Peripheral CD45RO, PD-1, and TLR4 expression in metastatic colorectal cancer patients treated with bevacizumab, fluorouracil, and irinotecan (FOLFIRI-B)
- First Online:
- Cite this article as:
- Formica, V., Cereda, V., di Bari, M. et al. Med Oncol (2013) 30: 743. doi:10.1007/s12032-013-0743-0
- 596 Views
CD45RO, PD-1, and TLR4 immune pathways have proven pivotal in regulating antitumor response and correlate with survival for localized colorectal cancer (CRC). We evaluated if their peripheral expression was associated with outcome in metastatic CRC (mCRC). Thirty-one mCRC patients were eligible for this prospective study (clinicaltrial.gov NCT01533740) and treated with first-line FOLFIRI-B. Blood was drawn before the first and third cycle and analyzed by flow cytometry for frequency (%) of CD4+, CD8+, CD45RO+, and PD1+ mononuclear cells and for TLR4 expression on neutrophils. Two cycles of chemotherapy determined changes in immune variables that were prognostically meaningful. Pre-third-cycle (ptc) CD45RO+CD8+cell% displayed a statistically significant association with progression-free survival (PFS) (median PFS 22.4 vs. 9.4 months for patients with CD45RO+CD8+cell%> vs. <the median value of 12 %, respectively, p 0.02) and overall survival (OS) (2-year OS rate 62 vs. 44 %, respectively, p 0.04). Surprisingly, ptc-PD1 overexpression was also associated with improved PFS of borderline statistical significance (HR 0.42, p 0.06). A Cox regression multivariate analysis for PFS including ptc-CD45RO+CD8+cell%, ptc-PD1+cell%, CEA, LDH, and Köhne risk class demonstrated CD45RO+CD8+cell% to be the only independent prognostic factor (HR 0.23, p 0.04). TLR4 and CD4 were not associated with the outcome. Peripheral CD8+CD45RO+ cells were confirmed to be of independent prognostic value in mCRC patients. Overexpression of the PD-1 immunosuppressor after two cycles of therapy may be a negative feedback mechanism, and therefore, an indirect sign of chemotherapy induced antitumor immune response with a favorable association with outcome. Enhancement of CD8+CD45RO+ cell response may be a fascinating therapeutic target to improve the efficacy of FOLFIRI-B.
Host immune and inflammatory responses may substantially contribute to patient outcome, as demonstrated by the improved prognosis of patients with high density of tumor-infiltrating lymphocytes (TIL), in particular memory T cells, in primary colorectal cancer lesions . High density of TILs in metastatic lesions is also predictive of higher radiological response to standard chemotherapy .
Human naïve and memory T cells can be identified by the reciprocal expression of the CD45RA or CD45RO isoforms . An analysis by Pagès et al.  of 415 colorectal tumors has demonstrated that high density of infiltrating CD45RO+ T cells correlated with a good clinical outcome. In another dataset by Mlecnik et al. , degree of cytotoxic CD8-positive and memory CD45RO-positive T cell infiltration (the so-called immune score) of the primary tumor had a discriminatory prognostic power superior to that of standard staging system with patients with high ‘immune score’ having significantly prolonged disease-free and overall survival.
Programmed death-1 (PD1, CD279) is an inhibitory co-receptor expressed on antigen-activated and exhausted T and B cells . There are two known ligands for PD1, B7-H1/PD-L1, and B7-DC/PD-L2. Several lines of evidence suggest that PD1 activation negatively regulates immune response. In murine tumor models, B7-H1 expression on cancer cells confers immune resistance and interrupting the PD1/B7-H1 interaction has clear antitumor effect [13–15]. In a phase I clinical trial of the fully human anti-PD1 monoclonal antibody MDX-1106, a durable complete response was observed in a patient with metastatic colorectal cancer (mCRC) .
TLR4 provides a crucial link between the recognition of potential biological harms, such as microbes or cancer cells, and the initiation of the host defense. Activation of TLR4 results in the release of antimicrobial peptides, inflammatory cytokines, and co-stimulatory molecules that modulate innate immunity . Stimulation of TLR4 on dendritic cells (DC), the antigen-presenting cells that trigger the T helper 1-mediated immune response, is crucial for effective DC-based cancer immunotherapy. It determines increased expression of DC maturation markers (e.g., the surface molecules CD40, CD80, and CD86, and MHC classes I and II), release of proinflammatory cytokines (IL-6, IL-12, IL-1b, TNF-a), and enhanced DC migratory capacity [18, 19]. Moreover, nude TLR4 −/− mice developed significantly increased tumor volumes when inoculated with the breast cancer cell line 4T1 .
