Effects of regorafenib on the mononuclear/phagocyte system and how these contribute to the inhibition of colorectal tumors in mice

REG, M-2, M-4, and M-5 were provided by Bayer AG. RAW264.7 (TIB-71) and CT26.WT (CRL-2638) cells were from the American Type Culture Collection. MC38 cells were from the National Cancer Institute. The syngeneic CT26 (classified as microsatellite stable) [22] and MC38 (classified as microsatellite instable) [23] CRC mouse cell lines were derived from carcinogen-induced BALB/c and C57BL/6 mice, respectively [24, 25].

For cell culture experiments, drugs were dissolved and diluted in 100% dimethylsulfoxide (DMSO) prior to a final dilution of 1:200 in the culture medium, resulting in working drug concentrations and 0.5% DMSO. For in vivo application, REG was dissolved in polypropylene glycol, polyethylene glycol 400 (PEG400), Pluronic F68, and water (34:34:12:20) (vehicle V1) or in PEG400 and 125 mM methanesulfonic acid in water (80:20; v/v) (vehicle V2). Details of antibodies and polymerase chain reaction (PCR) probes are provided in the Methods in Additional file 1.

RAW264.7 cells (10⁶) were seeded into cell culture dishes 6 cm in diameter and incubated in 4 mL RPMI/10% fetal bovine serum (Biochrom) for 24 h at 37 °C in a humidified 5% CO₂ atmosphere. Cells were washed once with phosphate-buffered saline (PBS) and starved in 3 mL RPMI/0.1% bovine serum albumin for another 24 h. Cells were then treated for 1 h at 37 °C with drugs at the indicated final concentrations, with 0.5% DMSO as vehicle control, or left untreated, and finally stimulated for 2 min at 37 °C with 100 ng/mL recombinant mouse CSF1 (R&D Systems) by adding 1 mL prewarmed starving medium containing 400 ng/mL CSF1. Dishes were placed on ice and cells were immediately washed with ice-cold PBS and lysed with 50 µL lysis buffer (50 mM HEPES pH 7.5, 150 mM NaCl, 1.5 mM MgCl₂, 10% glycerol, 1% Triton X-100, 0.5% desoxycholate, 0.2% sodium dodecyl sulfate, 1 mM ethylene glycol-bis[β-aminoethyl ether]-N,N,N′,N′-tetraacetic acid, 100 mM NaF, 10 mM Na₄P₂O₇, 1 mM Na₃VO₄, 1% Phosphatase Inhibitor Cocktail 3 [Sigma-Aldrich], 1 mM phenylmethyl sulfonyl fluoride, Complete Protease Inhibitor Cocktail [Roche CustomBiotech; 1 tablet/10 mL], and 50 U/mL Benzonase [Sigma-Aldrich]). Lysed cells were scraped into Eppendorf tubes, incubated on ice for 20 min, and centrifuged at 4 °C for 10 min at 13,000 rpm; supernatants (lysates) were transferred into new tubes and stored at −80 °C.

Protein concentrations were determined using Pierce bicinchoninic acid assay (Thermo Fisher Scientific) and 40 µg/lane of total lysate was separated by SDS-PAGE on 4–12% Criterion XT Bis–Tris Gels (Bio-Rad) and subsequently transferred to polyvinylidene difluoride membranes using the Trans Blot Turbo Blotting System (Bio-Rad). For antigen detection, membranes were blocked with PBS Odyssey blocking reagent (LI-COR Biosciences) for 1 h at room temperature (RT). Primary antibodies were incubated overnight at 4 °C and the fluorescent secondary antibodies were incubated for 1 h at RT. After each antibody incubation, blots were washed with PBS with 0.05% Tween 20 (Carl Roth) for 15–30 min at RT. Fluorescent signals were quantitatively captured using the Odyssey DLx Infrared Imaging System (LI-COR Biosciences) and the half-maximal inhibitory concentration (IC₅₀) values were calculated using GraphPad software (version 7 or newer).

