Imaging tumor-stroma interactions during chemotherapy reveals contributions of the microenvironment to resistance

Abstract

Little is known about the dynamics of cancer cell death in response to therapy in the tumor microenvironment. Intravital microscopy of chemotherapy-treated mouse mammary carcinomas allowed us to follow drug distribution, cell death, and tumor-stroma interactions. We observed associations between vascular leakage and response to doxorubicin, including improved response in matrix metalloproteinase-9 null mice that had increased vascular leakage. Furthermore, we observed CCR2-dependent infiltration of myeloid cells after treatment and that Ccr2 null host mice responded better to treatment with doxorubicin or cisplatin. These data show that the microenvironment contributes critically to drug response via regulation of vascular permeability and innate immune cell infiltration. Thus, live imaging can be used to gain insights into drug responses in situ.

Figure 1. Visualizing cellular responses to chemotherapy in vivo

(A) Doxorubicin-induced cell death in tumors of live MMTV-PyMT;ACTB-ECFP;c-fms-EGFP mice. Induction of cell death (red), visualized by propidium iodide (PI) uptake, can be seen in a tumor lesion (blue) infiltrated with myeloid cells (green). Red arrows point to cell death in the tumor, white arrow point to cell death in the stroma. Time after treatment (doxorubicin) or initiation of imaging (control) is indicated. Scale bar: 100 μm. (B) Dynamics of cell death in situ. Single cell death was observed as the appearance of PI staining (red arrow). This was followed by myeloid cell infiltration (white arrows). Time after doxorubicin treatment is indicated. Scale bar: 10 μm. (C) Dynamics and types of nuclear structural changes after doxorubicin treatment in MMTV-PyMT;ACTB-ECFP;H2B-EGFP mice. Examples of necrosis-like (nucleus becomes PI⁺ without major structural changes) and apoptosis-like cell death (nuclear structure breakdown before PI uptake). Time after treatment is indicated. Scale bar: 10 μm. (D) Doxorubicin induces necrosis-like cell death (mean ± SEM, 18 fields of view [FOV] from 3 mice were counted, p=0.009 at 32 hr, p<0.001 at 42 hr, Student’s t-test). Also see Figure S1 and Movies S1 and S2.

Figure 2. Doxorubicin sensitivity changes with tumor stage in vivo but not in vitro

(A) Doxorubicin treatment reduces tumor volume in MMTV-PyMT mice. Mice were given doxorubicin or PBS at days 0, 7 and 14 (indicated by arrows). Analysis includes 73 tumors from 11 doxorubicin-treated mice and 56 tumors from 12 PBS-treated mice (mean ± SEM, p<0.0001 for all time-points, Student’s t-test). (B) Doxorubicin-induced cell death preferentially occurs in early carcinomas and not in hyperplasias and late carcinomas. Different microenvironments imaged in the same MMTV-PyMT;ACTB-ECFP;c-fms-EGFP mouse after doxorubicin treatment. Time after treatment is indicated. Scale bar: 100 μm. (C) Cell death increases between 24 and 36 hr after doxorubicin treatment (n=25 fields of view [FOV]) as compared to control mice imaged for similar time frames (n=23 FOV, Fisher’s exact test, p=0.001). Cell death increases in early carcinomas (12 of 14 doxorubicin-treated vs. 1 of 8 control-treated FOV, Fisher’s exact test, p=0.002), but not in hyperplasias (1 of 4 vs. 1 of 8 fields) or late carcinomas (3 of 7 vs. 2 of 7). Four control- and 4 doxorubicin-treated mice were imaged. (D) DNA damage response measured by γ-H2AX immunostaining differs between tumor stages (mean ± SEM, ANOVA, p<0.0001), and is higher in early carcinomas than in hyperplasia (p<0.01, Bonferroni post-test) or late carcinomas (p<0.0001, Bonferroni post-test, 21 hyperplasia, 25 early carcinoma, and 24 late carcinoma FOV were evaluated in 5 tumors from 5 mice). (E) Cancer cell proliferation determined by BrdU labeling does not differ between tumor stages in MMTV-PyMT mice (mean ± SEM, n.s., ANOVA, 13 hyperplasia, 19 early carcinoma and 28 late carcinoma FOV were evaluated in 6 tumors from 3 mice). (F) Proliferating (BrdU⁺) cells are not preferentially killed by doxorubicin treatment. The fraction of BrdU⁺ cells with cell-death associated nuclear changes was scored (mean ± SEM, n.s. Student’s t-test. For each condition, 58-60 FOV were analyzed in 6 tumors from 3 mice). (G) Sensitivity to the ErbB1/ErbB2 inhibitor lapatinib is highest for cancer cells from late stage tumors in culture (mean ± SEM, p=0.04 or 0.02 as indicated, Student’s t-test. Values represent averages of 4 experiments, each done in triplicate with primary cells from independent mice). (H) Sensitivity to doxorubicin is not affected by tumor stage in culture (mean ± SEM, n.s., ANOVA. Values represent averages of 4 experiments, each done in triplicate with primary cells from independent mice). (I) Morphology of organoids generated from tumors at different stages after treatment with the indicated concentrations of doxorubicin. Scale bar: 50 μm. Also see Figure S2 and Movie S3.

