Bone marrow CD8 + T cell function declines in Multiple Myeloma and is consistently lower than in matched peripheral blood
Phenotyping of major T cell subsets in BM and PB mononuclear cells from age-matched controls, individuals with monoclonal gammopathy of undetermined significance (MGUS), asymptomatic MM and MM at diagnosis identified that BM CD8+ T cell abundance decreased with disease development (Supplementary Fig. 1A, Fig. 1A), whilst frequencies of BM CD4+ T cells (Fig. 1B), as well as PB CD8+ and CD4+ T cells remained similar (Supplementary Fig. 1B-C). BM CD56+ NK cell abundance also did not change with disease development, whilst CD4+CD127−CD25+ regulatory T cells (TReg) were enriched in MM BM at diagnosis, agreeing with previous literature (Supplementary Fig. 1A,D-E)(5, 7). In PB samples, NK cells and TReg frequencies were similar at all disease stages (Supplementary Fig. 1D,E). We next assessed the proportion of T cells within BM and PM samples expressing the key anti-tumour cytokines interferon-gamma (IFN-γ) and tumour necrosis factor-alpha (TNF-α), as well as interleukin 2 (IL-2) and the cytotoxic molecule Granzyme B. This identified decreased frequencies of IFN-γ- and IL-2-expressing CD8+ T cells, as well as multifunctional CD8+ T cells expressing all three cytokines, and those expressing Granzyme B in the BM with MM development (Fig. 1C). Frequencies of CD4+ T cells expressing these molecules did not change significantly (Fig. 1D). These observations were specific to BM CD8+ T cells, whilst PB T cell cytokine and granzyme expression also did not change (Supplementary Fig. 1F-G). Indeed, pairwise comparison of BM and PB samples identified that CD8+ T cells were consistently less functional within BM samples than matched PB samples. Frequencies of IFN-γ-, TNF-α-, IL-2- and Granzyme B-expressing CD8+ T cells were lower in BM than PB in all patient groups, including controls, albeit the disparity increased with disease development for IFN-γ and TNF-α. When all samples were combined, this difference was highly significant (Fig. 1E, Supplementary Fig. 1H). However, this was not the case for CD4+ T cells, which demonstrated similar functionality within BM and PB samples, and increased multifunctional cells in the BM (Fig. 1F, Supplementary Fig. 1I), indicating specific CD8+ T cell suppression within the BM microenvironment. Expression of cytokines was largely restricted to memory populations of T cells, with EMRA CD8+ T cells being a significant source of IFN-γ, TNF-α and Granzyme B but not IL-2 (Fig. 1G), and EM CD4+ T cells the major IFN-γ expressers, with contribution of CM cells for TNF-α (Fig. 1H). However, proportions of these subpopulations were broadly similar between BM and PB at all disease stages (Figure I-J), meaning altered overall function of CD8+ T cells in BM vs. PB is not explained by enrichment or loss of functionally-distinct subpopulations, and may rather relate to an effect of the microenvironment.
CD8 + T cell mitochondrial mass and lipid uptake alter during progression of Multiple Myeloma and in the bone marrow microenvironment
T cell function is underpinned by metabolic capacity, particularly mitochondrial mass(12, 13). Analysis of BM T cell mitochondrial mass using the fluorescent probe Mitoview Green (MVG) revealed this decreased within CD8+ and CD4+ T cells with MM development, and was reduced in MGUS, asymptomatic MM and MM at diagnosis compared to controls (Fig. 2A-B). Similar to immune effector function however, these changes were less apparent in matched PB samples (Supplementary Fig. 2A-B). Of note, BM CD8+ T cells again demonstrated significantly reduced mitochondrial mass compared to paired PB cells (Fig. 2C), which was not the case for CD4+ T cells (Fig. 2D), indicating a potential link between decreased mitochondrial mass and reduced functionality of CD8+ T cells in the BM microenvironment. To directly interrogate the relationship between mitochondrial mass and immune effector function, we stained for CoxIV, a subunit of ETC complex IV, as a mitochondrial marker, together with cytokines (since MVG cannot be fixed for intracellular staining) in a subset of samples from controls and MM at diagnosis. CoxIV expression was higher in cytokine-expressing CD8+ T and CD4+ T cells than cytokine-negative populations (Fig. 2E, Supplementary Fig. 2D), confirming a relationship between mitochondrial mass and effector function. In these assays, TNF-α expression was more tightly linked to CoxIV abundance than IFN-γ, potentially explained by the well-documented relationship between glycolysis and IFN-γ expression in CD8+ T cells(30, 31). Further confirmation that mitochondrial mass is a key determinant of T cell effector function across this patient cohort was provided by correlative analyses of MVG fluorescence vs. frequency of cytokine expressing cells in the BM sample, which identified significant positive relationships for IFN-γ and TNF-α-expressing CD8+ and CD4+ T cells (Fig. 2F, Supplementary Fig. 2D).
