Ascitic fluids and blood samples from individuals with ovarian cancer were collected from chemotherapy-naïve patients who underwent a tumor surgical resection at the Claudius Regaud Institute (IUCT Oncopole, Toulouse, France). We also analyzed blood samples of patients included in a phase I multicentric clinical trial who underwent an immunotherapy treatment with murlentamab (GM102, a low-fucose IgG1 anti-AMHRII antibody). Only blood samples from the six patients managed by IUCT were analyzed. The progression of the disease was established by the analysis of tumor markers (CA125 or inhibin B) and of tumor lesions evaluated by Response Evaluation Criteria In Solid Tumors (RECIST V.1.1). Blood samples from healthy volunteers were provided by the French Blood Establishment (EFS).

Blood samples and ascitic fluids were obtained using an aseptic technique from 28 chemotherapy-naïve patients with ovarian cancer. Blood samples were also obtained from seven healthy subjects and from six patients with ovarian cancer who underwent a tumor-targeted immunotherapy treatment.

Mononuclear cells were isolated from patients’ blood and ascites with a Ficoll density gradient (Lymphoprep, Stemcell Technologies). The fractions containing peripheral blood mononuclear cells or ascitic mononuclear cells were then washed two times. Cell number was evaluated and 250 000 cells were stained for flow cytometry analysis.

All analyses were gated on viable cells after live/dead staining (Molecular Probes LIVE/DEAD Fixable Violet Dead Cell Stain Kit, Life Technologies).

To study blood monocyte populations, CD14 and CD16 were respectively detected after staining with CD14-PerCPVio700 and CD16-PE antibodies (Myltenyi Biotec). To study the phenotype of blood monocytes and macrophages from ascitic fluid, cells were labelled with the following antibodies: CD14-PerCPVio700, CCR2-PEVio770, CD163-PEVio770, CD86- Vioblue, TLR-2-APC, CD36-FITC, MHC2-APCVio700 (Myltenyi Biotec) and CD206-APC (BD Biosciences). In ascites, analysis of the percentage of cells expressing studied receptors and/or the level of expression of the receptors (measured as the mean fluorescence intensity) was conducted in a population of cells characterized by medium-sized granules (Side Scatter (SSC)^med) and the expression of the monocyte/macrophage receptor CD14.

To study the ascitic lymphocyte population, cells were labelled with the following antibodies: CD45-Viogreen, CD3-Vioblue, CD4-APCVio770, CD8-PerCP, CD25-PE, CD56-APC, CD19-PEVio700 and CD183-FITC (Myltenyi Biotec). For intracellular staining, cells were fixed and permeabilized with the FoxP3 staining buffer set (Myltenyi Biotec) and stained with the FoxP3-APC antibody (Myltenyi Biotec).

Appropriate fluorochrome-matched isotype antibodies were used to determine non-specific background staining. All stainings were performed on 100 µL of phosphate-buffered saline (PBS) solution without calcium and magnesium (PBS−/−) and 1% heat-inactivated fetal calf serum. A population of 10,000 cells was analyzed for each data point. All analyses were carried out in a BD Fortessa flow cytometer with Diva software.

For ascitic fluid cytokine titration, a multiplex bead-based immunoassay was used according to the manufacturer’s instructions (LEGENDPlex Mix and Match System, Biolegend).

For blood monocyte comparison between healthy subjects and patients with ovarian cancer, the data were subjected to analysis of variance followed by the Bonferroni’s multiple comparisons test. Spearman rank correlation was used to investigate the relationships between IBMs and immune/clinical parameters of patients with ovarian cancer. Analysis was performed using GraphPad Prism V.7.0. A p value of <0.05 was considered statistically significant.

In our cohort of chemotherapy-naïve patients with tovarian cancer (see online supplementary table 1 for clinical parameters of subjects), complete blood count revealed a slight increase in leukocyte concentration compared with normal values established in healthy subjects (figure 1A). Moreover, although the blood counts of some patients with ovarian cancer deviate from healthy subject reference intervals, the average concentrations for leukocyte subtypes and other blood cells did not show any significant change in regard to normal values (figure 1A).

