Intratumoral T cells have a differential impact on FDG-PET parameters in follicular lymphoma

The identification of patients with follicular lymphoma (FL) that is destined to respond poorly to frontline therapy remains elusive.¹ To this end, quantifiable pretherapy ¹⁸F-fluorodeoxyglucose-positron emission tomography-computed tomography (FDG-PET) imaging parameters have been evaluated to identify such high-risk patients. The parameters include total metabolic tumor volume (TMTV), which gives insight into the burden of the tumor by 3-dimensional quantification encompassing all involved sites, and maximum standardized uptake value (SUV_max), a standardized measure of the highest FDG-uptake within a single lesion. Pooled analyses of predominantly cyclophosphamide, doxorubicin, vincristine, and prednisone+ rituximab (R-CHOP)–treated patients (of whom only a subset received R maintenance) found that increased pretherapy TMTV is a strong predictor of an adverse outcome in FL.^2,3  Interestingly, in the same group of treated patients, both a low and high SUV_max have been associated with a reduced progression-free survival (PFS).^4,5  However, FDG-PET analysis of the phase 3 GALLIUM study (58% treated by bendamustine induction), in which all patients received anti-CD20 antibody maintenance, found no association with either TMTV or SUV_max.⁶ Intriguingly, high pretherapy SUV_max was associated with shorter PFS in nonanthracycline-treated (mainly bendamustine or lenalidomide) but not in anthracycline (R-CHOP)–treated patients in another study.⁷ 

The predictive utility of a pretherapy T-effector gene signature in patients treated with CHOP/CVP (cyclophosphamide, vincristine, prednisone)-rituximab/obinutuzumab is reversed in patients treated with bendamustine-rituximab (B-R)/obinutuzumab.⁸ Unlike R-CHOP, B-R causes profound and sustained T-cell depletion.⁹ Rituximab maintenance after B-R induction may prolong CD4⁺ T-cell lymphopenia,¹⁰ involving reduced production of interleukin-17 and -15, which are critical for survival of CD4⁺ T-cell memory.¹¹ The extent to which clonally expanded T cells infiltrate the tumor microenvironment (TME) in FL is also prognostic.¹² Notably, the TME is distinct from that of other non-Hodgkin and Hodgkin lymphomas, with a high population of CD4⁺ T-follicular helper cells (T_FH).¹³ Although the FDG-tracer detects cellular glucose-uptake by all viable cells within the tumor site, to our knowledge, the role of intratumoral T cells on pretherapy FDG-PET parameters in FL remains undefined. Potentially, differences in the prognostic impact between SUV_max and TMTV in FL across studies, may in part reflect the differential depletion of glucose-avid intratumoral T cells after exposure to different chemoimmunotherapeutic regimens.

FDG-PET is integral to FL management. In this study, we assessed the relationship of TMTV and SUV_max with intratumoral T cells in FL. The findings have implications for the improved interpretation of FDG-PET metrics in FL, which remains an unmet need in this disease.

The study was approved by the institutional regulatory board (Metro South Human Research Ethics Committee) and was conducted in concordance with the Declaration of Helsinki. Eighty-three patients with histologically confirmed de novo grade 1 to 3a FL were included (characteristics in Table 1). The consort diagram (supplemental Figure 1) outlines sample testing, which was based solely on availability.

Pretherapy FDG-PET scans were assessed for SUV_max, TMTV, and total lesion glycolysis (TLG), using customized software (MIM Software, Cleveland, OH).^6,14 ,15  Biopsy samples (formalin-fixed, paraffin-embedded tissue and tumor-infiltrating lymphocytes [TILs] from deaggregated nodes) were used for multiplex gene hybridization, immunohistochemistry, T-cell receptor (TCR) repertoire sequencing and flow cytometry as published.^12,16 -19  Cryopreserved TILs were used to assess cellular glucose uptake by intratumoral CD4⁺ T cells, CD8⁺ T cells, and CD19⁺ B cells, by using the fluorescently labeled glucose analogue 2-NBDG uptake assay (Abcam).²⁰ The supplemental Methods provides a detailed description.

Multiplex gene hybridization was used to quantify intratumoral T-cell infiltration in 45 patients with FL and compared with TMTV. T-cell infiltration was calculated by a standardized CD4 and CD8A gene z-score, with T-cell^LO defined as quartiles 1 to 3 (n = 34), and T-cell^RICH as quartile 4. Patients with a T-cell^LO state had a significantly higher TMTV than T-cell^RICH (approximately fivefold, using the 41% SUV_max, and approximately sixfold using SUV ≥2.5; Figure 1A-B). TLG was also significantly higher in T-cell^LO states (approximately sevenfold; Figure 1C). Next, proportions of clonal (light-chain restricted) B cells and CD4⁺ and CD8⁺ T cells were enumerated by flow cytometry in 54 fresh nodal tissues. Those with greater than the median cutoff of FL B cells within tumor biopsy specimens (FL B cells^RICH) had approximately twofold higher TMTV than those with FL B cells^LO (Figure 1D). High Ki67 was also associated with increased TMTV (Figure 1E). However, TMTV was not associated with histological grade (1 and 2 vs 3A) or serum lactate dehydrogenase (supplemental Figure 2). Put together, the results indicate that TMTV reflects the burden of malignant B cells within FL nodes, which explains its inverse association with T-cell infiltration.

