Correlation of microscopic tumor extension with tumor microenvironment in esophageal cancer patients

With ever-increasing precision of radiation modalities, i.e., particle therapy, and improved imaging enabling adaptive radiotherapy, profound understanding of the gross tumor volume (GTV) and the clinical target volume (CTV), which includes the GTV and microscopic tumor extension (MTE), prior to and during radiotherapy (RT) is crucial. This holds true for many solid tumors including esophageal cancer (EC), in which large CTVs have historically been used. These large volumes are based on previous histological assessments of the MTE in EC and calculations of the CTV margins required to encompass 95% of the MTE, which ranges from 6 to 24 mm around the GTV in EC [1‐4]. However, these studies neither investigated the possible influence of the tumor microenvironment (TME) on MTE, nor the exact information on the in vivo localization of the specimens in the patients.

Factors of the TME, such as kinases, cancer stem cells, hypoxia, tumor cell proliferation, and immune cells, are hypothesized to be involved in MTE [5, 6]. FAK and ILK, known to be overexpressed in EC, are strongly linked to tumor cell survival, proliferation, invasion, and migration [7‐9]. The cancer stem cell marker CD44 characterizes a tumor subpopulation highly associated with tumor metastasis and radioresistance in EC patients, ultimately leading to tumor recurrence [10, 11]. HIF-1α, an important biomarker of the hypoxic TME, contributes significantly to the survival of tumor cells and to resistance to radiochemotherapy (RCHT) [12, 13]. Lastly, Ki67 is a marker of cell proliferation and a well-known prognostic marker in a variety of solid cancers, including EC [14].

In a previous retrospective cohort of EC patients, we compared changes in the TME in patients with esophageal adenocarcinoma (AC) or squamous cell carcinoma (SCC) treated with neoadjuvant RCHT followed by resection (NRCHT+R) with those in patients who underwent resection only (R) [15]. However, these findings could not be directly correlated with MTE since the study was based on pre-existing paraffin-embedded tumor blocks, and their in vivo localization could not be reconstructed. For proton beam therapy of patients with EC, our department has introduced use of fiducial markers, which are endoscopically implanted at the cranial and caudal borders of the macroscopic tumor prior to RT planning [16]. Besides increased treatment precision, these fiducial markers, which remain in situ in the tumor resection specimen, hold information on the position of the former GTV in the paraffin-embedded tissue blocks and can be back-projected to the pretreatment computed tomography (CT) scans. Consequently, both the precise position of the tissue blocks and the potential postoperative tissue shrinkage impacting the MTE can be estimated.

Therefore, this study aimed at assessing various markers of the TME in a cohort of EC patients having undergone NRCHT+R and whose resection specimen contained implanted fiducial markers, and to correlate the possible MTE with the marker expressions.

The clinicopathological characteristics of the patients are outlined in Supplementary Table 1. Since there was only one patient with AC in this cohort, no comparative analyses of the markers of the TME in tumor nests or tumor stroma between the two tumor histologies were performed.

In tumor resection specimens of EC patients, the overall percentages of FAK⁺, CD44⁺, HIF-1α⁺, and Ki67⁺ cells were higher in tumor nests than in the tumor stroma, with Ki67⁺ cells reaching statistical significance (p < 0.001; Fig. 2). Conversely, expression of ILK⁺ cells was higher in tumor stroma compared to tumor nests, albeit not statistically significantly. In line with this, a heatmap illustrating the co-expression of markers of the TME showed more overall co-expression in tumor nests compared to tumor stroma (Fig. 3).

Of interest, MTE beyond the fiducial markers was found in three patients (all cT3N1): two SCC patients (P1 and P2) showed MTE of 14 and 15 mm, and one AC patient (P3) still had MTE of 31 and 17 mm in the cranial and caudal directions, respectively (Fig. 4). In those patients, a stronger expression of FAK⁺, CD44⁺, and ILK⁺ cells was found in tumor nests between the fiducial markers (former GTV) and beyond (former CTV; Fig. 5), even after NRCHT. The expression of TME markers of these three patients in the former GTV versus the former CTV are shown for tumor nest and tumor stroma in Supplementary Fig. 3. Also, the expression of the assessed markers was compared between the three patients with MTE and the seven patients without MTE. However, no significant differences were found, neither for the tissue type (tumor nest, tumor stroma) nor upon combining the markers or using them separately (Fig. 6). Of note, a co-expression analysis of Ki67 with all the markers was done to determine the proliferation status of the selected TME markers even after NRCHT.

Overall, the results of this study showed higher expression of cells positive for the selected TME biomarkers (FAK⁺, CD44⁺, HIF-1α⁺, and Ki67⁺) within tumor nests compared to tumor stroma in tumor resection specimens of EC patients after NRCHT+R. Conversely, ILK-positive cells were highly expressed in the tumor stroma compared to the tumor nests. However, the marker expressions in the resection specimen of the three patients with residual MTE after treatment did not differ from those of patients without MTE, neither in the tumor stroma compared to the tumor nests, nor when combining marker expression.

