Monitoring CD8a⁺ T Cell Responses to Radiotherapy and CTLA-4 Blockade Using [⁶⁴Cu]NOTA-CD8a PET Imaging

The remarkable, yet variable, responses to cancer immunotherapy highlight the pressing need for biomarkers that can assist in patient selection and evaluation of treatment response. Alongside the diverse efficacies reported among patients and different tumor types comes the potential development of severe autoimmune-like adverse events that could be spared if appropriate treatment was selected at an early stage [1‐3].

The clinical need is not limited to suitable, predictive biomarkers, but a strong urgency also exists for reproducible and standardized methods to systematically monitor the expression and dynamics of these [4]. Routine clinical biomarkers should preferably be assessed in a minimally invasive way, and ideally, longitudinally. Standard Response Evaluation Criteria in Solid Tumors (RECIST) are considered inadequate in discriminating responders from non-responders in patients receiving immune checkpoint blockade. This is due to atypical response patterns such as the associated inflammatory response that can cause an increase in tumor volume, i.e., pseudo-progression [5‐7]. To account for this, immune-related response assessment systems (irRC, irRECIST, iRECIST) for evaluation of response to immunotherapy have been implemented [8]. However, these modified criteria are merely related to tumor size and limited for early assessment of treatment response [5, 9]. While providing useful information regarding metabolic status, positron emission tomography (PET) Response Criteria in Solid Tumors (PERCIST) are correspondingly insufficient to identify patients with a favorable response to immunotherapy due to the inflammatory flare causing increased uptake by the metabolically active immune cells, i.e., also a pseudo-progression [10]. Consequently, much research has been committed to develop molecular imaging techniques to evaluate and monitor immune status to spare patients from expensive and ineffective treatments with potential severe side effects.

CD8⁺ cytotoxic lymphocytes play an essential role in the anti-tumor immune response and their activation and infiltration in tumors are considered feasible candidates for assessment of immunotherapeutic response. This is owing to the fact that clinical responses to PD-1/PD-L1 immune checkpoint inhibitors arise most often in patients with high preexisting numbers and/or infiltration of CD8⁺ cells during therapy [11‐14]. Molecular imaging probes targeting CD8⁺ have indeed been pursued in preclinical studies primarily utilizing engineered antibody fragments [15‐18] and successfully demonstrated the value of CD8⁺ PET imaging for evaluation of immunotherapeutic response [16‐18].

In the current study, we developed a Cu-64 labeled (t_1/2 = 12.7 h) F(ab)′2 fragment targeting murine CD8a⁺ for non-invasive in vivo detection and quantification CD8⁺ cytotoxic lymphocytes. We sought to investigate the utility of [⁶⁴Cu]NOTA-CD8a as a prognostic imaging biomarker of therapeutic response in an immunocompetent mouse model of colon adenocarcinoma. We used a combination therapy approach as these are currently widely investigated clinically due to the inherent differential response rates observed with monotherapy immune checkpoint inhibition [19‐21]. Radiotherapy combined with immune checkpoint inhibition of CTLA-4 effectively inhibited tumor growth, increased the overall survival of CT26 tumor-bearing mice, and [⁶⁴Cu]NOTA-CD8a PET identified the responding mice before tumor growth inhibition was evident.

The molecular and cellular features associated with a successful immune response to a tumor are continuously investigated and mapped to identify key denominators of response to immunotherapy. One of the major challenges is the effective stratification of responders from non-responders in the clinical setting. While immune monitoring of patients with serial biopsies or peripheral blood samples is extremely informative, the location of sampling, i.e., tumor heterogeneity or organ-dependent differences, timing of sampling, and interpretation of data before and after therapy are all potential contributors to biological variation that complicate the examination of data [22]. Therefore, we developed a specific radiotracer to be used for non-invasive detection, tracking, and quantification of CD8a⁺ lymphocytes in mice with the possibility of repeated imaging and spatial mapping over time.

