Effect of chemotherapy on cancer stem cells and tumor-associated macrophages in a prospective study of preoperative chemotherapy in soft tissue sarcoma

The study was approved by the University of Minnesota IRB. Thirty-one patient samples were collected before chemotherapy by biopsy and after chemotherapy at the time of primary excision, during our prospective clinical trial of PET imaging in STS [15] (Additional file 1: Table S1). Samples were obtained from 2006 to 2014 from patients ≥ 16 years old with high-grade (FNCLCC grade 3) STS of the extremities or body wall whose tumors were greater than 5 cm in maximum diameter. Patients received pre-operative chemotherapy followed by wide surgical excision of their tumor and subsequent external beam radiation; the goal of the clinical trial was to correlate treatment response with early PET changes [15]. The preferred chemotherapy regimen was pegylated-liposomal doxorubicin (PLD) at 45 mg/m² intravenously (IV) on day 1 every 28 days, with ifosfamide given by continuous intravenous infusion (CIVI) at 1.5 g/m²/day for 6 days (total dose over 6 days of 9 g/m²), in conjunction with mesna 1.5 g/m²/day for 7 days [19]. In some early cases, free doxorubicin was used because of insurance restrictions at a dose of 9.3 mg/m²/day by continuous IV infusion over 7 days (total dose over 7 days of 65 mg/m²). For both the PLD and free doxorubicin regimens, granulocyte colony stimulating factor (G-CSF) was used prophylactically. FDG PET-CT was performed at baseline before chemotherapy and after cycle 4 immediately before wide surgical excision of the tumor as previously described [15]. External beam radiation was administered to patients after wide surgical excision, and patients were then followed for recurrence.

IHC was performed using standard techniques following the manufacturer’s instructions with antibodies to ALDH1 (BD Clone 44, BD Transduction Laboratories) and CD133 (clone 7, kindly provided by Dr. John Ohlfest, University of Minnesota). IHC for CD31, CD44, and CD68 was performed by a CLIA-certified laboratory (University Hospital, Minneapolis) using the antibodies JC70 (CD31) (Santa Cruz Biotechnology), DF1485 (CD44, Santa Cruz Biotechnology), and KP-1 (CD68) (Santa Cruz Biotechnology). Stained slides were reviewed by a pathologist with expertise in sarcomas (JDW). For CD31, staining was scored on a scale of 0–3 (reflecting vessel density). For CD68, staining was scored at 100× magnification as the percent of cells, including stroma, that were positive: 0–5%, > 5–25%, > 25–50%, > 50–75%, or > 75%. For CD133, staining was scored as the percent of tumor cells positive, on a scale of 0, 1–33% positive (score 1), 34–66% positive (score 2), or > 67% positive (score 3). Colon was used as a positive control for CD133 expression. For ALDH1 and CD44, staining was scored both as a percent of tumor cells positive on a scale of 0, 1–33% positive (score 1), 34–66% positive (score 2), or > 67% positive (score 3) and also for intensity of staining as 0, 1, 2, or 3.

Grading of percent necrosis was performed by assessing the percentage of viable tumor and percentage of non-viable tumor; the latter specifically addressed the extent of necrosis, fibrosis, hemorrhage, and cystic change, each in percentages. This assessment was obtained by correlating the gross features and the histologic sections. In addition, each case was assigned a specific histologic diagnosis and histologic grade according to the NCI system and the FNCLCC system. Histologic response percentage was defined as 100 − percentage of viable remaining tumor.

Staining of CD133, ALDH1, CD44, CD31, and CD68 was available in up to 30 pre-chemotherapy biopsy samples and up to 28 post-chemotherapy resection samples (Additional file 1: Table S1 and Additional file 2: Table S2).

CD133 expression was detected in only 2 of the 25 pre-chemotherapy biopsy samples (cases 10 and 19, Additional file 2: Table S2), with representative images shown in Fig. 1. In case 10 at baseline, 66–100% of tumor cells were positive at a score of 1, while 33% of the tumor cells had a score of 3; after chemotherapy 33% had a score of 1, but only in the cells that retained spindle characteristics.