All mouse experiments were approved by the relevant regulatory agency (State Office for Health and Social Affairs, Berlin, Germany; approval number: Reg 0010/19) and conducted in compliance with the German Animal Welfare Act. Animals were kept in a 12 h light/dark cycle at a housing temperature of 23 °C. Food and water were available ad libitum. Therapies were well tolerated and decreases in body weight did not exceed 10% of body weight at the start of treatment; no drug-related deaths were observed.

CT26 tumor cells were cultivated in RPMI 1640 with 2 mM L-glutamine (Gibco, Invitrogen GmbH) containing 10% fetal bovine serum (Gibco) at 37 °C and 5% CO₂. Cell lines were maintained in culture for no longer than 6 months, and mycoplasma contamination was excluded by enzymatic test (MycoAlert, Lonza) prior to in vivo application. Tumor cells (2 × 10⁶) were transplanted subcutaneously into 6- to 8-week-old female BALB/c mice. Control mice were subjected to sham surgery without tumor cell transplantation. Blood and tumor samples were collected from untreated control and tumor-bearing mice on days 4 and 8 (n = 3 mice/group). The remaining tumor-bearing mice were randomized on day 8, by which time the tumor size had reached approximately 100 mm³, to the control groups V1 and V2 (n = 6 mice/group) and to the treatment groups REG(V1) and REG(V2) (n = 18 mice/group), and treatment was immediately started. REG was administered at 10 mg/kg daily by oral gavage (5 mL/kg) for up to 10 days. For interim analysis, blood and tumor tissue were collected from each group (n = 3 mice/group) after 7 days of treatment. For pharmacokinetic (PK)/pharmacodynamic (PD) relationship analysis, tissues were taken 24 h after the penultimate dose and 1, 3, 7, and 24 h after the ultimate dose. Tumor diameter was measured by caliper and body weight was determined at least twice weekly. Tumor volume was calculated using the formula 0.5 × a × b², where a and b are the long and short diameters of the tumor, respectively.

For tissue collection, mice were killed at predetermined time points. Blood samples were collected after retrobulbar venous plexus puncture in MiniCollect tubes containing ethylenediaminetetraacetic acid (EDTA; Greiner Bio-One). Typically, 100 µL of blood was used per flow cytometry (FC) staining. The remaining material was centrifuged for 10 min at 3000 g at 4 °C and the plasma supernatant was snap-frozen for further analyses. Tumors were removed, and their weights determined, then divided into three pieces, one of which was snap-frozen, one formalin fixed and paraffin embedded, and one weighed and used for FC analyses.

For tumors, single-cell suspensions were prepared from the fresh tumor pieces using a mouse Tumor Dissociation Kit (Miltenyi Biotec). The final cell pellets were resuspended in 1 mL FC buffer (PBS with 2 mM EDTA, 0.5% bovine serum albumin). Then, 100 µL of the tumor single-cell suspensions and the EDTA blood samples were stained by addition of the antibody cocktail and incubated for 10 min at 4 °C according to the manufacturer’s instructions. After antibody incubation, blood samples were incubated with 2–3 mL lysis buffer (0.17 M NH₄Cl, 10 mM KHCO₃, 0.1 mM EDTA) for 10–20 min at RT to lyse and remove red blood cells. Subsequently, samples were centrifuged at 300 g for 5 min at 4 °C and cells were washed with 2–3 mL FC buffer. Tumor samples were washed immediately after antibody incubation with 2–3 mL FC buffer without the lysis procedure and centrifuged at 300 g for 5 min at 4 °C. Cell pellets from blood and tumor samples were resuspended in 100 µL and 200 µL FC buffer, respectively. Next, 50 µL of blood and 70 µL of tumor cell suspensions were analyzed by FC using MACSQuant Analyzer 10 and MACSQuant Analyzer 16 (Miltenyi Biotec). FC data were analyzed with the FlowLogic software (Inivai Technologies distributed through Miltenyi Biotec; version 7.2.1) and gated as shown in Fig. S1 (see Additional file 2), whereby up to seven subpopulations were identified in the CD45⁺ fraction using antibodies against F4/80 and CD115. In blood, the analysis focused on the CD115^hiF4/80⁻ population, henceforth referred to as CD115^hi. In tumor samples, the analysis focused on the F4/80^hiCD115^lo population, henceforth referred to as F4/80^hi.