T cells proliferate in response to antigenic stimulation into clonally expanded populations.²¹ Therefore, glucose metabolism may differ because of the degree of clonal expansions within T-cell compartments, reflected in differential TMTV. There are currently no data on the relationship between T-cell clonality in specific T-cell subsets and TMTV in FL. In 21 FL TILs, sorted into CD8⁺ T-cell (Figure 1F), T_FH cell, regulatory T-cell (T_REG), and non-T_FH/T_REG CD4⁺ T-cell (Figure 1G) subsets, TCR sequencing showed that large T-cell clones predominantly resided in the CD8⁺ T-cell population (mean clonality index [CIx], 0.067; ∼33% of all T cells). T-cell clonality was intermediate in T_FH cells (CIx, 0.029; ∼30% of all T cells), low in T_REG (CIx, 0.022; ∼7%), and lowest in non-T_FH/T_REG cells (CIx, 0.013; ∼31%). Importantly, there was modest enrichment of expanded T_FH and CD8⁺ T-cell clones in low TMTV states (Figure 1H-I). Further interrogation of the intratumoral CD8⁺ T-cell population suggested that the clonal expansions were primarily driven by activated, rather than exhausted or resting, CD8⁺ T-cell subsets (Figure 1H). These findings require validation in larger, independent cohorts.

To evaluate the relative contribution of intratumoral B and T cells to TMTV, 2-NBDG was used to determined cellular glucose uptake in TILs (Figure 2A). The CD8⁺ and CD4⁺ T cells had significantly higher glucose uptake than did the CD19⁺ B cells (Figure 2B). However, given that CD19⁺ B cells contribute approximately two-thirds of the total lymphocyte population, it was the B cells that contributed the most to overall glucose uptake (Figure 2C). This further supports our initial finding that TMTV is reflective of the burden of the malignant population. Notably, cellular metabolism of TILs may be altered by cryopreservation and subsequent thawing. Experiments in fresh TILs are needed to confirm our findings.

Finally, we assessed SUV_max. TMTV is a volumetric assessment, whereas SUV_max is a measure of highest FDG avidity. Therefore, the contribution of the TME to SUV_max may differ from that of TMTV. To accurately assess the difference, we identified a subset of patients with FDG-PET imaging before their diagnostic FL biopsy, from which we obtained SUV_lesional (ie, SUV_max within the subsequently excised node) values. After excisional biopsy, we assessed immune gene infiltration by multiplex gene hybridization and cellular glucose uptake and compared the results with the paired SUV_lesional value. Those with high SUV_lesional values had significantly higher CD4 and CD8A, but not CD19 gene expression (Figure 2D). Unsurprisingly (as macrophages make up only a small percentage of the TME),¹² there was no association with CD68. There was a trend toward a positive correlation between CD4⁺, but not CD8⁺ or CD19⁺, cellular glucose uptake and SUV_lesional (supplemental Table 1). These results demonstrate that intratumoral T cells play an important role in SUV_max of individual FL lesions. Intratumoral T-cell spatial heterogeneity, with T_FH cells particularly pronounced, is an established feature of FL^22,23  and may influence SUV variability between lesions to contribute to the inconsistent prognostic efficacy of SUV_max. It may also influence the potential inability of pretherapy SUV_max to predict subsequent histological transformation.²⁴ 

In conclusion, our study suggests that intratumoral T cells differentially influence pretherapy FDG-PET parameters and can affect the predictive utility of FDG-PET in patients with FL who are treated with different chemoimmunotherapy regimens that vary in the extent of T-cell depletion that they induce. How chemotherapy-free immunomodulatory and/or checkpoint blockade therapies influence glucose-uptake in intratumoral T cells to affect interim and end-of-therapy FDG-PET parameters in FL is another avenue to be explored.

The visual abstract was created with Biorender.com. No writing assistance was used in the production of the manuscript.

This work was supported by a Haematology Society of Australia and New Zealand and Leukaemia Foundation scholarship (K.N.); the Leukaemia Foundation, the Mater Foundation, and the National Health and Medical Research Council (M.K.G.); and the Translational Research Institute FACSymphony A5 Reagent Support Fund which supported panel development. The Translational Research Institute is supported by the Australian Government.

Contribution: K.N., S.-C.L., and P.L. contributed to the study design, performed the experiments and collected and analyzed the data; M.B.S., J.G., L.M.d.L., D.S., M.S., H.T., J.W.D.T., S.-J.H., A.H., and D.C. performed experiments and contributed to data interpretation; S.J., R.J.B., and D.T. provided patient samples, collected data and contributed to data interpretation; C.K. and J.T. contributed to data analysis and interpretation; M.K.G. designed the project and contributed to patient samples and data analysis and interpretation; and all authors were involved in writing the manuscript.

Conflict-of-interest disclosure: D.T. has received research funding, honoraria and has held an advisory board membership from Amgen; honoraria and speakers bureau and advisory board membership for Janssen; honoraria for Novartis; honoraria, speakers’ bureau, and advisory board membership for Roche; and honoraria for Takeda. M.K.G. has received honoraria and travel support from Roche and honoraria from Janssen, Merck, Amgen, and Gilead. The remaining authors declare no competing financial interests.

Correspondence: Karthik Nath, Level 7, Translational Research Institute, Brisbane, QLD 4102, Australia; e-mail: karthik.nath@uqconnect.edu.au; and Maher K. Gandhi, Level 7, Translational Research Institute, Brisbane, QLD 4102, Australia; e-mail: maher.gandhi@mater.uq.edu.au.