ALDH1 staining was observed in 11 of the 30 pre-treatment samples and was generally similar in pre- and post-chemotherapy specimens (Additional file 2: Table S2). In 1 synovial sarcoma, only cells with epithelioid morphology were positive, while the spindle cell component was negative. ALDH1 expression in tumor cells ranged from 0 (Fig. 2B–D) to prominent staining (score 3). Of the 14 evaluable paired cases after chemotherapy, ALDH1 expression scores in tumor cells increased in 4, were unchanged in 7, and decreased in 3 (Additional file 2: Table S2). In addition to staining some tumor cells, ALDH1 was also found to be prominently expressed in macrophages, as has been reported [20‐22] (Fig. 2A).

CD44 staining of tumor cells was seen in 26 of 30 pre-chemotherapy samples. Tumor cell expression of CD44 ranged from no expression (Fig. 3A) to strong expression (Fig. 3C, D). CD44 expression increased in tumor cells after chemotherapy in 7 cases, was unchanged in 15 cases, and decreased in 2 cases of the 24 evaluable paired cases. CD44, like ALDH1, also stained macrophages, as expected [23, 24] and can be seen in Fig. 3A.

As described above, significant staining of macrophages for both CD44 and ALDH1 was observed, as has been reported [20, 22‐24], and macrophage infiltration of the tumors was more prominent than expected. We therefore performed IHC with CD68 to better quantify the degree of macrophage infiltration of tumor specimens. Some areas had more macrophages than others, and there were more macrophages in areas of tumor necrosis. We quantified CD68 positivity in both viable and necrotic regions of pre-chemotherapy biopsy specimens and post-chemotherapy resection specimens, and confirmed that the density of macrophages was greater, and often markedly so, in necrotic areas (not shown). Staining for CD68 was scored as the percent of all cells, including stromal cells and macrophages, that were positive. A representative image showing 25–50% staining is shown in Fig. 4. In general, necrotic regions contained 75–100% CD68 + macrophages in both the pre-and post-chemotherapy specimens. In some cases, nearly 100% of the remaining cells in necrotic areas after chemotherapy were macrophages. Since macrophages could be recruited to the tumor both by factors actively secreted by tumor cells, and also by the products of necrosis, it was not surprising to find many macrophages in necrotic regions. Subsequent detailed analyses of CD68-positive cells were performed only on viable regions of tumor.

Macrophages comprised from < 5% to, in some cases, 50–75% of cells present in viable regions of the tumor in the 29 paired pre-chemotherapy biopsy specimens. Only 4 of the 29 evaluable samples had 0–5% macrophages (2 synovial sarcoma, 1 leiomyosarcoma, and 1 malignant peripheral nerve sheath tumor), 9 had 5–25% macrophages (5 UPS, 2 synovial sarcoma, 1 fibrosarcoma, 1 spindle cell sarcoma), and 11 had 25–50% macrophages (7 UPS, 1 synovial sarcoma, 1 fibrosarcoma, 2 dedifferentiated liposarcoma), while 4 tumors were composed of 50–75% macrophages (4 UPS) in the pre-chemotherapy viable tumor (Additional file 2: Table S2). Thus, macrophage infiltration was most prominent in UPS. After treatment, the number of macrophages in viable tumor areas increased in some cases and decreased in others without a clear pattern (Additional file 2: Table S2). Macrophage density in pre- and post-chemotherapy tumor specimens was variable but prominent in many samples (Additional file 2: Table S2).

We also evaluated angiogenesis by IHC staining for the endothelial marker CD31. CD31 staining was detected in all 28 of the pre-treatment specimens and was generally similar in the 20 paired pre-and post-chemotherapy specimens evaluated (Additional file 2: Table S2). CD31 staining increased in 5 cases, remained the same in 14 cases, and decreased in 1 case (Additional file 2: Table S2). Representative staining for CD31, reflecting vessel density, is shown in Fig. 5.