Human blood monocytes are heterogeneous and are classified into three subsets based on CD14 and CD16 expressions. Fluorescence-activated cell sorting analysis performed in the blood of patients with ovarian cancer demonstrated that the proportions of monocyte subsets were altered (figure 1B–D). As expected, in healthy subjects (n=7), classical monocytes (CD14^high CD16^neg) represented 85%–90% of total monocytes, whereas non-classical (CD14^low CD16^high) and intermediate (CD14^high CD16^low) populations, these represented only 5%–10% (figure 1C,D). Interestingly, in 28 chemotherapy-naïve patients with ovarian cancer, there was a significant increase in CD14^high CD16^low monocytes (on average, 34%±4% of total circulating monocytes) to the detriment of CD14^high CD16^neg monocytes (on average, 61%±4% of total monocytes). The proportion of CD14^low CD16^high monocytes was the same as in healthy subjects (on average, 3%±1% of total monocytes) (figure 1C,D). These data reveal a robust expansion of the IBM population in patients with ovarian cancer.

We then characterized the phenotype of the three monocyte subsets and analyzed whether it could change in patients with ovarian cancer (online supplementary figure 1). In healthy subjects, classical, intermediate and non-classical monocytes displayed different protein levels of markers, such as CCR2, CD206, CD163, CD86, CD36 and MHC2. However, the expression level of these markers remained unchanged in patients with ovarian cancer compared with those observed in healthy donors. Together, these results support that ovarian tumors impact the proportion of monocyte subsets without altering their phenotype.

To establish whether intermediate monocyte expansion in the blood of patients with ovarian cancer may reflect the immune microenvironment in the tumor site, we evaluated the correlation between the circulating IBM population and innate and adaptive immune cells in tumor ascites.

We found significant positive correlations between the percentage of IBMs and the percentage of regulatory T cells (Spearman rank correlation coefficient (r)=0.6503, p=0.0002), of CD4+ T cells (r=0.5907, p=0.0012) and of B cells (r=0.4946, p=0.0102) in the ascites (figure 2A). Conversely, we observed significant inverse correlations between the proportion of IBMs and cytotoxic CD8+ T cells (r=−0.5509, p=0.0035), natural killer (NK) cells (r=−0.515, p=0.0068) and the CD8+/regulatory T-cell ratio present in the ascites of patients with ovarian cancer (figure 2A). Together, these results clearly demonstrate that circulating IBM expansion is associated with a decrease in the effector/regulatory T-cell ratio in tumor ascites.

The phenotypical characterization of macrophages (CD45+ CD14+ cells) in tumor ascites of the 28 patients with ovarian cancer demonstrated that almost all of them express CCR2, CD163, CD206 and CD86 (online supplementary figure 2). However, considering the variation in the expression level (geomean) of these markers on CD14+ cells, we established significant positive correlations between the proportion of IBM subset and the expression on CD14+ cells from tumor ascites of CCR2, CD163 and CD206, markers associated with ovarian tumor progression and poor prognosis (figure 2B). Conversely, we established an inverse correlation between the percentage of IBMs and the expression of CD86, costimulatory molecule for the priming and activation of T cells, on macrophages from tumor ascites (figure 2B).

The quantification of protein levels of ascites soluble mediators showed a significant negative correlation between the frequency of IBMs and interferon-γ (IFN-γ), CXCL10 and CCL3, factors that are involved in the Th1 response (figure 2C). Moreover, this inverse correlation was also observed with granzyme B, which is well described as a mediator of the cytotoxic response of T lymphocytes and NK cells. However, the frequency of IBMs was significantly correlated with the protein level in ascites of interleukin (IL)-10 and IL-6 immunosuppressive cytokines (figure 2C). Finally, the increase in IBM frequency significantly correlated with the augmentation of the proangiogenic vascular endothelial growth factor (VEGF) (figure 2C).

