Granzyme B PET Imaging of Immune Checkpoint Inhibitor Combinations in Colon Cancer Phenotypes

Purpose Immune checkpoint inhibitor (ICI) monotherapy and combination regimens are being actively pursued as strategies to improve durable response rates in cancer patients. However, the biology surrounding combination therapies is not well understood and may increase the likelihood of immune-mediated adverse events. Accurate stratification of ICI response by non-invasive PET imaging may help ensure safe therapy management across a wide number of cancer phenotypes. Procedures We have assessed the ability of a fluorine-labelled peptide, [18F]AlF-mNOTA-GZP, targeting granzyme B, to stratify ICI response in two syngeneic models of colon cancer, CT26 and MC38. In vivo tumour uptake of [18F]AlF-mNOTA-GZP following ICI monotherapy, or in combination with PD-1 was characterised and correlated with changes in tumour-associated immune cell populations. Results [18F]AlF-mNOTA-GZP showed good predictive ability and correlated well with changes in tumour-associated T cells, especially CD8+ T cells; however, overall uptake and response to monotherapy or combination therapies was very different in the CT26 and MC38 tumours, likely due to the immunostimulatory environment imbued by the MSI-high phenotype in MC38 tumours. Conclusions [18F]AlF-mNOTA-GZP uptake correlates well with changes in CD8+ T cell populations and is able to stratify tumour response to a range of ICIs administered as monotherapies or in combination. However, tracer uptake can be significantly affected by preexisting phenotypic abnormalities potentially confusing data interpretation. Electronic supplementary material The online version of this article (10.1007/s11307-020-01519-3) contains supplementary material, which is available to authorized users.


Introduction
Immune checkpoint receptors are crucial molecules that regulate the immune system, dampening T cell activation. Tumours exploit these checkpoints to evade the immune system. Immune checkpoint inhibitors (ICIs) have changed the landscape of cancer treatment; however, responses vary depending on tumour type and only a minority of patients show a durable response [1][2][3][4]. Combination ICI therapy is being actively pursued as a strategy to improve the rate of response. PD-1 and CTLA-4, the two most commonly studied checkpoints, are not functionally redundant and while they both act to modulate distinct populations of tumour-infiltrating lymphocytes (TILs), including CD4+, CD8+, regulatory T (T reg ) cells, they do so with different mechanistic profiles. Thus, combination therapies may act in a complementary or synergistic fashion depending on the tumour microenvironment [5]. However, the biology surrounding combination therapies is not well understood and combined use may also increase the incidence of potentially severe side effects. Recent studies using a combined PD-1/ CTLA-4 strategy have shown increased immune-related side effects and unexpected off target effects including cardiac toxicity [6]. It is hoped that newer ICIs may improve efficacy when combined with established inhibitors such as PD-1, while reducing unwanted side effects. The negative immune checkpoints TIM-3 (T cell immunoglobulin and mucin domain-3) and LAG-3 (lymphocyte-associated gene 3) are co-expressed with PD-1 on T cells and are being actively assessed in the clinic [7][8][9]. TIM-3 is expressed on numerous immune cell types including T cells, dendritic cells, and macrophages and is thought to be involved in immune tolerance by mediating T cell exhaustion [8]. LAG-3 is also a checkpoint modulator, is also thought to be involved in immune tolerance by mediating T cell exhaustion [7] and has been shown to be expressed on expressed on natural killer (NK), activated T cells and T reg TILs [5,10]. Alternative strategies include targeting OX40, a T cell costimulator expressed on CD4 and CD8 T cells during antigen specific priming which leads to T cell expansion and differentiation. OX40 antibody agonists have shown significant therapeutic efficacy both as monotherapies and in combination with PD-1 blockade leading to significant tumour regression [11]. Certain phenotypes, such as microsatellite instability (MSI), mutational burden, and neoantigenic load have been demonstrated to enrich response to PD-1 monotherapy [2,12]. Currently, there is little evidence to determine whether these will also enrich response to the newer immune therapies and importantly no way to accurately predict individual patient response. Accurate stratification of treatment response to monotherapies and especially combination therapies, across all tumour phenotypes, is paramount to ensure safe and effective patient management.
Currently, there is a dearth of biomarkers capable of accurately stratifying ICI treatment response. Non-invasive imaging of the tumour microenvironment is being intensively researched to accurately assess immune therapies. Several studies have shown that TILs are necessary for checkpoint inhibitor efficacy; thus, radiopharmaceuticals have been developed to target T lymphocyte populations including CD3 and CD8. The presence of TILs alone, however, may not be enough to accurately stratify response due to immune cell regulation or tolerance [13][14][15]. Granzyme B is released upon activation of cytotoxic CD8+ T cells, providing information about T cell location and tumouricidal activity. A 68 Ga-labelled peptide that targets GZB has been shown to reliably stratify response to ICIs in preclinical tumour models [16][17][18]. In the current study, we have developed and assessed a new fluorine-labelled derivative, [ 18 F]AlF-mNOTA-GZP, to improve PET imaging sensitivity and spatial resolution [19]. The ability of [ 18 F]AlF-mNOTA-GZP to stratify immunotherapy response was assessed in response to ICI monotherapy and combined ICI therapies in CT26 and MC38 syngeneic colon cancer models, frequently associated with MSI-low phenotype, and MSI-high phenotype [20], respectively. PET imaging results were then correlated to differences in TIL subpopulations by flow cytometry.