Possible associations between IHC staining and outcomes from our prospective PET imaging study [15] were examined. Specifically, baseline and post-chemotherapy ALDH1 (% positive and intensity), CD44 (% positive and intensity), CD133, CD31, and CD68 were tested for correlation with baseline (pre-chemotherapy) SUVmax, post-chemotherapy SUVmax, % change in SUVmax, diagnosis (synovial vs other), survival, percent viable cells post-chemotherapy, and change in tumor size (Table 1). Correlations between CD31 and CD68 were also tested.

For CD133, there were too few non-zero values for statistical analysis. However, both cases that showed CD133 staining at baseline demonstrated excellent chemotherapy effect based on tumor cell viability.

ALDH1 did not correlate strongly with the measured outcomes, although some correlations were found (Table 1). For example, post-treatment ALDH1 staining intensity correlated positively with the change in tumor size; i.e., more intense ALDH1 staining correlated with tumor growth, and pre-chemotherapy ALDH1 staining intensity correlated weakly negatively with survival; specifically, those who died had a lower intensity of ALDH1 staining. The percent of pre-chemotherapy ALDH1-positive cells did not correlate with any endpoints in the analysis. The percent of post-treatment ALDH1-positive cells was weakly positively correlated with the percent of viable tumor cells and with change in tumor size. In other words, tumor growth tended to be higher in tumors with more ALDH1-positive cells.

The percent of pre-chemotherapy CD44-positive cells was positively correlated with post-chemotherapy SUVmax. The percent of post-treatment CD44-positive cells was positively correlated with baseline SUVmax and post-treatment SUVmax; and negatively correlated with tumor size. Synovial sarcomas had a lower percent of both pre-chemotherapy and post-treatment CD44-positive cells than the other sarcomas.

Pre-treatment CD44 staining intensity was positively correlated with baseline SUVmax and post-treatment SUVmax. Pre-chemotherapy synovial sarcomas had lower CD44 staining intensity than other sarcomas. Post-treatment CD44 staining intensity was positively correlated with baseline SUVmax and negatively correlated with change in tumor size.

Pre-chemotherapy CD68 staining correlated positively with baseline SUVmax, and negatively with the percent of viable tumor cells in post-chemotherapy resection samples. In other words, cases with more CD68+ cells at biopsy had fewer viable tumor cells at resection, suggesting a better response to chemotherapy. Synovial sarcomas had lower CD68 staining when compared with all other diagnoses.

Post-treatment CD68 staining correlated positively with baseline SUVmax and post-treatment SUVmax (pre-operative SUVmax), and correlated negatively with the percent of viable tumor cells in the resection specimen. Synovial sarcomas had lower CD68 staining than other sarcomas.

Baseline (pre-chemotherapy) CD31 staining was negatively correlated with post-chemotherapy SUVmax, and positively correlated with the percent change in SUVmax from baseline; in other words, a lower CD31 at baseline correlated with a smaller drop in SUV from pre-chemotherapy to post-chemotherapy. Baseline CD31 also correlated positively with change in tumor size (i.e., more tumor growth and higher baseline CD31). Post-treatment CD31 staining correlated positively with change in SUVmax and change in tumor size. Baseline CD31 did not correlate with baseline or post-chemotherapy CD68.

Correlations of the changes in expression of the markers examined between pre- and post-chemotherapy specimens are shown in Table 1. The post- minus pre-chemotherapy for ALDH1-positive cells was weakly positively correlated with change in tumor size, such that an increase in ALDH1 positive cells was associated with an increase in tumor size. The post- minus pre-chemotherapy increase in ALDH1 staining intensity was positively correlated with percent of viable tumor cells in the resection specimen and change in tumor size. Change in the percent of CD44-positive cells showed no correlation with measured outcomes. Post-minus pre-treatment increase in CD44 staining intensity was negatively correlated with post-treatment SUVmax, and was higher in synovial sarcomas than the other diagnoses. Post- minus pre-chemotherapy increase in CD31 staining was negatively correlated with the percent increase in SUVmax. Post– minus pre-treatment CD68 staining scores were lower in synovial sarcomas than other sarcomas. There was no correlation between the baseline or post-treatment CD31 and CD68.