Figure 3. Distribution and uptake of doxorubicin in different tumor sizes

(A) Doxorubicin leakage and uptake varies with tumor size. MMTV-PyMT mice with small (just palpable tumors), medium (~160 mm³), or large tumors (~845 mm³) were injected intravenously with AngioSPARK 680 (green) and doxorubicin (red). Minimal leakage and uptake were seen in small tumors. Leakage was high in medium and some large tumors, with variable nuclear retention. Time after doxorubicin injection is indicated. The last column shows nuclear uptake of doxorubicin (higher magnification of the areas outlined in the fourth column). Scale bar: 100 μm. (B) Decay of intravascular doxorubicin levels determined by imaging is modeled as one-phase exponential decay (black line) (mean ± SEM. Eight FOV from two mice were analyzed). (C) Kinetics of doxorubicin leakage from vessels in tumors of different sizes. Quantification of the percentage of extravascular pixels with doxorubicin signal above intensity threshold (150% of background) and as a ratio of extravascular doxorubicin to vascular area (mean ± SEM. Eight small, 4 medium and 11 large FOV were analyzed in 6 mice). (D) Doxorubicin concentration is higher in tumors than hyperplasias and normal mammary glands (means are shown, p<0.0001, ANOVA, and Bonferroni post-tests for comparison of groups, p<0.05 or n.s. as indicated. Ten normal mammary glands from 5 FVB/n mice, and 14 hyperplastic and 47 tumors (13-81 mm³) from 5 MMTV-PyMT were analyzed). (E) Doxorubicin concentration is higher in tumors that were soft upon dissection than in hyperplastic tissue or solid tumors of similar size (means are shown, p=0.01, ANOVA, and Bonferroni post-tests for comparison of groups, p<0.01 or 0.05 as indicated). Ten hyperplastic, 15 soft medium-sized (60-350 mm³), 10 solid medium-sized (60-350 mm³), and 21 solid large tumors (351-1400 mm³) from 5 MMTV-PyMT mice were analyzed. Also see Movies S4 and S5.

Figure 4. Leakage and distribution of intravenously injected dextran varies with tumor stage

(A) Early carcinomas exhibit a higher degree of dextran leakage than hyperplasias and late carcinomas. More extensive leakage of 10 kD than 2 MD dextran is shown for all tumor stages. Kinetics of 10 kD (red) and 2 MD dextran (green) leakage from vessels in three different tumor stages from the same MMTV-PyMT;ACTB-ECFP mouse are shown. Regions with leakage of both dextrans are depicted in white and the epithelium in blue (ACTB-ECFP). Time after dextran injection is indicated. Scale bar: 100 μm. (B) Quantification of the percentage of pixels above intensity threshold (150% of background) for the two dextrans in the epithelial and stromal compartments (mean ± SEM, 4 hyperplasias from 2 mice, 5 early carcinomas from 4 mice, and 5 late carcinomas from 4 mice were analyzed). (C) The extent of c-fms-EGFP⁺ myeloid cell infiltration in MMTV-PyMT;ACTB-ECFP;c-fms-EGFP mice and the degree of 10 kD dextran leakage were scored independently using the indicated pixel-grid. Co-injected 2 MD dextran was used to differentiate between intra- and extravascular 10 kD dextran. Examples of pixel-fields scored as maximal infiltration and leakage are indicated by the white boxes. Scale bar: 100 μm. (D) Myeloid cell infiltration correlates with dextran leakage (r=0.56 and r=0.47, for observer A and B, respectively, p<0.0001, Spearman’s rank correlation coefficient, 720 FOV from two mice were scored). Also see Figure S3 and Movie S6.