CCL2 was detected in plasma samples and tumor tissue using the U-Plex mouse MCP-1 ELISA Kit (Meso Scale Diagnostics) according to the manufacturer’s instructions. Measurements were performed at least in duplicate. For plasma, 12.5 µL was analyzed per animal.

For tumors, snap-frozen tissue was subjected to cryogenic grinding using a Mixer Mill MM 400 and stainless steel beads (7 mm in diameter; Retsch) and subsequently lysed in Tris Lysis Buffer (Meso Scale Diagnostics). The lysates were centrifuged at 25,000 g at 4 °C for 10 min, protein concentrations were determined using the 660 nm protein assay kit (Pierce), and 2–20 μg of lysate was used for CCL2 analysis. CCL2 concentrations were related to the amounts of tumor tissue taken for the lysis procedure.

Unless otherwise specified, formal hypothesis testing was conducted at the 5% confidence level for each statistical method. Analyses and computations were performed using SAS 9.4 software, except for the PK/PD analysis in the CT26 study, which was carried out using R 3.6.1.

The repeated tumor volume caliper measurements were compared to evaluate the difference in the growth slopes between vehicle and treated groups. All measurements starting from day 8 to day 14 of the experiment were included. A covariance pattern model was applied, assuming linear growth for the log₁₀-transformed values, which corresponds to exponential growth when using the original values, and an AR(1) correlation structure between measurements of the same animal (see Additional file 1. Statistical analysis) [26].

One-sided Welch’s t-tests were applied for comparison of the average logarithmic CCL2 concentrations and logarithmic CD115^hi cell percentages of CD45⁺ cells, respectively, in PB of sham-treated mice without tumors and tumor-bearing mice on days 4 and 8 of the experiment.

An overall comparison between treatment groups of the endpoints measured on CT26 and MC38 mice after they had been killed was performed using different statistical models. All available measurements starting from day 15 of the experiment (CT26 study) and from day 14 of the experiment (MC38 study) were used for the evaluation, except for intra-daily observations in the final 24 h.

Five analysis of variance (ANOVA) models were applied to analyze the log₁₀-transformed values of the tumor weight in the CT26 study, the CCL2 levels in the tumor in the CT26 and MC38 studies, and the CCL2 levels in the PB in the CT26 and MC38 studies. The standard assumption of equal variances between treatment groups was applied for all endpoints, except for CCL2 levels in the tumor in the MC38 study, for which unequal variances between treatment groups were allowed. A negative binomial regression model was applied to evaluate percentages of CD115^hi and F4/80^hi cells within the CD45⁺ cell population in PB and tumor tissue, respectively. Notably, tumor CCL2 measurements are the result of the geometric average of two to four replicates obtained from various ELISA measurements.

A PK/PD analysis using values from 24 h after the penultimate REG application (considered as baseline) and the subsequent intra-daily measurements at 1, 3, 7, and 24 h after the last REG application was performed to assess the correlation of the endpoints and the REG concentration at the intra-daily level. Response ratios with confidence interval [27] as deviation from baseline were derived for all endpoints, as well as for REG concentration in tumors and plasma. No formal hypothesis testing was conducted.

