Background
As a variety of immunotherapies progress toward clinical approval, it is becoming more important to identify biomarkers to assess the clinical activity of these drugs; both to begin to understand what immunobiological changes are induced, and to identify those patients who are likely to benefit from these potentially toxic and often costly treatments. The addition of chemotherapy to immunotherapy in combination treatments is under intense investigation, however there is limited understanding of how concurrent chemotherapy may affect putative biomarkers of immunotherapy, and how to analyse and interpret these in the context of the cyclical changes in immunological parameters induced by cytotoxic treatments. Here, we present further immune biomarker data from a recent chemoimmunotherapy clinical trial conducted in patients with mesothelioma, and discuss the complexity of interpreting this information in the context of prediction and prognosis.
CD40 is a member of the TNF receptor superfamily primarily expressed on antigen-presenting cells (APC), e.g. dendritic cells (DC), B cells and monocytes, but also found on some non-lymphoid cells such as epithelial and endothelial cells, fibroblasts, and some tumours [
1]. During the T cell-mediated immune response, CD40 ligand (CD40L; CD154) -expressing CD4 helper T cells can activate APC through CD40 signalling. These APC can in turn provide a ‘licence-to-kill’ signal to CD8 cytotoxic T cells - the main effectors in immune-mediated tumour regression [
2].
Extensive preclinical studies of anti-CD40 therapy have shown efficacy in various tumour model systems, and several clinical agents targeting the CD40 signalling axis have been or are currently under investigation [
3]. CP-870,893 is a fully human IgG CD40 agonist antibody that has shown promise as a single agent in patients with solid tumours, although overall response rates are still low [
4]. Although CP-870,893 infusion was predicted from preclinical studies to induce tumour-specific cell-mediated immune responses, this remains to be fully confirmed in the clinical setting.
Agonistic anti-CD40 can synergise with chemotherapy and cure advanced tumours in mice, especially when administered after chemotherapy [
5,
6]. This post-chemotherapy activity of agonistic anti-CD40 is hypothesised to occur by activating DC that have become ‘loaded’ with antigen from chemotherapy-induced tumour cell death, inducing expression of costimulatory molecules CD80 and CD86 and increased production of IL12 amongst other cytokines [
7]. CP-870,893 has been investigated in conjunction with chemotherapy in early phase clinical trials, mostly in patients with advanced, treatment-resistant disease [
7,
8]. In studies in pre-treated patients, around 20% of participants achieved objective tumour regression. In our recent first-line mesothelioma trial (a phase Ib dose escalation study in combination with chemotherapy), 40% of patients achieved a partial response [
9].
Currently identified pharmacodynamic effects of CP-870,893 as a monotherapy are most obvious in the B cell compartment, with depletion and activation of peripheral B-cells occurring within 72 h of infusion [
4]. Previous studies have also reported detectable modulation of DC by CP-870,893; in particular, depletion of CD11c
+CD123
dimCD14
− DC from peripheral blood, and in vitro increases in HLA-DR expression by monocyte-derived DC [
10,
11]. Weekly CP-870,893 monotherapy halved circulating lymphocyte concentrations after 48 h before returning to pre-treatment levels; this depletion was not observed when dosing occurred every three weeks [
4,
10]. Thus, the pharmacodynamic effects of CP-870,893 are still somewhat undefined – particularly when combined with chemotherapy. Specifically, prolonged immune modulation over the longer term of a full course of treatment has not been characterised.
We recently showed that B-cell depletion and activation also occurs over several cycles of combined CP-870,893 plus pemetrexed and cisplatin chemotherapy in patients with mesothelioma [
9]. Here, we present further in-depth flow cytometric analysis of patient peripheral blood mononuclear cells (PBMC) collected longitudinally throughout this study, in order to enhance our understanding of the immunobiology of combination chemoimmunotherapy and the unique challenges and considerations of analysis in this setting. We identify cyclical variations in PBMC subpopulations repeated with each cycle of chemo-immunotherapy, and identify potential relevant biomarkers of clinical activity in dendritic cells, CD8+ effector cells and regulatory T cells in response to anti-CD40 agonist treatment in the context of chemotherapy. We present statistical analysis techniques which may inform other investigators in serial immunomonitoring of chemoimmunotherapy.
Discussion
It is becoming increasingly important to identify immune biomarkers in patients treated with the variety of immunotherapies currently being developed; both to assist with clinical decision making and to help understand the immunological basis of response, or lack thereof. Given that activating anti-CD40 has strong preclinical and early clinical evidence of efficacy, we undertook this study of systemic immune parameters to establish the pattern of changes induced by this agent in the context of chemotherapy, and to undertake an exploratory analysis of any correlation between these changes and patient outcomes.