Figure 5. Matrix metalloproteinase-9 is expressed by myeloid cells and influences vascular leakage and response to doxorubicin

(A) MMP9 is expressed by tumor-infiltrating myeloid cells, including those marked by 7/4 (neutrophils/monocytes) and F4/80 (macrophages). Scale bars: 50 μm. (B) Vascular structure of tumors in MMTV-Neu;Mmp9^+/+ and MMTV-Neu;Mmp9^−/− mice analyzed by perfusion with FITC-conjugated tomato lectin (green, labels all blood vessels) and rhodamine-conjugated Ricinus communis agglutinin I (red, labels basement membrane exposed to the vascular lumen in leaky vessels). Scale bar: 100 μm. (C) Vascular volume as determined by perfusion with tomato lectin does not differ between MMTV-Neu;Mmp9^+/+ and MMTV-Neu;Mmp9^−/− mice (mean ± SEM, n.s., Student’s t-test, 33 FOV in 10 tumors from 5 MMTV-Neu;Mmp9^+/+ mice and 29 FOV in 8 tumors from 3 MMTV-Neu;Mmp9^−/− mice were analyzed). (D) Blood vessels are leakier in MMTV-Neu;Mmp9^−/− tumors. The percentage of the vasculature that is positive for Ricinus communis agglutinin I shown (mean ± SEM, p=0.004, Student’s t-test, 33 FOV in 10 tumors from 5 MMTV-Neu;Mmp9^+/+ mice and 29 FOV in 8 tumors from 3 MMTV-Neu;Mmp9^−/− mice were analyzed). (E) Immunostaining for phospho-Y731 VE-cadherin. Scale bar: 100 μm. (F) Depletion of MMP9 increases phosphorylation of VE-cadherin in endothelial cells (mean ± SEM, p<0.0001, Student’s t-test, 113 vessels from tumors of 9 MMTV-Neu;Mmp9^+/+ mice and 101 vessels from tumors of 9 MMTV-Neu;Mmp9^−/− mice were examined). (G) Pericyte coverage is decreased in the absence of MMP9. Double immunofluorescence was used to determine the ratio of αSMA-positive pericytes to CD31-positive endothelial cells (mean ± SEM, p=0.002, t-test, 87 vessels from tumors of 10 MMTV-Neu;Mmp9^+/+ mice and 71 vessels from tumors of 9 MMTV-Neu;Mmp9^−/− mice were examined). (H) MMTV-Neu;Mmp9^−/− (n=7 tumors from 5 mice) respond better to treatment with doxorubicin than MMTV-Neu;Mmp9^+/+ tumors (n=11 tumors from 6 mice) (mean ± SEM, * indicates p<0.05 (Student’s t-test). Tumors below 256 mm³ at the beginning of treatment were excluded from the analysis. Also see Figure S4.

Figure 6. Myeloid cells are recruited to areas of tumor necrosis

(A, B) The infiltration of (A) alternatively activated macrophages (n=40 FOV from tumors of four mice per condition) and (B) the total macrophage population (n=22-30 FOV from 4-5 mice per condition) is increased in Mmp9^−/− host mice transplanted with MMTV-PyMT tumor cells. The infiltration of these cells decreases 48 hr after doxorubicin treatment (mean ± SEM, n.s. or significant as indicated, Student’s t-test). (C) Myeloid cells are recruited to tumors after doxorubicin treatment (12 of 16 movies) as compared to tumors of PBS-treated control mice (3 of 16 movies, p=0.004, Fisher’s exact test. Three mice were analyzed per condition). (D) Dynamics of myeloid cell infiltration (arrow) into an area of necrosis in a doxorubicin-treated MMTV-PyMT;ACTB-ECFP;c-fms-EGFP mouse. Time after treatment is indicated. Scale bar: 100 μm. Also see Figure S5 and Movies S7, S8 and S9.