Materials and Methods
General H-Asp(OtBu)-H NovaSyn TG resin (0.21 mmol/g) was obtained from Merck. Fmoc-amino acids, HATU and HOAt were obtained from Advanced Chemtech. Fmoc-glutamic acid was t-butyl protected. (p-SCN-Bn)-NOTA was purchased from Boc Sciences and Macrocyclics. Glacial acetic acid was purchased from JT Baker. Sep-Pak® light (46 mg) Accell™ plus QMA carbonate cartridges and Sep-Pak® C18 light cartridges were purchased from Waters Corporation. Saline solution (0.9 % w/v) was purchased from Braun Medical Industries. All other chemicals and reagents were purchased from Sigma-Aldrich, Fisher Scientific, and Tokyo Chemical Industry. No-carrier-added aqueous [ 18 F]fluoride ion was produced via the [ 18 O(p,n) 18 F] nuclear reaction (GE PETtrace 860 cyclotron). Quality control analytical radio-HPLC was performed on a UFLC Shimazdu HPLC system equipped with dual wavelength UV detector and a NaI/ PMT-radiodetector (Flow-Ram, LabLogic). Radioactivity measurements were made with a CRC-55tPET dose calibrator (Capintec, USA).
trapped on a Sep-Pak® light (46 mg) Accell™ plus QMA carbonate cartridge (pre-conditioned with 10 mL water) and washed with a further 5 mL of water. The trapped [ 18 F]fluoride anion was eluted with 0.9 % w/v saline (0.2 mL) and adjusted to pH 4 with glacial acetic acid (1 μL). To the pH-adjusted [ 18 F]fluoride solution was added 2 mM AlCl 3 in 0.1 M pH 4 NaOAc buffer (24 μL) and ethanol (0.2 mL). This solution was then added into a reaction vial containing mNOTA-GZP (0.1 mg). The reaction vial was sealed and heated at 100°C for 15 min without stirring. After cooling to room temperature, the crude reaction mixture was diluted with water (40 mL) and loaded on a Sep-Pak® C18 light cartridge (pre-conditioned with 5 mL ethanol and 10 mL water). The cartridge was washed with a further 5 mL of water.

PET-CT/MR Imaging
Animals were imaged 14 days after treatment initiation using a Siemens Inveon PET-CT and a Mediso nanoscan 3 T MR-PET. Briefly, animals were anaesthetised using inhalational isoflurane (maintained at 1.5 % alveolar concentration) and injected with [ 18 F]AlF-mNOTA-GZP (~10 MBq) via the lateral tail vein. Static PET acquisitions were performed at 60-80 min post-injection (p.i.) and CT and MRI scans were used for co-registration. Animals were monitored for maintenance of body temperature and respiration rate during imaging studies using the Biovet physiological monitoring system. Post-analysis of reconstructed calibrated images were performed with FIJI and Amide software (version 10.3 Sourceforge). Uptake of radioactivity in the tumour was determined by the placement of a volume of interest (VOI) delineated by CT and MR imaging. A VOI was also placed in the quadriceps muscle to provide reference tissue values. Data are expressed as % of the injected dose per gram (%ID/ g) of tumour tissue in the VOI.

Fluorescence-Assisted Cell Sorting (FACS) and Analysis
Tumours were excised after in vivo PET imaging on day 14 and immediately processed for flow cytometry. A single-cell suspension was generated by incubating in modified RPMI (Gibco) supplemented with 10 % heat-inactivated foetal bovine serum (Gibco, Life Technologies), 20 μg/ml DNAse1 (Sigma-Aldrich) and 200 μg/ml Collagenase (Sigma-Aldrich). The samples were mechanically diced and incubated for 1 h at 37°C and dissociated into single cells by passing through a 100 μm cell strainer. The samples were then counted and assessed for

Statistical Analysis
Data were analysed using a Kruskal-Wallis 1-way ANOVA with a Dunn's post-test using GraphPad Prism version 8.0.0 for Windows, GraphPad Software, San Diego, CA, USA, www.graphpad.com; P G 0.05 was considered statistically significant. Data are expressed as mean ± SD unless otherwise indicated.  [17], with an inhibition potency of K i7 8 ± 36 nM (Supplementary Materials, Fig. S1).