We recently published the results from a phase Ib clinical trial combining the anti-CD40 agonist antibody CP-870,893 with cisplatin/pemetrexed chemotherapy in patients with mesothelioma [
9]. We demonstrated the combination was well tolerated, and had at least equivalent efficacy to chemotherapy alone. In an earlier study, Vonderheide et al. documented short-term changes in various immunological parameters immediately following CP-870,893 infusion on B cells in particular, and in our original report we described similar changes in B cell markers over 6 treatment cycles [
4]. Here, we expand this longitudinal analysis of immune parameters to include data on dendritic cells and T cells, and show how, in the context of chemotherapy, CP-870,893 induces alterations in these cell populations.
Changes in B cells are one of the clearest immunopharmacodynamic indicators of CP-870,893 activity both in this and the previous study. This is not unexpected since B cells are the largest leukocyte subset that express the target of the drug, CD40. Notwithstanding this, the proposed mechanism of action of the drug is via agonistic activation of DC through the CD40 receptor [
24,
25]. However, it remains to be clarified which DC subsets may be involved in anti-CD40 mediated tumour immunity.
BDCA-1+ (CD1c+) DC are the most numerous myeloid DC population, maturing into both tissue and lymph node DC that are potent activators of CD4+ T cells. BDCA-2+ (CD303+) plasmacytoid DC are present in the blood at similar concentrations to BDCA-1+ DC and are primarily lymphoid DC precursors; these play a key role in responding to viral infection through type 1 interferon release. BDCA-3+ (CD141+) myeloid DC are a smaller population, typically making up around 10% of total blood DC, and have a major role in cross-presentation of antigen to CD8+ T cells. See refs [
26,
27], and for recent reviews on DC subtypes and nomenclature [
26,
27]. We report that the proportion of the two major DC subsets, BDCA-1+ (CD1c
+ myeloid) and BDCA-2+ (CD303
+ plasmacytoid), decreased sharply as a proportion of total PMBC in response to dexamethasone treatment as previously described, but it was not apparent that anti-CD40 treatment itself (i.e. between Day 8 and Day 15) induced any changes away from baseline in either proportional presence of any DC subsets at the time points chosen for this study [
14]. A previous study by Rϋter et al. reported that depletion from peripheral blood of CD11c
+CD123
dim CD14
− DC was observed in the shorter term, although data were not shown [
10].
It may be that direct binding to the CD40 molecule by CP-870,893 can modulate CD40 expression, but this was not observed - either because no such modulation occurred, or alternatively because it was simply not detected due to post-CP-870,893 blood collection being scheduled one week following drug administration, with any modulations occurring in the shorter term. Indeed, this was the case with data on B cell populations in previous work from this study [
9]. Although expression of the MHC-class II molecule HLA-DR had a downward trend over 6 cycles of chemoimmunotherapy in all three DC subsets examined, this was only statistically significant in BDCA-1+ DC (
p = 0.022) and we do not consider the magnitude of this decrease to be biologically relevant. A 2007 study by Hunter and colleagues reported that HLA-DR expression increased in response to CP-870,893 treatment of monocyte-derived DC in vitro over a 24 h time window. Similarly, whilst profound immediate and short term effects of CP-870,893 on B cells were identified, these parameters returned to baseline by day 8 after treatment [
4]. Therefore it is possible that any increase was transient and not detected by us 7 days after treatment [
11].
Weekly CP-870,893 monotherapy has been shown to reduce lymphocyte concentrations in peripheral blood to below 50% of baseline counts two days after treatment at MTD, followed by rebound to pretreatment levels, although this has not been observed for less frequent dosing schedules [
4]. Across the longer term, however, our combination of cisplatin/pemetrexed and anti-CD40 gives similar outcomes to anti-CD40 monotherapy, with some variation between patients but generally no overall change in lymphocyte concentration observed [
7,
10]. Ruter and colleagues also noted a bias toward depletion of CD4+ over CD8+ T cells and indeed our linear mixed modelling shows that over 6 cycles of chemotherapy this may be the case – particularly at Day 1 of each cycle, directly prior to administration of chemotherapy [
10]. However our observation is also small in scale, and we are unable to say definitively whether anti-CD40 may be responsible. A high intrapatient variability in the proportional presence of Tregs was observed, but was generally modulated consistently within individual patients in response to treatment, no long term changes appeared to be uniformly induced with this 6-cycle treatment schedule in agreement with past studies [
7,
10].