CSF1R is involved in the differentiation of TAMs from blood monocytes and is inhibited by REG and its metabolites M-2 and M-5 [5]. This finding was confirmed in vitro in murine RAW264.7 macrophages. REG, M-2, M-4, and M-5 potently inhibited CSF1R with cellular IC₅₀ values between 18 and 273 nM across the two tested pY559 and pY721 phosphorylation sites, and these values correlated well with the biochemical values (Fig. 1A–D; see Additional file 2: Fig. S2). pY559 is described as being involved in receptor ubiquitination and degradation [28], and its inhibition was found to be consistently associated with concomitant reduction of ubiquitinated receptor (Fig. 1B). The activation of the intracellular kinases ERK and AKT, which mediate CSF1R signaling, was also inhibited by all three compounds, with IC₅₀ in the range of 239–783 nM and 63–276 nM, respectively [29] (Fig. 1A–D). This is considered to be a result of CSF1/CSF1R signaling blockade because there was no significant biochemical inhibition of ERK and AKT by any of these compounds [4, 5]. The shutdown of the entire CSF1/CSF1R signaling cascade in macrophages by REG could explain the previously observed TAM effects.

To further investigate possible CSF1R-mediated immunomodulatory effects of REG in vivo, we first evaluated the effect of 10 mg/kg/day REG, which corresponds to the maximal clinical dose of 160 mg REG per day, on the subcutaneous growth of CT26 tumors in mice (Fig. 2A) [5, 30]. To rule out potential secondary vehicle effects on the immune system, REG was administered in a second vehicle (V2) in addition to the standard vehicle (V1). Pharmacokinetically, REG exposure, in terms of area under the curve during 24 h at steady state (AUC_(0–24)ss) and maximum concentration (C_max), was lower (1.5–2-fold) in both plasma and tumor tissue after administration in V2 compared with its administration in V1 (see Additional file 2: Fig. S4). Besides underlining the importance of PK analyses to assess vehicle-dependent drug exposures, this provided the possibility to evaluate dose-dependent effects. Tumor weight was significantly reduced with REG(V1) and REG(V2) vs vehicle control; tumor volume was significantly reduced with REG(V1) only vs vehicle control (Fig. 2B–E; see Additional file 2: Fig. S5). REG(V1) was more effective than REG(V2), which correlates with the higher exposure of REG in REG(V1) vs REG(V2) and demonstrates the dose-dependent antitumor activity of REG. The vehicle groups V1 and V2 were not significantly different.

Protumorigenic TAMs are derived from blood monocytes, which are attracted by cytokines released by tumors, such as CCL2 [31]. We analyzed whether CT26 tumors affected monocytes and CCL2 levels in PB using FC gating for the REG target CD115 (see Additional file 2: Fig. S1A) and ELISA, respectively. Both the concentration of CCL2 and the percentage of CD115^hi cells were significantly higher in mice with tumors than in sham-treated control mice without tumors 8 days after the start of the experiment, but not after 4 days (Fig. 3; see Additional file 2: Fig. S6). This indicates that CT26 tumors may attract monocytes by secretion of CCL2 and suggests that surgery did not have an effect. As REG inhibits CSF1R, we then examined its effects on monocytes and CCL2.

CD115^hi monocytes were analyzed by FC in PB samples taken 24 h after the last dose of 7, 10, or 11 daily treatments with REG to assess the steady-state treatment effect (Fig. 2A). REG(V1) significantly reduced the number of CD115^hi cells in PB vs V1 by about half (Fig. 3, Fig. 4A [CD115^hi cells]). No significant effect was observed for REG(V2) in comparison with V2, which is in line with the lower REG exposure in REG(V2) and indicates that this effect may occur only at a higher REG exposure. V1 and V2 did not differ from samples of untreated mice, suggesting that these two vehicles did not affect these cells. REG did not change CCL2 levels in corresponding plasma samples in this steady-state analysis vs vehicle or untreated samples (Fig. 4A [CCL2]). Similarly, REG reduced the number of CD115^hi cells in the blood of C57BL/6 mice with subcutaneous MC38 CRC tumors without affecting the CCL2 levels (see Additional file 2: Fig. S7A and C), indicating that this effect is not model specific.