Assessment of Treatment Efficacy Using Tumour Growth Inhibition
Balb/c mice bearing CT26 and C57BL/6 mice bearing MC38 colon tumours were treated with either control IgG, an ICI monotherapy regime (either ⍺PD1, ⍺CTLA4, ⍺OX40, ⍺TIM3 or ⍺LAG3 alone) or an ICI combination regime (⍺CTLA4, ⍺OX40, ⍺TIM3 or ⍺LAG3 combined with ⍺PD1) as shown in Fig. 1a. Initial tumour volume assessment at day 5 after tumour cell inoculation showed similar tumour volumes~100 mm 3 for the two tumour types. However, the CT26 and MC38 colon tumours showed very different responses to ICI monotherapy or combined ICI therapies as determined by tumour growth inhibition, as shown in Figs. 1b and c and Supplemental Tables S2 and S3.

Assessment of Treatment Efficacy Using [ 18 F]AlF-mNOTA-GZP PET Imaging
[ 18 F]AlF-mNOTA-GZP PET imaging showed adequate tumour-to-background contrast for tumour visualisation in the two tumour models assessed. [ 18 F]AlF-mNOTA-GZP showed heterogeneous in vivo tumour tracer uptake across the different treatment arms in each of the tumour models (Fig. 2). Quantitative tumour VOI analysis from PET-CT/ MRI images at 60-80 min p.i. showed that [ 18 F]AlF-mNOTA-GZP is able to detect significant differences in tracer uptake between the different treatment arms and was able to differentiate responders from treated non-responders (TNR, an aggregate of non-responders across all treatment arms) or control treatment in the CT26-tumour bearing mice (Table 1, Fig. 3a, c).

[ 18 F]AlF-mNOTA-GZP Tumour Uptake Correlates with FACS-Defined Immune Profile
To quantitate the changes in CT26 TILs in response to therapy, we analysed the distribution of different immune cell markers ( Table 2, Fig. 4a-c). Significant differences were only observed when assessing T cell populations, including the total T cell population (marked by CD3+ relative to the CD45+ lymphocyte marker) and cell numbers associated with T cell effector function (namely with CD8+ T cells and GZB+ CD8+ T cells).

