference may be attributable to small sample size and the generally low level expression of this marker. Given the associations with poor outcomes outcome [26, 27], we also assessed expression of soluble immune modulatory receptors. We found that HVEM plasma levels in MM patients were higher compared to healthy controls and increased further following ASCT, as were soluble TIM3 levels. These data further support the role of soluble ligands as diagnostic indicators, which is a focus of our ongoing work.

One of the major limitations of our findings is the lack of functional studies of PBTL before and after ASCT and the focus on the circulating T-cell pool. Evaluating TCR signaling and activation states of PBTL subsets will be needed to determine antigen responsiveness, proliferation, and cytotoxic capacity in the context of ASCT. Several reports have described dysfunctional PBTL in MM patients [8, 34], Treg function in response to IMiD therapy [35], and responsiveness to checkpoint blockade [8, 33–36]. Others have explored circulating lymphocytes in comparison to the bone marrow niche, demonstrating more severe impairment of T-cell proliferation and function within the bone marrow microenvironment [13]. Therefore, while opportunities remain to evaluate T-cell function in MM patients, our findings are an important step toward understanding these defects so they can ultimately be targeted therapeutically. Our findings indicate that CD8⁺ subsets post-ASCT exhibit a more senescent phenotype, although it is unclear if this population of cells is able to restore anti-tumor immunity. We previously reported that ASCT accelerates and induces the molecular aging of PBTL with increased expression of p16^INK4a [16], and also that increased PBTL p16^INK4a expression has a modest relationship with post-transplant fatigue [20]. Understanding potential therapeutic targets on PBTL, particularly checkpoint inhibitors on cells that are preferentially exhausted or anergic, is a key step toward improved treatment approaches. To this end, our data suggest that LAG3 might be an ideal target to preferentially reactivate T-cells in MM patients. Inhibition of PD-1 in combination with IMiD therapy has not been successful in MM, resulting in early trial discontinuations[37, 38]. In the post-transplant setting, a recent study involving PD-1 blockade myeloma patients who did not achieve a complete response after ASCT, was terminated early after failing to meet the interim endpoints and was considered feasible in combination with lenalidomide, yet ineffective[39]. However, newer evidence demonstrates efficacy of simultaneous blockade of LAG3 and PD-1 in pre-clinical models [40, 41] and acceptable safety profiles in early phase clinical trials in predominately solid tumor trials [42]. Thus, our findings support future clinical investigations of LAG3-directed immune checkpoint blockade in MM, alone or in combination with other agents.

In summary, by evaluating PBTL subsets transcriptionally and phenotypically, we have identified differential senescence in CD8 populations and anergy/exhaustion in CD4 populations. Moreover, we identified LAG3 as an early biomarker of adverse events. Future work targeting immune defects in MM is warranted to augment or restore T-cell responses using the next generation of immune checkpoint inhibitors.

Highlights.

• ASCT induces global transcriptional T-cell changes and altered molecular levels of markers of T-cell subtypes, T-cell activation and exhaustion in multiple myeloma patients.

• High LAG3 transcript expression in T-cells detectable as early as 3-months post-transplant is associated with inferior clinical outcomes in myeloma patients.

• ASCT affected soluble levels of molecules with immunomodulatory function by increasing plasma HVEM and TIM3.

• ASCT promotes an exhausted immunophenotype in CD4⁺ cells and a senescent/anergic immunophenotype in CD8⁺ subsets.

• Future work targeting immune defects in Multiple Myeloma is warranted to augment or restore T-cell responses using the next generation of immune checkpoint inhibitors.