Together, these data link the expansion of the IBM population in patients with ovarian cancer to the peritoneal ascites protumor immunosuppressive status.

To evaluate whether the higher frequency of circulating IBMs in patients with ovarian cancer might be related to tumor burden, and hence might reflect disease progression, we examined the correlation between the circulating IBM population and the peritoneal cancer index (PCI) used to assess the extent of peritoneal cancer throughout the peritoneal cavity.16 Interestingly, we established a correlation between the percentage of IBMs and the PCI (r=0.6503, p=0.0004) (figure 2D), highlighting the circulating IBM subset as a potential signature for the progression of ovarian cancer.

In line with the link between IBMs and ovarian tumor progression, we also established a positive correlation between IBM proportion and the platelet count (figure 2E), which has been previously described to be related to poor prognosis and unfavorable clinicopathological parameters for patients with ovarian cancer.17 Except for platelets, no other correlations could be established between IBM proportions and blood cell counts (online supplementary figure 3).

We analyzed blood samples, at diagnosis and following treatment, from six patients with ovarian cancer who underwent a tumor targeted immunotherapy with murlentamab (GM102) (see online supplementary table 2 for clinical parameters of subjects). Preliminary data have demonstrated various responses under GM102 treatment from stabilization to partial response.18 Moreover, GM102 has been shown to activate the antitumor T-cell immune response.19

As observed in the first cohort of 28 chemotherapy-naïve patients with ovarian cancer (figure 1C,D), the 6 patients with ovarian cancer included in the GM102 phase I clinical trial presented a high proportion of CD14^high CD16^low IBMs and a low proportion of CD14^high CD16^neg classical blood monocytes prior to treatment (C1D1 SOI) (figure 3A).

Interestingly, we observed a decrease of IBMs, which was mirrored by an increase of classical blood monocytes over time following GM102 infusion in three out of six patients with ovarian cancer (patients 1, 2, and 3) (figure 3A). These data support that GM102 treatment can orient monocyte subset proportions towards those found in healthy subjects. In patient 1, the orientation of classical and IBM subsets towards standard proportions was accompanied by a stabilization of both tumor markers (figure 3B) and tumor lesions evaluated by RECIST V.1.1 (figure 3C). During the first few hours after GM102 treatment, we observed in patient 3 a profile similar to patient 1, that is, a decrease in IBM proportion associated with a moderate increase in tumor marker, as well as a stabilization of tumor lesions (figure 3A–C). Surprisingly, in patient 2, although the orientation of classical and IBM subsets towards standard proportions was accompanied by a stabilization of tumor marker (figure 3B), the tumor lesions gradually increased after GM102 infusion (figure 3C). These mitigated results are further corroborated by the fact that, contrary to the first two patients (patients 1 and 3), the reduction of IBM proportions observed in patient 2 during GM102 treatment was accompanied by a strong increase in CD14^low CD16^high non-classical monocytes whose deleterious role has already been described in tumors.20

In patient 4, 5 and 6, GM102 infusion was not followed by reduced IBM proportions and simultaneous increased classical monocyte proportions (figure 3A). In patient 4, the continuous IBM elevation after GM102 infusion was associated with the progression of the disease, as indicated by the increase in tumor markers and lesion development (figure 3B,C). In patient 5, while the target lesion evaluated by RECIST V.1.1 appeared to decrease (figure 3C), the failure of GM102 treatment to orient IBM proportions towards those found in healthy subjects was accompanied by the appearance of new non-target lesions resulting from a rapid tumor progression that led to the withdrawal of this patient from the study (figure 3A–C). Similarly, in patient 6, the increase in IBMs during treatment was also accompanied by an increase in tumor markers and tumor progression, as evidenced by the appearance of non-target tumor lesions in the brain, resulting in the exclusion of this patient from the study (figure 3B,C).

In addition to the established positive correlation between IBM frequency and the PCI (figure 2D), these results suggest that the expansion of IBM is linked to tumor load and could therefore be used to follow up tumor growth and to monitor treatment adaptation in patients with ovarian cancer.