On the other hand, a hyperactive TLR4 signal may lead to enhanced systemic inflammation, which is known to be proatherogenic. For example, the human TLR4 polymorphism Asp299Gly, which impairs the efficacy of TLR4 signaling and the capacity to elicit inflammation, is associated with a decreased risk of atherosclerosis . Currently, such a risk is of special concern in the care of metastatic colorectal cancer, taken into consideration the implementation, in the standard first-line therapy, of new antiangiogenic drugs (e.g., bevacizumab) with the increased risk of arterial thromboembolic events .
The aim of the present prospective observational study was to evaluate whether CD45RO, PD1, or TLR4 expression in the peripheral blood, at baseline or after treatment, were associated with efficacy and safety outcome of a standard first-line chemotherapy containing fluorouracil, irinotecan, and bevacizumab in mCRC patients.
Patients and methods
Patients with histologically confirmed diagnosis of colorectal cancer and measurable metastatic disease at standard CT scan, not suitable for radical metastasectomy, were eligible for this prospective observational study.
They were required to be chemotherapy-naïve for metastatic disease.
Adjuvant treatment was allowed provided that it was completed >6 months prior to study entry.
Patients had to have ECOG performance status 0–1 and adequate hematologic, hepatic, and renal functions. Prior treatment with monoclonal antibodies was not allowed.
A standard first-line therapy containing fluorouracil–leucovorin–irinotecan (FOLFIRI) in association with bevacizumab was chosen for the protocol. Even though published data have not shown a clear difference between irinotecan- and oxaliplatin-based regimens [3, 23], FOLFIRI was preferred as oxaliplatin might have been administered before study entry as a part of adjuvant therapy.
All patients were treated with the following drug doses and schedules: irinotecan 180 mg/m2/day 1, bolus 5-fluorouracil 400 mg/m2 days 1 and 2, folinic acid 200 mg/m2 days 1 and 2, continuous infusion of 5-fluorouracil 1,200 mg/m2 for 48 h starting from day 1, and bevacizumab 5 mg/kg day 2; cycles were repeated every 2 weeks.
Exclusion criteria for study entry were as follows: coagulation disorders, clinically relevant cardiovascular disease, or other cancers within the previous 5 years, except for adequately treated squamous or basal cell carcinoma of the skin or carcinoma in situ of the cervix.
Blood sample for flow cytometry analysis of circulating immune cells was collected for all patients on day 1 of the first and third cycle, before the treatment administration.
Treatment was continued until tumor progression, unacceptable toxicity, or request of termination by individual patients.
Clinical and biochemical evaluation of toxicity was performed before each cycle and graded according to the NIH Common Terminology Criteria for Adverse Events version 3.0 (http://ctep.cancer.gov/protocolDevelopment/electronic_applications/ctc.htm).
Assessment of response was evaluated after every six cycles of treatment by comparison of a thorax/abdomen/pelvis CT scan, according to RECIST criteria, with the baseline CT scan that had been performed within 45 days prior to chemotherapy initiation .
Data on gender, age, ethnicity, primary tumor site (colon vs. rectum), number of organs involved by metastatic disease, ECOG performance status (PS), histologic features, circulating tumor markers (carcinoembryonic antigen (CEA) and carbohydrate antigen 19.9 (CA 19.9)), routine hematology, and biochemical blood tests were collected before treatment commencement. ECOG PS evaluation, CEA, CA 19.9, routine hematology, and biochemical blood tests were also performed during treatment at each cycle.
Baseline data were considered evaluable if collected within 2 weeks before the start of the treatment.
Metastatic sites of disease were grouped into five categories (liver, lymph nodes, peritoneum, lungs, and other). CEA and CA19.9 testing were performed in a single laboratory with the laboratory-defined upper limit of normal (ULN) being 5 ng/mL and 35 UI/mL, respectively. Biochemical variables known to have prognostic value in mCRC were evaluated for their association with patient outcome by using a Cox regression multivariate analysis (see below).
All patients signed an informed consent to blood collection and study participation according to the protocol.