Long-chain fatty acid uptake from tumour microenvironments is increasingly implicated in immune cell suppression through mechanisms including mitochondrial loss and dysfunction(19, 20) and lipid peroxidation-induced ferroptosis(23, 32). This may be pertinent in the lipid-rich BM, where decreased mitochondrial mass of CD8+ T cells was observed, therefore T cell capacity to take up long-chain fatty acids was assessed using a fluorescent palmitate probe, C16-BODIPY. Uptake of this probe did not change in BM or PB T cells during disease development (Fig. 2G, Supplementary Fig. 2E-F) however, comparison of paired BM and PB samples revealed markedly elevated lipid uptake in BM cells, particularly CD8+ T cells (Fig. 2H-I, Supplementary Fig. 2G). Among CD8+ T cells, C16-BODIPY uptake particularly mapped to antigen-experienced CM and EM populations and was also significantly higher in cells expressing the immune checkpoints TIGIT and PD-1 than checkpoint-negative cells (Fig. 2J). Lipid peroxidation within T cells from MM patients was also assessed using BODIPY 581/591, which fluoresces at different wavelengths before and after peroxidation. This identified that, in addition to increased lipid uptake capacity, BM T cells also demonstrate elevated lipid peroxidation than PB T cells (Fig. 2K, Supplementary Fig. 2H).
CD8+ T cell function is impaired by lipid uptake from the BM microenvironment via FATP1
Next, we established an in vitro model to interrogate the relationship between long-chain fatty acid uptake and T cell mitochondrial and immune dysfunction within the BM microenvironment. Specifically, we cultured BM mononuclear cells from control individuals or MM patients in autologous PB or BM plasma, stimulated T cells and assessed their mitochondrial mass and cytokine production. This established that BM plasma significantly decreased mitochondrial mass and IFN-γ expression by CD8+ T cells compared to PB plasma, which occurred in both control and MM samples (Fig. 3A-B). Conversely, CD4+ T cell mitochondrial mass was not altered in BM plasma, despite reduced IFN-γ expression (Supplementary Fig. 3A-B). Total secreted IFN-γ was also decreased in BM plasma cultures, as was TNF-α (Fig. 3C-D). Next, to probe a role for lipids in this suppressive activity of BM plasma, they were depleted from BM samples and compared with lipid-replete BM samples. Removal of BM lipids increased mitochondrial mass in control and MM BM CD8+ T cells (Fig. 3E), accompanied by restoration of IFN-γ and TNF-α expression (Fig. 3F-H). In CD4+ T cells however, mitochondrial mass remained unchanged, whilst IFN-γ expression was not restored (Supplementary Fig. 3C-D) indicating an alternative, lipid-independent mechanism of suppression, in line with the cohort data. Taken together, the data indicate that BM CD8+ T cells substantially take up lipids from the BM environment, leading to loss of mitochondrial mass and impaired functionality. This occurs in control BM, consistent with decreased functionality of BM vs. PB CD8+ T cells (Fig. 1), but may compound disease-driven loss of T cell mitochondrial mass (Fig. 2) to further compromise BM CD8+ T cell function in MM. To probe a role for lipid peroxidation-induced damage in this, BM mononuclear cell cultures were treated with ferrostatin. This increased IFN-γ secretion in BM but not PB plasma (Fig. 3I-K), confirming BM lipid peroxidation contributes to T cell suppression.