F(ab)′2 fragments of a CD8a⁺ antibody were successfully produced and concurrently purified by preparative HPLC. Importantly, there were no full length CD8a⁺ antibodies or Fc fragments in the final product as confirmed by SDS-PAGE. Intact monoclonal antibodies targeting lymphocyte surface markers can have substantial immunomodulatory effects and can deplete the target cell population in vivo limiting their utility as imaging agents [23, 24]. The purified CD8a⁺-F(ab)′2 fragments were radiolabeled with Cu-64 with a high radiochemical purity. Many immuno-PET tracers have been coupled to Zirconium-89 (Zr-89, t_1/2 = 78.4 h) due to the relatively long biological half-lives of antibodies. The use of Cu-64, however, reduces radiation exposure and allows for multiple scanning sessions to be conducted in the same animal. Furthermore, the shorter positron range of Cu-64 compared with Zr-89 (0.7 vs. 1.18 mm) enables more exact quantification of uptake in smaller organs such as lymph nodes.

Our in vivo studies demonstrated that [⁶⁴Cu]NOTA-CD8a targeted lymphoid tissue, i.e., the spleen, lymph node, and the thymus of immunocompetent mice. The optimal dose of CD8a⁺-F(ab)′2 for tumor visualization has previously been established in immunocompetent mice by our group [25]. The optimal dose was applied in this study to block endogenous CD8a⁺ levels and thereby increase the amount of tracer circulating in the blood. In vivo biodistribution identified the optimal imaging timepoint as 24 h after injection of tracer, where the mean [⁶⁴Cu]NOTA-CD8a tumor uptake peaked. Although an increase in image contrast was observed from 24 to 48 h p.i. (Table 1), the fast tumor growth of the CT26 model was the rationale for selecting 24 h for evaluation of treatment efficacy. This was based on the fact that the magnitude of non-specific uptake of antibodies increases with tumor size due to the enhanced permeability and retention of macromolecules [26, 27]. Thus, the potential differences in tumor volume between different treatment groups could possibly be more pronounced 48 h as opposed to 24 h p.i. thus introducing a bias in interpretation of the imaging data. Moreover, tumors that are prominently different in size does not eliminate the possibility that the observed differences in CD8a⁺ infiltrate could be a consequence of tumor growth rather than a response to treatment. Importantly, tumor volumes were not significantly different at the day of PET imaging in this study. At 24 h p.i., approximately 2 %ID/g tracer remained in the circulation. Therefore, target uptake was corrected for background levels by normalizing to the heart content of tracer. In this way, tumoral and splenic uptake was corrected for potential treatment-induced changes in perfusion thereby reflecting residual CD8a⁺ numbers.

Other CD8a⁺-targeting antibody-derived probes have shown similar tumor uptake in syngeneic mouse tumor models. The [⁸⁹Zr]malDFO-169 cDb (~ 55 kDa) [16] presented with a PET-derived tumor-to-blood ratio of ~ 2 at 24 h p.i. in CT26 tumor-bearing mice, which is in good agreement with the tumor-to-heart ratio of 1.89 ± 0.28 obtained in this study. In contrast, the [⁸⁹Zr] PEGylated VHH-X118 camelid (~ 15 kDa) [18] exhibited a tumor-to-muscle ratio of ~ 10–20 at 24 h p.i. in B16F10 and Panc02 tumors. Typically, lower molecular weight corresponds to lower bioavailability and thereby amount of tracer available for tumor accumulation. Contrary, improved tumor penetration and the faster clearance from the blood stream increases image contrast at earlier timepoints. This, together with the fact that the authors utilized different tumor models likely explains the superior tumor-to-muscle contrast of [⁸⁹Zr] PEGylated VHH-X118.