The study was approved by the local ethics committee and institutional review board and therefore performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki. The study was registered at clinicaltrial.gov (Registration Number: NCT01533740).
Purification of immune cells
30–50 mL of blood was drawn from patients and heparinized (10 U/mL). Peripheral blood mononuclear cells (PBMCs) were extracted by centrifugation over Lymphoprep (Axis-Shield, Oslo, Norway). The cells were washed twice in PBS before, and all were frozen in liquid nitrogen before use. With the lymphoprep separation, part of the neutrophile was also collected and assessed for TLR4 expression (see below).
Flow cytometric analysis
PBMCs were stained with fluorescently labeled antibodies to CD4 (clone RPA-T4), CD8 (clone RPA-T3), CD45RO (clone UCHL1), and PD-1 (clone BNI3) (BD Pharmingen). All stainings were performed in phosphate-buffered saline (PBS), 2.5 % FCS, 5 mM EDTA.
TLR4 staining was performed on neutrophiles using a primary antihuman TL4 biotinylated antibody and a secondary avidin-FITC fluorophore. Mean TLR-4 expression (mTLR4) was measured on neutrophiles population and expressed as fluorescence intensity unit (FIU).
All flow cytometry was performed on a BD FacsCalibur and analyzed using the Summit v3.1 analysis program (build 839 Dakocytomation, Fort Collins, Colorado).
A positive control (blood drawn from a single healthy donor) was always analyzed in each experiment to minimize inter-experiment variability. A 5 % variation of the positive control was considered tolerable to accept overall results as reliable; otherwise, the experiment was repeated.
The primary endpoint was to evaluate the predictive value of baseline and one-month immune variables for progression-free survival (PFS), defined as the period beginning on the first day of treatment and ending with the first observation of disease progression or death from any cause. If a patient had not reached the progression endpoint, PFS was censored at the time of the last follow-up.
The secondary endpoint was the association between immune variables and the following: (1) overall survival (OS), defined as time from study entry to death from all causes— if a patient had not died, OS was censored at the time of the last follow-up— (2) overall RECIST-defined radiological response rate (ORR) divided into two main categories: presence (complete + partial responses) or absence (stable + progression of disease) of radiological tumor response; and (3) toxicity.
The association between immune variables and PFS or OS was examined using Kaplan–Meier curves. In the univariate survival analysis, the estimated hazard ratio (HR) and its associated 95 % confidence interval (95 % CI) were based on the log-rank test.
Cox proportional hazards regression model was used to evaluate the association between immune cell values along with clinical and biochemical covariates known to have prognostic value in mCRC and the primary endpoint (PFS) in the multivariate analysis .
The entered method was used to enter independent variables into the model (i.e., all variables were entered into the model in one single step without checking for significance).
The association between immune cells frequencies or TLR4 expression dichotomized as below or above the median and ORR was evaluated by Fisher’s exact test.
Changes of immune variables over the first month of treatment were assessed with the Wilcoxon test for paired samples.
All statistical tests were two-sided. Analyses were done using MedCalc for Windows, version 188.8.131.52 (MedCalc Software, Mariakerke, Belgium).
Patients’ characteristics and treatment delivery
Characteristics of the 31 study patients
Other than liver
Liver + other sites
Primary tumor status
6.8 × 103/mm3
3.9–10.4 × 103/mm3
Köhne risk class
All included patients were treated with FOLFIRI-B as per protocol for at least two cycles. The median age at the study entry was 69 years (range 41–83). Twenty-nine patients were Caucasian and two Hispanic. ECOG PS was 0 for 21 patients and 1 for 10 patients. In five cases, the primary tumor was located in the rectum (within 15 cm from the anal verge).
Most of the patients presented with synchronous liver metastases. Primary tumor was resected in 79 % of cases.
Routine biochemical tests, some of which are well recognized prognostic factors, were performed at baseline (values reported in Table 1). Adverse biochemical features of WBC > 10,000, ALP > ULN (i.e., 129 U/L), LDH > ULN (245 U/L), CEA > ULN (i.e., 5 ng/mL), CA19.9 > ULN (i.e., 37 U/mL) were observed in 5, 74, 32, 68, and 37 % of cases, respectively. Vast majority of the patients (79 %) fell into the good prognostic class according to Kohne criteria.
Detailed demographic and clinical features are summarized in Table 1.