Many characteristics of tumors are derived in varying degrees from infiltrating normal host cells, either stromal or blood-derived. Chronic inflammation has been linked to cancer development since the time of Virchow [48, 51], and tumors have been described as “wounds that do not heal” [52]; normal infiltrating inflammatory cells, including TAMs and even myofibroblasts, could contribute to tumor growth (reviewed in [18, 53‐55]). In fact, infiltration of tumors by macrophages, in some cases representing 50% of the cells in the tumor, was recognized by Virchow in the nineteenth century, terming them lymphoreticular cells [2, 51]. In some tumors, the number and/or distribution of TAMs correlated with prognosis [56]. While lymphocytes were sometimes seen in the samples, there was no clear correlation with macrophage density.

Macrophage infiltration of STS has been previously reported [57‐62], and CD68 expression has been reported to have prognostic significance in leiomyosarcoma [58, 59]; in addition, some leiomyosarcoma cells were shown to produce M-CSF in vitro. Prominent infiltration of many STS tumors was also recently reported in a phase II trial of PD-1 inhibition [62]. Thus, the prominent macrophage infiltrate of many sarcomas suggests a potential role for targeting tumor macrophages in some sarcomas, as has been suggested in some other tumors [18, 58‐61, 63]. In our study, both pre- treatment and post-treatment CD68 staining in viable tumor correlated positively with the baseline SUVmax and negatively with percent viable tumor cells at resection (after chemotherapy). Post-treatment CD68 staining also correlated weakly positively with the post-treatment SUVmax. While the negative correlation of post-treatment CD68 staining with percent viable tumor cells in the resection specimen could reflect a treatment effect, the negative correlation of the pre-treatment CD68 staining with percent viable tumor cells after chemotherapy further suggests that infiltrating macrophages may play an important role in STS biology.

The role of paracrine signaling in tumors is dramatically evident in giant cell tumor of bone where RANKL secretion by tumor cells recruits normal osteoclasts to the tumor from circulating precursors [64]. Elimination of osteoclasts by a monoclonal antibody to RANKL results in dramatic effects on the neoplastic tumor cells as well, with associated clinical benefit, presumably by eliminating factors produced by the osteoclasts that are beneficial for growth and maintenance of the tumor phenotype of the neoplastic cells [64‐66]. Analogous to these observations, interfering with signaling from CSF1 produced by the tumor cells in pigmented villonodular synovitis can markedly decrease macrophage infiltration and have striking beneficial effects [67‐69]. TAM infiltration in some other sarcomas could possibly affect tumor cell biology in a similar manner, and targeting macrophages has shown anti-tumor activity in a model of leiomyosarcoma [70‐72]. Thus, it is possible that infiltrating CD68 positive macrophages in some STS could produce factors that have a trophic effect on the tumor cells, and that chemotherapy effects on these macrophages could result in less production of tumor supporting factors. Another possibility could reflect primary damage to the tumor resulting in new antigens then recognized by the macrophages resulting in augmented immune destruction.

Similarly, TAMs and other stromal cells may be involved in a variety of processes critical to tumor development including immuno-suppression, angiogenesis, invasion, metastasis, response to chemotherapy and radiation therapy (reviewed in [73, 74]), and macrophage content has been correlated with prognosis in several cancers [73]. For example, TAMs could contribute to tumor growth by several mechanisms, including releasing factors that promote tumor cell growth and inhibiting the immune response to the tumor (reviewed in [54]).

Macrophages can secrete a variety of cytokines capable of paracrine signaling and potentially affecting tumor biology [75]; one study suggested that CSF-1 production with macrophage infiltration may promote neovascularization [72], and angiogenesis plays an important role in tumor growth [17]. In our study, CD31 staining was similar before and after chemotherapy in most cases evaluated. Pre-treatment CD31 staining was negatively correlated with post-chemotherapy SUVmax and weakly with percent viable tumor cells in the resection specimen, and it was positively correlated with change in tumor size (i.e., more tumor growth and higher baseline CD31).