Combining conventional anti-cancer therapies with immune checkpoint inhibitors is currently investigated in numerous clinical trials to improve patient benefit [28]. Especially, the synergy of immunogenic cell death induced by irradiation leading to local and systemic immune responses in combination with immunotherapy has gained much attention [29]. Both the local and systemic CD8a⁺ infiltration was investigated with [⁶⁴Cu]NOTA-CD8a in this study, i.e., splenic and tumoral uptake (Fig. 2c, d). In tumors treated with radiotherapy in combination with anti-CTLA-4 therapy, a significantly increased tumor-to-heart ratio was observed. We found no effect of XRT or XRT + anti-CTLA-4 on the [⁶⁴Cu]NOTA-CD8a spleen-to-heart or spleen-to-muscle ratio suggesting that our probe was not able to detect splenic changes or that our dosing regimen was not sufficient in producing systemic effects of focal radiation. Sparse preclinical data exist on the abscopal effect of radiation therapy on splenic CD8⁺ levels in combination with immune checkpoint inhibitors. Increased CD8⁺ numbers in the spleen of mice have previously been demonstrated on day 8 following a fractionated 3 × 9.18 Gy dose in combination with anti-PD-1 [30]. The exact T cell infiltration kinetics indeed depend on murine tumor model and XRT dose schedule [31] and it is plausible that our dosing regimen or the timing of PET imaging was not optimal to detect systemic changes in CD8a⁺ numbers. This notion is supported by the flow cytometric ex vivo analysis, which did not show differences in splenic CD45⁺CD8a⁺ numbers among treatment groups on day 8 following therapy initiation (Fig. 4b). Further, the increase in the [⁶⁴Cu]NOTA-CD8a tumor-to-heart ratio in tumors of mice treated with XRT and anti-CTLA-4 matched the increased number of CD45⁺CD8a⁺ cells detected by flow cytometry in this treatment group. Overall, the ex vivo evaluation of tissue confirmed the impact of our efficacy protocol and the specificity as well as accuracy of our CD8a⁺ PET imaging probe for monitoring treatment-induced responses in CD8a⁺ dynamics in this particular setup.

In tumors treated with combined XRT and immune checkpoint inhibition of CTLA-4, we identified TRs and TNRs. There was no pronounced effect of treatment on tumor volumes between groups at the time of imaging. Interestingly, TRs presented with an increased [⁶⁴Cu]NOTA-CD8a tumor-to-heart ratio compared with TNRs. This indicates that the utility of [⁶⁴Cu]NOTA-CD8a could extend beyond monitoring and might be a useful tool for predicting response to therapy. However, the arbitrary cutoff value (median [⁶⁴Cu]NOTA-CD8a tumor-to-heart ratio of XRT + anti-CTLA-4-treated mice) used for stratification was not an indicator of overall survival.

The utility of CD8⁺ as a predictive and prognostic biomarker of response has been lively debated. The main question arising is that the presence of CD8⁺ cells in the tumor microenvironment is not a prerequisite for response to immunotherapy, as the CD8⁺ cells could be anergic and/or exhausted and not capable of promoting tumor-mediated killing. Indeed, the activation state and clonality of T cells have been proposed as critical determinants of treatment response to immunotherapy [32, 33]. To address this issue, PET tracers targeting activation markers of T cells or soluble proteins such as cytokines are increasingly being developed [23, 34, 35]. Although providing significant information, this approach might be limited since activation markers can be short-lived following an immunotherapeutic treatment regimen necessitating a well-timed scanning schedule for accurate prediction or multiple scanning sessions—limiting its clinical utility. In addition, targeting secreted markers such as cytokines might face issues with solubility and redistribution within the tissue.

Although our data suggest that [⁶⁴Cu]NOTA-CD8a could be a suitable in vivo biomarker for response prediction and monitoring of immunotherapy protocols, additional experiments are warranted including different combination strategies, tumor models as well as extending to orthotopic models, as the immune infiltrate and immunotherapeutic efficacy are known to differ with implantation site [36]. Also, it would be interesting to apply the [⁶⁴Cu]NOTA-CD8a criterion dictating response in this study to a new cohort of mice to confirm the accuracy and reproducibility in determining response.

In the present study, we developed a Cu-64 labeled antibody-based PET tracer for the non-invasive detection and quantification of murine CD8a⁺ cells. We show that [⁶⁴Cu]NOTA-CD8a can be used to detect and quantify CD8a⁺ cells and monitor response to a common immunotherapy combination protocol. Further, our data suggest that [⁶⁴Cu]NOTA-CD8a PET may serve as a predictive imaging biomarker of response. Future work including additional preclinical examination across multiple cancer types and treatment regimens will elucidate whether [⁶⁴Cu]NOTA-CD8a is feasible for clinical translation.