As of March 2013, 19 patients died and 12 were alive. Median follow-up of surviving patients was 23 months (range 10–36). Twenty-three out of 31 patients had reached the progression endpoint as per protocol definition.
Median number of treatment cycles administered was 12 (range 3–24). Dose reduction (>25 % of the planned dose for any of the drug) or treatment delay (>1 week of delay for two or more cycles) was necessary in 52 % of patients.
General immune cell analysis
Absolute number of WBCs, PBMCs, and neutrophiles were analyzed using routine hematology test. There were no significant changes between baseline and after two cycles in absolute numbers WBCs (6.8 × 103/mm3 vs. 6.1 × 103/mm3), neutrophiles (4.4 × 103/mm3 vs. 3.5 × 103/mm3), or PBMCs (2.3 × 103/mm3 vs. 2.1 × 103/mm3), all p > 0.05.
Median values of immune variables at baseline and after 1 month (2 cycles) of therapy
Wilcoxon p value
Wilcoxon test for paired samples was used to assess for differences between pre- and post-one-month of therapy (pre-third-cycle, ptc). No significant changes were detected with p values ranging between 0.09 and 0.952; however, a mild increase of borderline statistical significance in CD45RO+ cell frequency (%) was seen after two cycles of therapy.
To test if individuals with high or low immune values represented two distinct subsets of patients, variation in clinical features (some of which known to be clinical prognostic factors) was investigated between subgroups with below or above the median PD1+, CD45RO+CD8+, or CD45RO+CD4+ cell frequencies and mean neutrophil TLR4 expression. No significant differences were detected (p values ranging between 0.220 and 0.989), arguing that no clear interaction between immune variables and known clinical prognostic factors can be hypothesized.
Prognostic value of analyzed immunophenotype
We first evaluated if baseline immunophenotypic values, or trend after 1 month, were predictive of PFS, with variables dichotomized based on their median.
Immune variables at baseline and after 1 month and progression-free survival (PFS) in the 31 enrolled patients
mPFS (months) above versus below the median
95 % CI
12.4 versus 12.9
12.4 versus 12.9
20.1 versus 11.4
12.9 versus 11.4
20.4 versus 11.4
11.4 versus 12.7
12.7 versus 12.4
CD4 + cell %
13.4 versus 10.6
CD8 + cell %
12.9 versus 11.4
PD1 + cell %
22.4 versus 11.9
CD45RO + cell %
20.1 versus 11.4
CD45RO + CD8 + cell %
22.4 versus 9.4
12.9 versus 12.4
12.6 versus 12.5
Cox multivariate regression analysis for PFS of the 31 enrolled patients
95 % CI of Exp(b)
ptc-CD45RO+CD8+cell% (>vs<12 %)
ptc-PD1+cell% (>vs<24 %)
CEA (>vs<5 ng/mL)
Kohne risk class (high-intermediate vs. low)
LDH (>vs<245 U/L)
Correlation between dynamics of circulating immune profile and RECIST response
To assess if changes over the time of immune variables (increase or decrease) were associated with tumor growth or shrinkage, Fisher’s exact test was used to test for differences in the proportion of radiological responders or nonresponders according to the one-month trend of analyzed immune variables. The trend of CD45RO+CD8+cell% after two cycles (increase vs decrease) was the only immune variable that significantly correlated with RECIST objective response rate, ORR = 83 vs. 14 %, respectively, p 0.01.
As per protocol, a secondary objective was the association between immunophenotype and toxicity, and in particular between TLR4 expression and risk of arterial thromboembolic events (ATEs). One CNS ischemic event was observed in a patient after ten cycle of treatment confirmed by a repeated CT scan and that required chemotherapy interruption. In this patient, neutrophil mTLR4 expression was above the median (65 FIU).
Treatment was in general well tolerated with an anticipated safety profile. Overall grade 3–4 toxicities were observed in 55 % of cases, with the most common being leuconeutropenia (43 %), mucositis (13 %), diarrhea (7 %), and hypertension (2 %). None of the immune variables was significantly associated with any of the common toxicities.
In the present prospective observational trial, we demonstrated for the first time that findings obtained from primary tumor tissue of radically resected colorectal cancer may be repeated in the peripheral blood of metastatic patients treated with standard first-line chemotherapy. In particular, circulating CD45RO+CD8+ cells after two cycles of FOLFIRI-B is a favorable predictor of prolonged PFS and OS with a 67 and 69 % reduction in the risk of progression and death, respectively, for CD45RO+CD8+ cells above the median of 12 %. Moreover, baseline ptc-CD45RO+CD8+ cells were the only independent prognostic factor in a Cox regression multivariate analysis, including commonly used prognosticators such as CEA, LDH, and Köhne risk class [25–27].