Our exploratory analyses of the relationship between immunological parameters and patient outcomes highlighted a potential relationship between either higher-than-median BDCA-2 + DC% of PBMC at baseline, or lower-than-median BDCA-3 + % at C2D8 and C3D8, and longer patient survival. Mechanistically, it is possible that a higher proportion of BDCA-2+ DC at baseline indicates a larger number of target cells available for the action of anti-CD40 therapy, potentially giving a useful prognostic indicator for those patients likely to achieve better responses. The seemingly beneficial link between a reduction in antigen cross-presenting BDCA-3+ DC and better OS, however, might seem contrary to expectation if an anti-tumour immune response is thought to be underlying the difference in survival. However, it may be that these BDCA-3+ are migrating out of the peripheral blood and into other tissues, potentially as a result of anti-CD40 inducing the maturation of these cells, and may therefore be indicative of those patients achieving better responses. Exploratory analyses in the T cell compartment revealed a correlation between OS and the ICOS + % of CD8+ T cells at baseline. The postulated mechanism of CP-870,893 action is to increase APC-mediated CD8+ T cell activation and it may be that those patients with higher-than-median CD8 activation prior to treatment have an immune landscape that is already predisposed toward this outcome; thus, addition of anti-CD40 is sufficient to tip the balance further in favour of a better anti-tumour response in those individuals. It is interesting to note that both parameters showing differences in OS at baseline, BDCA2+ DC and ICOS+ CD8 T cells, correlated in this sample set and may be worth further investigation. We acknowledge these data were obtained from a small number of patients and serve only as hypothesis-generating, however they may inform investigators in future clinical trials that include greater numbers of participants.
The interpretation of our results is complicated because each of the three administered agents have potential immune-modulating effects, and each was given at different stages of the 21-day treatment cycle. Nevertheless, this complexity is an emerging reality in human cancer immunotherapy, where single agents have promising but insufficient activity and where combinations are now under intense investigation. Firstly, characteristics of samples taken at Day 1 of each cycle (taken immediately before infusion of pemetrexed/cisplatin) were influenced by three doses of the glucocorticoid steroid dexamethasone, with a cumulative dose of 12 mg given in the 24 h leading up to chemotherapy. Dexamethasone is given prior to and for three to five days after pemetrexed/cisplatin for control of skin rash, emesis and inflammatory side effects, and has been shown to cause alterations in the majority of immune parameters examined in this paper; for example lymphodepletion and alteration of DC subsets [
14]. Secondly, the effects of chemotherapy itself, given on Day 1 of each treatment cycle here and evident one week later at Day 8. The effects of cisplatin/pemetrexed are well known, particularly with respect to depletion of proliferating and activated cells as can be observed from our lymphocyte data, and may well alter the immunological background of patients in a manner that can affect the subsequent dose of anti-CD40. We have previously demonstrated a profound reduction in the proportion of Ki67+ CD8+, CD4+ and Treg as well as activated effector CD8+ T cells on day 8 after chemotherapy, which recovers to baseline prior to the next chemotherapy cycle [
18]. Thirdly, the effects of the agonistic anti-CD40 antibody, administered on Day 8. As described above, pharmacodynamics of CP-870,893 with respect to parameters of the adaptive immune system are generally observed over a shorter time window (3–4 days) [
4,
10]. It is therefore probable that the majority of direct effects of anti-CD40 have reset by the time the next blood sample is collected at Day 15, hence were not observed as part of the inter-cycle variation in our study.
An important message from this study is that investigators must undertake careful preliminary pharmacodynamic immunological studies when selecting time points for immunological biomarker investigation in patients who are undergoing concurrent chemoimmunotherapy. Although longitudinal samples from patients provide the opportunity to identify predictive biomarkers for response to combination therapy, it is more complicated to attribute changes in defined immunological parameters to individual drugs within this combination without the inclusion of control arms to the study. Ideally, control arms would monitor immunological parameters in individual agents, however this is not always practical or feasible with small signal-finding studies and phase Ib studies. Consistency in timing of immunological sampling is paramount, with the potential for alteration in immunological parameters by supportive care medications and chemotherapy in addition to immunotherapies. A pre-corticosteroid baseline is also essential; some chemotherapy regimens have an absolute requirement for corticosteroid premedication which cannot be modified. Furthermore, in many clinical trials there is an allowable window for blood sampling which may include +/− one to two days which may risk adding to the complexity of result interpretation. Statistical techniques such as the linear mixed modelling used in this study can enhance the clarity of interpretation of complex, cyclical data and facilitate an understanding of change over time. However, only the availability of a control group receiving chemotherapy and supportive care alone will allow the contributions of immunotherapy to biomarker changes to be definitively evaluated.
Acknowledgements
The authors acknowledge the facilities, and the scientific and technical assistance of the Australian Microscopy & Microanalysis Research Facility at the Centre for Microscopy, Characterisation & Analysis, the University of Western Australia, a facility funded by the University, State and Commonwealth Governments.