Since BM lipids suppress CD8+ T cell function, effective blockade of relevant transporters may restore this within the BM environment. The long-chain fatty acid transporter CD36 is implicated in mediating immune-suppressive and ferroptotic effects of lipids on NK and T cell populations(19, 23). We therefore assessed its expression in BM samples and found it to be more highly expressed by CD8+ than CD4+ T cells (Fig. 3L, Supplementary Fig. 3E), consistent with C16-BODIPY uptake data. However, analysis of cytokine expression revealed CD36-positive cells more highly expressed IFN-γ, TNF-α and IL-2 than CD36-negative counterparts, indicating activity of this transporter may not suppress BM T cell function (Fig. 7M, Supplementary Fig. 7F). Consistent with this, CD36 blockade did not increase CD8+ or CD4+ T cell IFN-γ expression in BM plasma (Fig. 7N, Supplementary Fig. 7G) and indeed decreased overall IFN-γ secreted (Fig. 7O). Therefore, to identify other relevant transporters expressed by BM CD8+ T cells, we explored single cell RNA-sequencing analysis of BM mononuclear cells from MM patients. Specifically, we assessed expression of the Fatty Acid Transport Proteins (FATP) 1 to 6 (SLC27A1-6). These mediate uptake of long-chain fatty acids and convert them into acyl-CoA esters, which retains them within the cell. This analysis identified that FATP1 and FATP5 are the most highly expressed family members in BM CD8+ T cells in MM, whilst transcripts for FATP2, 4 and 6 were scarce and for FATP3 undetectable (Fig. 3P). Analysis of control BM single cell RNA-sequencing dataset with cellular indexing of transcriptomes and epitopes by sequencing (CITE-Seq) data(33) revealed broadly similar expression patterns, with FATP1 and FATP5 being more highly expressed than FATP2, 4 or 6 (undetectable in this dataset) (Supplementary Fig. 3H-I). In contrast, FATP3 was more abundant in control than MM BM, yet significance of this is unclear, since FATP3 does not have confirmed lipid uptake activity(34). We therefore assessed whether blockade of FATP1 or FATP5 could restore function of CD8+ T cells from MM patients activated in presence of autologous BM or PB plasma. These experiments identified that specific blockade of FATP1, but not FATP5 increased T cell IFN-γ secretion (Fig. 3Q, Supplementary Fig. 3J-K). This did not occur in PB plasma, indicating a specific effect of BM long-chain fatty acids transported via FATP1 on CD8+ T cells (Fig. 7R-S). Therefore, BM long-chain fatty acids decrease CD8+ T cell mitochondrial mass and impair their effector function, which can be rescued by blockade of the lipid transporter FATP1.
Functional and metabolic features of CD8 + T cells are restored in MM patients in remission but not in relapsed patients
Finally, to understand T cell functional and metabolic phenotypes associated with poor vs. effective control of MM, we analysed samples from patients who had either relapsed on therapy or were in remission. In both groups, irrespective of treatment response, CD8+ T cell frequency was increased in BM and PB compared to MM patients at diagnosis (Fig. 4A, Supplementary Fig. 4A) and decreased CD4+ T cell frequency was observed (Fig. 4B, Supplementary Fig. 4B). Proportions of naïve, CM, EM and EMRA CD8+ and CD4+ T cell subsets were also similar between patient groups, in BM and PB (Fig. 4C-D, Supplementary Fig. 4C-D), as were those of immune-checkpoint expressing cells (Fig. 4E-F, Supplementary Fig. 4E-F). However, clear differences between the two groups emerged when interrogating T cell function and metabolism, particularly within BM CD8+ T cell populations. Proportions of these cells expressing IFN-γ, TNF-α, IL-2 and Granzyme B were all significantly increased in patients in remission compared to those who had relapsed (Fig. 4G), as well as CD4+ T cells expressing IL-2 (Fig. 4H). This increased functionality was less pronounced for PB CD8+ T cells (Supplementary Fig. 4G), albeit increases in PB CD4+ T cell function were observed (Supplementary Fig. 4H). Frequency of IFN-γ-, TNF-α, IL-2-expressing and multifunctional BM CD8+ T cells inversely correlated with malignant plasma cell abundance within aspirate samples from MM patients at diagnosis, upon relapse and in remission (Fig. 4I), supporting a role for BM CD8+ T cells controlling disease progression. This was also true for TNF-α, IL-2-expressing and multifunctional CD4+ T cells (Supplementary Fig. 4I). Of note, the profound restoration in BM CD8+ T cell function in patients in remission occurred alongside significantly decreased long-chain fatty acid uptake and increased mitochondrial mass, such that, in this group, BM and PB CD8+ T cell mitochondrial mass were now comparable (Fig. 6J-K). Whilst drivers of these metabolic and functional T cell changes cannot easily be determined from these cross-sectional samples from patients on diverse treatment regimens (Table S2), the data nevertheless highlight that effective MM control associates with specific changes in T cell metabolism, implying strategies to modulate this, such as FATP1 blockade, could be beneficial.