Identification of reliable prognostic factors for colorectal cancer patients is the focus of intensive clinical and translational research [28, 29], and the base of the present study comes from the researches by Pagès et al.  demonstrating that the combined immunohistochemical analysis of CD8+ and CD45RO+ cells in specific tumor tissue regions may be predictive of tumor recurrence and overall survival in patients with early-stage, radically resected colorectal cancer.
In that study, authors demonstrated as first step that in 15 tumors with high density of memory CD45RO+ cells, upregulation of gene clusters referring to CD8 cytotoxicity and TH1 orientation (e.g., CD8, granzymes, perforin, T-bet, interferon-gamma, interleukin 12Rb 1 and 2, and IL-18) was present. They then demonstrated, in a cohort of 411 colorectal cancer with apparently good prognosis (all stage I or II), that patients with low density of both CD8+ and CD45RO+ had sixfold–sevenfold increased risk of death (HRs 6.17–6.72) compared to patients with high density of these immune cells. In the multivariable analysis, infiltration of CD8+ and/or CD45RO+ cells was an independent prognosticator together with pT stage and clinical presentation with bowel perforation. These results were further confirmed in an independent cohort of 191 patients.
In the Pagès study, multivariate analysis was performed with the factors that are related to risk of relapse for early-stage tumors (i.e., clinical presentation and pT), while we introduced risk factors that are more related to systemic cancer dissemination such as blood tumor markers (CEA) or inflammation markers (LDH, Khone risk class). Moreover, Pagès et al. separately analyzed CD8 and CD45RO expression, whereas flow cytometry of peripheral cells allowed us to evaluate the combined expression of both markers and therefore distinguish whether the prognostic value of this immunophenotype relied on two distinct cellular compartments (i.e., CD45RO+CD8− cells and CD45RO−CD8+ cells) or on a single cell subset co-expressing both CD45RO and CD8. We demonstrated, in our small cohort, that prognosis depends on the double positive CD45RO+CD8+ cells, but not on the whole population of CD8+ or CD45RO+ cells (Table 3).
In our study, the modulation of immune response determined by FOLFIRI-B was crucial for metastatic patients. High CD45RO+CD8+ cells at baseline were associated with prolonged PFS, but this did not reach statistical significance (HR 0.59, p value 0.33). Conversely, immune response following two cycles of FOLFIRI-B was a reliable marker of improved outcome, meaning that the influence of standard chemotherapy on antitumor immunity has a significant impact on prognosis.
In our cohort, we apparently found opposing results on the role of PD1 as compared to the literature. Initial reports suggested a possible adverse effect of PD1 expression on colorectal cancer progression as PD1 inhibition with specific monoclonal antibodies had antitumor activity in mCRC patients . However, PD1 pathway was found not significant for colorectal cancer in larger trials [31, 32], and this was in line with the absence of association between baseline PD1+ cells and PFS in our study. However, post-treatment PD1 expression may indicate a triggered immune response as it functions as immunosuppressor signal following activation of cytotoxic immune defence to avoid excessive autoimmunity. This would explain the longer PFS observed for high ptc-PD1+% as it may be an indirect sign of chemotherapy-elicited antitumor immune response.
Due to the limited sample size, only one ATE was observed during the study and no inference could be drawn on possible risk of cardiovascular events in presence of high TLR4 expression. Overall, no clear association with toxicity was revealed for analyzed immune variables.
In conclusion, we confirmed that CD45RO+CD8+ population evaluated in the bloodstream may have meaningful prognostic value also in metastatic colorectal cancer patients, but this has to be evaluated in the light of chemotherapy–immune system interaction. Measuring this T cell population in the peripheral blood is methodologically simple; furthermore, the enhancement of CD45RO+CD8+ response may be a fascinating therapeutic target to ameliorate clinical outcome of this setting of patients treated with standard chemotherapy.
This study has been carried out within the Ph.D. program in Experimental and Systems Medicine, XXV cycle, ‘Tor Vergata’ University of Rome.
Conflict of interest
Authors declare that there is no conflict of interest in connection with this paper.