Discussion
Immune checkpoint inhibitor (ICI) monotherapy has been shown to trigger activation of compensatory T cellassociated checkpoints [5], such as PD-1 blockade upregulating TIM-3, an effect that could perhaps be mitigated by combined therapy. Emerging evidence supports the use of combined ICI regimens, although toxicity profiles, especially those associated with combined CTLA-4 blockade, may be of concern [3]. With the substantial number of clinical trials evaluating ICI combination therapies, the ability to rapidly and reliably assess therapy efficacy is paramount to ensure safe patient management. Non-invasive molecular imaging of immunotherapy has predominantly focused on the expression of immunologically relevant targets within the tumour microenvironment. The majority of studies have investigated the utility of measuring PD-L1 expression to predict therapy response and have shown that while they may have excellent utility for patient selection, PD-L1 expression is dynamic and context dependent and high PD-L1 expression does not guarantee therapy response [23,24]. Immunotherapy success is dependent on TILs and numerous biomarkers have been developed to assess changes in specific T cell subsets by the specific targeting of cell-surface CD molecule expression (CD3, CD8) [13][14][15]; however, high TIL expression does not guarantee therapy response either, due to tolerance and regulatory mechanisms [25,26]. Detection of molecules that represent an effector T cell phenotype has proven to be more successful. The radiolabelled GZB targeting peptide [ 68 Ga]Ga-mNOTA-GZP has been shown to be well correlated with immunotherapy efficacy [16][17][18]. In the current study, we have radiolabelled the GZB targeting peptide using [ 18 F]AlF to improve PET imaging duration of use, sensitivity and spatial resolution [19]. The addition of [ 18 F]AlF did not substantially change the enzymatic activity of the peptide (Supplemental Fig. S1), nor did it change the in vivo uptake and excretion kinetics compared to the 68 Galabelled peptide (Supplemental Fig. S2a and b). Furthermore, when compared to our previously published data on the 68 Ga-labelled peptide, very similar tracer uptake was observed in CT26 tumours in both ICI treatment responder and non-responder groups [18].
Interestingly, both radiopharmaceutical uptake and response to therapy was very different in the CT26 and MC38 tumours. These differences are likely due to the MSI phenotype found in MC38 tumours. MSI or deficiency of DNA mismatch repair system (MMR) occurs in approximately 15 % of colorectal cancers [20] and renders cancer cells unable to correct errors that occur at microsatellite regions during DNA replication. This leads to the accumulation of mutations which likely increase the amount and immunogenicity of neoantigens [12,[27][28][29]. Even with such an immunostimulatory environment, only~50 % of MSI colorectal cancers respond to classical ICIs clinically while MSI negative colorectal cancers generally do not respond [20]. The effects of this immunostimulatory environment can be clearly observed in the tumour responses to ICIs and subsequent [ 18 F]AlF-mNOTA-GZP uptake. Overall, the CT26 tumours showed a modest (~40 %) response to ⍺PD1 monotherapy and an improved (~70 %) response to ⍺CTLA4 or ⍺OX40 monotherapy or combined ⍺PD1 with ⍺CTLA4 or ⍺PD1 with ⍺OX40 therapy, as determined by tumour growth inhibition. The newer combination therapies ⍺TIM3 and ⍺LAG3, however, did not effectively inhibit tumour growth in the CT26 tumour bearing mice when used as monotherapies. Furthermore, neither ⍺TIM3 or ⍺LAG3, effectively inhibited tumour growth when used in combination with ⍺PD1, despite ⍺PD1 monotherapy proving efficacious in 40 % of tumours when used as a monotherapy. This lack of response may be due to the relatively low total number of CT26 resident cytotoxic T cells, despite tumour-infiltrating CD8+ cells having been shown to express TIM-3 and LAG-3 [19,30]. The low number of T cells associated with the CT26 tumours leads to a relatively low tumour uptake of [ 18 F]AlF-mNOTA-GZP at~0.2%ID/g in nonresponsive tumours, peaking at~0.6%ID/g in treatment responsive tumours. The MC38 tumours, with a much greater number of cytotoxic T cells, displayed a higher response rate to all therapies (~70-90 % response to ⍺PD1, ⍺CTLA4, ⍺OX40 or combinations and successfully responded to ⍺TIM3, ⍺LAG3 monotherapies or combinations (~40-60 %). [ 18 F]AlF-mNOTA-GZP uptake was similarly affected by the immunostimulatory environment, with significantly higher uptake in all the MC38 tumours (in controls, TRs and TNRs alike) ranging from~0.6%ID/g in non-responsive tumours (levels comparable to peak responders in the CT26 tumour type) and up to~1.1%ID/g in treatment-responsive tumours. In both tumour types, uptake of [ 18 F]AlF-mNOTA-GZP in TRs were positively correlated with CD3+, CD8+ and CD8+GZB+ TILs (Figs. 4 and 5), however, the variable levels of immune cells prevalent in the MSI high MC38 tumours, obscured the changes in T cell populations diminishing the correlation between TILs and [ 18 F]AlF-mNOTA-GZP uptake when compared to the MSIlow CT26 tumours. The difference in tumour immunostimulatory environments may be a limiting factor affecting the utility of [ 18 F]AlF-mNOTA-GZP, potentially obfuscating response stratification in immune-sensitive tumours.
Furthermore, while the number of non-responsive MC38 tumours was relatively low, those that did not respond displayed an intriguing reduction in CD8+ TILs compared with the control treated group, potentially linked with the regulatory suppression mechanisms that caused these tumours to be non-responsive. While a small increase in CD4+ FOXP3+ CD25+ Tregs was observed in the MC38 TNRs when compared to the control treated tumours this was not significant. Certainly, the presence of mutations can influence the function of several cellular processes and these may include those that are designed to nullify the host response.

Conclusions
In summary, the current study shows that in vivo molecular imaging of intra-tumoural granzyme B using [ 18 F]AlF-mNOTA-GZP is able to stratify tumour response to a range of ICIs administered to preclinical colon cancer models given as monotherapies or in combination. However, care should be taken in interpreting the images, as tracer uptake can be significantly affected by pre-existing phenotypic abnormalities. Further studies will be needed to elucidate other factors affecting the interpretation of granzyme B targeting tracers before they could be used for stratification of patient response across multiple tumour phenotypes.
Funding Information. This work was supported by Singapore's Health and Biomedical Sciences (HBMS) Industry Alignment Fund Pre-Positioning (IAF-PP) grants H18/01/a0/018 and H17/01/a0/010, administered by the Agency for Science, Technology and Research (A*STAR). The content is solely the responsibility of the authors and does not necessarily represent the official views of A*STAR.