To the best of our knowledge, this is the first study to thoroughly assess markers of the TME after NRCHT+R in EC patients and compare them to MTE. Some of our findings are in line with reports on the expression of markers of the TME in previous studies performed on pretreatment specimens in a variety of solid tumors, including EC, while others differ, as discussed below.

Previous studies using immunohistochemical analyses on tissues samples of patients with EC, colon cancer, breast cancer, and sarcomas compared the expression of FAK proteins in tumor compartments with normal tissues of either the same patients or between different patient cohorts [7, 29, 30]. These investigations showed higher expression of FAK in the cytoplasm of tumor cells and in the invasive front of tumor nests compared to the tumor stroma. Furthermore, higher expression of FAK was associated with tumor infiltration depth, lymph node metastases, and advanced disease stage, leading to poor prognosis and lower survival rates. Similar investigations comparing tumor and non-tumor tissue associated CD44 expression with tumor invasion, metastasis, progression, treatment resistance, and poor overall survival [31‐33]. Many investigations on HIF-1α involvement in a hypoxic TME concentrated on its association with metastasis and biological response to treatment. Two studies using immunohistochemistry examined HIF-1α expression from resection tissues of ESCC patients who underwent RCHT [34, 35]. Their findings showed that high HIF-1α expression is significantly correlated with a poor degree of differentiation, lymph node metastases, and depth of invasion. Additionally, three studies assessed tissues samples of ESCC patients who underwent surgical resection using immunohistochemistry [36‐38]. Thus, it was shown that high HIF-1α expression is associated with regional and distant tumor cell metastases, advanced tumor stage, depth of infiltration, positive surgical margins, and poor overall survival. Hu et al. [39] used immunohistochemistry to compare tumor tissues with adjacent normal tissues in patients with ESCC who underwent surgery. They revealed that HIF-1α expression was higher in cancer tissues than in surrounding normal tissues, and that higher HIF-1α expression was associated with tumor cell motility, recurrence, and poor prognosis. A few studies investigated the value of Ki-67 in predicting response in ESCC patients following NRCHT [14, 40‐42]. These investigations found a link between a high Ki-67 index and a poorer postoperative survival rate. They indicated that elevated Ki67 expression in tumor tissues is a risk factor for tumor progression and poor survival in ESCC patients. Finally, contrary to our results, two immunohistochemical studies comparing tumor tissues with adjacent normal tissues of EC patients after resection demonstrated an overexpression of ILK in differentiated areas of tumor tissues as compared to adjacent normal tissues [9, 43].

Our findings revealed MTE of up to 31 and 17 mm in the cranial and caudal directions from the fiducial markers (GTV borders) after completion of NRCHT+R. However, both the treatment regimen and analysis method of our cohort differ from previous studies that investigated MTE in resection specimens of EC patients in order to estimate CTV margins for RT. Gao et al. [4] used H&E staining to examine MTE in surgical specimens of SCC or AC patients with pT1-4 EC treated by primary resection. SCC had a mean MTE beyond the GTV of 10.5 ± 13.5 mm and 10.6 ± 8.1 mm in the cranial and caudal directions, respectively, whereas AC had cranial and caudal MTE of 10.3 ± 7.2 mm and 18.3 ± 16.3 mm, respectively. In a further analysis, they found a statistically significant correlation between the extent of MTE and the pathological tumor stage. Based on these findings, they suggested that a craniocaudal CTV margin of 30 mm for SCC, and 30 mm cranial and 50 mm caudal for AC is adequate to cover MTE within the esophagus in 94% of patients. In a similar study, Song et al. [44] used H&E staining to estimate CTV margins in patients with pT2‑3 SCC of the esophagus having undergone esophagectomy. They reported MTE of 30–40 mm in either direction. To obtain 95% coverage of MTE, they advised CTV margins of 30 and 40 mm in the cranial and caudal directions from the GTV. Furthermore, two investigations used H&E staining to examine MTE in resection specimens of EC patients with pT1‑4 treated with surgical resection and reported cranial and caudal distances of 34 ± 22 mm and 40 ± 34 mm, respectively [45, 46]. In these studies, the MTE was not linked with the CVT margins.

Our study has some limitations, which may influence its findings. First, the study was based on a small sample size, due to the limited number of EC patients amenable to proton beam therapy, and to the labor-intense acquisition and processing of the resection specimen. Second, in some patients, only one fiducial marker could be implanted prior to treatment (stenosing tumor), or one of the two implanted markers was lost during NRCHT. Thus, fiducial markers were absent on pre- or posttreatment imaging and/or in the resection specimen, and those patients had to be excluded. The third limitation is the absence of a complementary cohort of patients who underwent resection alone. This cohort of resection alone would be important, since the length of MTE may be underreported in the studied cohort having undergone NRCHT+R. Finally, implantation of the fiducial marker used here is burdensome, such that replacement by a liquid fiducial marker is envisioned.

Our findings indicate that in EC patients having undergone NRCHT+R, the overall expression of FAK, HIF-1α, Ki67, and CD44 was higher in tumor nests, whereas ILK was higher in tumor stroma. Differences in the TME between patients with residual tumor cells in the original CTV compared to those without were not found. We have successfully established a panel of TME markers and their link to MTE. However, there is insufficient evidence that the TME influences the required CTV margin on an individual patient basis.