To evaluate PK/PD relationships, we analyzed blood samples taken at various time points within 24 h after the last REG dose (Fig. 2A). Both CD115^hi monocytes and CCL2 were significantly reduced by REG(V1) and REG(V2) over time. This reduction became evident 1 h after the last dose and persisted until 3 h and 7 h, respectively, before levels returned close to steady-state baseline by 24 h (Fig. 3 [Peripheral blood], Fig. 4B). The highest concentration of REG (C_max) in corresponding plasma samples occurred by 3 h before returning to baseline (Fig. 4B [REG]). A negative correlation was observed comparing the time curves of the relative concentration estimates between REG and both PD measures, CD115^hi cells and CCL2 (Fig. 4B [PK/PD]). This strongly indicates the involvement of REG, and thereby establishes a clear PK/PD relationship.

Fresh tumor tissue was dissociated and analyzed by FC using the same gating strategy as applied for the PB cells (see Additional file 2: Fig. S1B). In untreated animals, a new intratumoral immune cell population, not found in the blood, was identified, which expressed high levels of F4/80 and low levels of CD115 (Fig. 3 [Tumor tissue]; see Additional file 2: Fig. S1B). These F4/80^hi cells, considered to reflect TAMs, amounted to approximately 25% of the CD45⁺ fraction in untreated mice or vehicle controls. They were detected on day 8 of tumor growth and remained at a constant level until the end of the study (Fig. 3 [Tumor tissue]). In contrast to the blood, only very small amounts (< 0.5%) of CD45⁺CD115^hi cells were detected in the tumor. In the steady-state analysis, REG(V1) reduced the number of F4/80^hi cells significantly and dose dependently about threefold, and REG(V2) about twofold, vs V1 and V2, respectively. As expected from their different exposures, REG(V1) was more effective than REG(V2). The vehicle groups did not differ from untreated animals (Fig. 5A). When the CCL2 concentrations were determined in corresponding tumor lysates, a trend toward higher levels was found in the tumors after treatment with REG vs vehicle controls (Fig. 5A [CCL2]). Comparable results were also obtained with MC38 tumors. REG significantly reduced the F4/80^hi population, which also appeared in these tumors, by more than twofold, even at a dosage of 5 mg/kg/day, vs vehicle (see Additional file 2: Fig. S7B and C). REG also dose dependently increased CCL2 concentrations in MC38 tumors (see Additional file 2: Fig. S7B and C).

For PK/PD relationships in the tumor, samples from the same time points as for blood were analyzed (Fig. 2A). Marginal changes were observed for both PD markers, which statistically did not deviate significantly from baseline levels (Fig. 5B). When the concentration–time curves between REG and F4/80^hi cells and CCL2 were compared, an overlapping profile was observed for both PD markers (Fig. 5B [PK/PD]). At REG’s C_max, both markers exceeded baseline in the REG(V1) group, which had a higher REG exposure; they were below baseline in the REG(V2) group, in which the REG exposure was only half of that in the REG(V1) group. This observation relates to previous findings [11, 17] and supports the suggestion that REG may exert its immunomodulatory activity at an optimal biological dose.

REG treatment did not deplete tumor tissue entirely from TAMs (Fig. 5A [F4/80^hi]). We applied quantitative real-time PCR (qRT-PCR) to characterize the remaining TAM fraction. We selected various TAM-associated genes, including candidate genes specific for recently identified subpopulations [32], and compared REG(V1) with V1 tumor samples. REG significantly reduced the transcript frequencies of various genes, such as Adgre1 (F4/80) and CSF1R (see Additional file 2: Fig. S8), which is in line with our FC data, as well as Mrc1 (CD206), a marker for protumorigenic M2-TAMs, previously shown by immunofluorescence staining to be reduced by REG in orthotopic CT26 tumors [12]. Furthermore, the frequencies of Maf and Mgl2, markers of TAM subpopulations considered to be associated with sensitivity to CSF1R inhibitors, were also reduced, whereas the frequencies of MafB, VEGFA, and Hilpda, indicators of resistance to CSF1R inhibitors [32], as well as Arg1 and Nos2, were unchanged. The fact that REG does not reduce Nos2, a M1-TAM marker, in contrast to Mrc1, is an indicator of an increased M1:M2 ratio impeding tumor growth.