In addition to the possible effects of inflammatory cells in the tumor microenvironment on tumor growth and angiogenesis, tumor macrophages may influence response to chemotherapy [56, 61, 76‐78]. Depending on the model, depletion of TAMs increased or decreased the efficacy of chemotherapy [73]. In fact, there is evidence that some of the anti-tumor effects of trabectedin may be related to a reduction in inflammatory mediator production by tumor cells and by reducing infiltrating inflammatory cells such as macrophages [76‐78].

The presence of large numbers of macrophages in some tumors also raises the question of their role in the findings of our study correlating changes in PET scan measurements after chemotherapy with outcome [15]. The positive correlation of both pre-and post-treatment CD68 staining with the pre- and post-treatment SUVmax, respectively, suggests that intra-tumor macrophages may contribute significantly to the observed SUVmax on PET imaging in some cases. An analogous case occurs in giant cell tumor of bone, where normal osteoclasts are recruited by RANKL secreted by tumor cells [64, 79]; osteoclasts contain large amounts of H+ transporting ATPase and may avidly take up FDG [80, 81]. PET activity falls rapidly following treatment with denosumab, an antibody to RANKL, which may reflect the change in energy use by osteoclasts rather than reflecting a change in activity of neoplastic cells [79]. Similarly, FDG uptake by tumor macrophages may play a role in changes in PET activity in some sarcomas and other solid tumors, and this may contribute to less correlation between PET findings and pathologic findings of tumor viability. This possibility is supported by the positive correlation between CD68 staining and SUVmax in both the pre-and post-chemotherapy specimens. Macrophage infiltration also may have a significant effect on tumor volume and thereby contribute to traditional measures of tumor size and response to therapy such as RECIST.

Our findings suggest that while CD133, ALDH1, and CD44 may be useful markers of CSC in epithelial tumors, their clinical utility in STS seems limited. CD133 was reported to be expressed in synovial sarcoma [36], where CD133 expression was found in 5/5 tumor samples and 3/3 cell lines with the percentage of cells expressing CD133 ranging from ~ 2 to ~ 20%. A recent study found that a CD133+ subpopulation of a synovial sarcoma cell line had several CSC properties, including self-renewal and resistance to chemotherapy [32]. CD133 expression has also been reported in other sarcomas [35, 37], and correlated with lung metastases and poor prognosis in osteosarcoma [82], and poor survival in embryonal rhabdomyosarcoma [11]. In the current study, no CD133 staining was observed in 5 synovial sarcomas. While only 2 STS cases in the current study stained for CD133, the changes in CD133 staining following chemotherapy were compatible with a decrease in “stemness” after treatment.

ALDH1 has been proposed to be a marker of both normal and cancer stem cells, and expression in breast cancer has been correlated with survival in some studies [6]. However, the utility of ALDH1 as a marker of CSCs is not clear in some solid tumors such as ovarian cancer [12‐14]. Liu et al. found ALDH1 expression in ~ 1% of cells in the synovial sarcoma cell line SW982 [32]. In the current study, ALDH1 staining of tumor cells was detected in 11 of 30 pre-treatment biopsy samples and was similar before and after chemotherapy in half of the cases, while in some cases staining was higher and in some lower. Overall, patients whose pre-treatment ALDH1 staining intensity was high had worse survival outcomes, and patients whose post-treatment ALDH1 staining intensity was high had increased tumor growth; however, these correlations were weak. Thus, ALDH1 expression does not seem to be a useful marker of CSC in STS.

CD44 has been described as a marker of CSC in some carcinomas [6, 8]. In our study, CD44 expression was detected in tumor cells in 26 of 30 pre-treatment samples and was similar before and after chemotherapy. As with ALDH1, CD44 does not appear to be a useful CSC marker in STS.