Figure 7. Myeloid cells are recruited to doxorubicin-treated tumors through a stromal CCL2/CCR2 chemokine/chemokine receptor axis

(A) Protein array identifies CCL2 as the most upregulated chemokine in tumor lysates 48 hr after doxorubicin treatment (p=0.09 for CCL2 and p=0.03 for CCL12, Student’s t-test). Each column represents a tumor from a different mouse. (B) The number of CCL2-expressing cells increases 48 hr after treatment with doxorubicin (mean ± SEM, p=0.04, Student’s t-test, 80 FOV from 6 PBS-treated and 66 FOV from 6 doxorubicin-treated mice were analyzed). (C, D) Endothelial cells, pericytes and fibroblasts do not express CCL2. Tumor tissue from MTMV-PyMT mice isolated 48 hr after treatment with doxorubicin was immunostained for CCL2 and (C) α-smooth muscle actin (αSMA, a pericyte and fibroblast marker), or (D) MECA-32 (an endothelial cell marker). >300 αSMA or MECA-32 positive cells from PBS and doxorubicin-treated tumors were observed and none were positive for CCL2. Scale bar: 50 μm. (E) Doxorubicin treatment results in infiltration of 7/4⁺CCR2⁺ cells with monocytic but not polymorphonuclear (PMN) morphology. Double immunostaining for CCR2 and 7/4 with scoring of nuclear morphology (mean ± SEM, Student’s t-test, significance levels as indicted, 104 FOV from 4 PBS-treated and 113 FOV from 5 doxorubicin-treated mice). (F) Doxorubicin results in CCR2-dependent myeloid cell infiltration. The percentage of Gr1⁺7/4⁺ of all CD11b⁺ myeloid cells in tumors was determined by flow cytometry (mean ± SEM, significance levels as indicated, Student’s t-test, n=10-11 mice). (G) FACS plots with indication of the percentages of the gated cell populations from representative tumors. Also see Figure S6.

Figure 8. Host CCR2 regulates response to doxorubicin

(A) Tumors in MMTV-PyMT;Ccr2^−/− (26 tumors from 10 mice) respond better to doxorubicin than those in MMTV-PyMT;Ccr2^+/− (15 tumors from 8 mice). Results for tumors with a pretreatment volume of 250-750 mm³ (mean ± SEM, * indicates p<0.05, Student’s t-test). (B) Tumors in Ccr2^−/− host mice respond better to doxorubicin than those in C57BL/6 hosts. Two cohorts were treated with doxorubicin and showed similar results. The results of one cohort are shown (mean ± SEM, * indicates p<0.05, Student’s t-test. Eight tumors in 6-8 hosts were analyzed per condition. One Ccr2^−/− and two C57BL/6 hosts were euthanized on days 13-16 due to poor health). (C) Tumors in Ccr2^−/− hosts respond better to cisplatin than those in C57BL/6 hosts mean ± SEM, * indicates p<0.05, ** indicates p<0.01, Student’s t-test. n=22-24 tumors in 11-12 hosts). (D) Tumors in Ccr2^−/− hosts are more cystic and contain fewer cancer cells acutely (48 hr) after doxorubicin treatment. Relapsed tumors (6 weeks after treatment) in Ccr2^−/− hosts are low-grade with decreased cellularity and necrosis (p=0.005, Fisher’s exact test, 10 low-grade and 5 high-grade tumors in Ccr2^−/− hosts vs. 1 low-grade and 11 high-grade tumors in C57BL/6 hosts). Scale bar: 100 μm. (E) Vascular structure of tumors in Ccr2^−/− and C57BL/6 hosts analyzed by perfusion with FITC-conjugated tomato lectin (green) and rhodamine-conjugated Ricinus communis agglutinin I (red). Nuclei are stained with DAPI. Scale bar: 100 μm. (F) Vascular volume is increased in tumors in Ccr2^−/− hosts compared to C57BL/6 wild type hosts as determined by perfusion with tomato lectin (mean ± SEM, p=0.005, Student’s t-test, analysis of 12 tumors (from 6 mice) per genotype and 5-10 fields of view per tumor). (G) The total volume of leaky vasculature does not differ between tumors in Ccr2^−/− and C57BL/6 hosts (mean ± SEM, n.s., Student’s t-test, analysis of 12 tumors from 6 mice per genotype and 5-10 FOV per tumor). (H) Relative leaky vasculature is decreased in tumors of Ccr2^−/− hosts compared to C57BL/6 hosts as determined by the percentage of the vasculature that is positive for Ricinus communis agglutinin I (mean ± SEM, p=0.01, Student’s t-test, analysis of 12 tumors from 6 mice per genotype and 5-10 FOV per tumor). Also see Figure S7.