Prognostic impact of CD34 and SMA in cancer‐associated fibroblasts in stage I–III NSCLC

Abstract Background Epithelial‐to‐mesenchymal transition (EMT) is a crucial step in lung cancer pathogenesis. Among others, cancer‐associated fibroblasts (CAFs) are reported to regulate this process. Objectives To investigate the prognostic and clinical impact, we analyzed CD34+ and SMA+ CAFs in non‐small cell lung cancer (NSCLC). Methods Retrospectively, immunohistochemistry was performed to study stromal protein expression of both CD34 and SMA in 304 NSCLC patients with pTNM stage I‐III disease. All tissue samples were embedded on tissue microarrays (TMAs). Results Our analysis revealed an association for CD34+ CAFs with G1/2 tumors and adenocarcinoma histology. Moreover CD34+ CAFs were identified as an independent prognostic factor (both for progression free survival [PFS] and overall survival [OS] in stage I‐III NSCLC). Besides, SMA+ expression correlated with higher pTNM‐tumor stages and lymphatic spread (pN stage). In turn, SMA‐negativity was associated with improved PFS, but no prognostic impact was found on OS. Of interest, neither CD34+ CAFs nor SMA+ CAFs were associated with the primary tumor size, localization and depth of infiltration (pT stage). Conclusions CD34 was identified as an independent prognostic marker in pTNM stage I‐III NSCLC. Moreover, loss of CD34+ CAFs might influence the dedifferentiation of the NSCLC tumor from its cell origin. Finally, SMA+ CAFs are more prevalent in NSCLC tumors of higher stages and lymphonodal positive NSCLC. Key points Expression of CD34 on cancer associated fibroblasts (CAFs) is an independent prognostic factor in stage I‐III NSCLC. SMA+ cancer associated fibroblasts are associated with higher tumor stages in NSCLC and might contribute to tumor progression in NSCLC.


Introduction
Today, lung cancer is still one of the most common and most lethal cancer types worldwide with rising incidence rates. 1,2 With a focus on histology, two major subtypes must be distinguished from each other: While non-small cell lung cancer (NSCLC) is the more frequent subtype, small cell lung cancer (SCLC) is considered to be the most aggressive. 1,3 NSCLC comprises over 80% of the cases with a rising incidence of adenocarcinoma (ADC) and decreasing levels of squamous cell (SCC) and large cell carcinoma (LCC). 4 Therapeutically, besides cytotoxic chemotherapies 5 and molecular targeted therapies (eg, EGFR-, 6 ALK-and ROS-mutation antagonists 7 or anti-angiogenic antibodies against VEGF 8,9 ), immune-checkpoint-inhibition [10][11][12][13][14][15][16] has recently augmented our therapeutic armamentarium for locally advanced and metastatic NSCLC.
When addressing stromal tumor tissue, three relevant factors need further consideration. Besides tumor infiltrating cells, such as macrophages, T-lymphocytes, natural killer cells and dendritic cells, 17 tumor microenvironment predominantly comprises tumor induced angiogenesis 18 as well as cancer-associated fibrocytes and fibroblasts (CAFs). 19 Antivascular therapies and tumor infiltrating cytotoxic T-lymphocytes have already been reported to increase the therapeutic efficacy against lung cancer. However, analysis of stroma in NSCLC is challenging due to its cell-heterogeneity and complex cell-cell interactions. [20][21][22] Metastatic conversion, a crucial step in cancer progression, is driven by epithelial-to-mesenchymal transition (EMT). EMT results in loss of cell polarization, reduction of cell-celljunctions and in turn gain of motility and invasiveness. 23 Upon activation via transforming growth factor-β (TGFβ), receptor tyrosine kinases (RTK) or the Wnt-pathway, Snail and Zinc-finger proteins downregulate the transcription of essentially involved proteins, such as E-cadherin. 23 With a focus on pathogenesis, in vitro TGFβ was able to transform clusters of differentiation 34 (CD34)+ fibrocytes into alphasmooth muscle actin (SMA)+ fibrocytes. 24 Moreover, in contrast to regular fibroblasts, in vitro cell culture analyses of SMA+ CAFs revealed SMA+ CAFs induced EMT in different NSCLC cell lines. 25 In other cancer types, such as pancreatic cancer, 26 cervical cancer, 27,28 oral squamous cell cancer 29 and cholangiocarcinoma 30 as well as colorectal cancer 31 and thyroid papillary carcinoma 32 stromal CD34+ and SMA+ CAFs could be associated with distinct malignant features such as lymphonodal metastasis. So far, these analyses have not been performed for NSCLC. CD34 was first detected by Civin et al. in 1984 and was subsequently characterized as a hematopoietic progenitor cell marker. 33 Recently, increased expression of CD34 and CD31 has been reported on vessel endothelium and endothelial progenitor cells, especially in tumor vascularization. 34 Here, monoclonal antibodies against CD34 and CD31 have been used to analyze the microvessel density (MVD) in central and peripheral regions of NSCLC tissue. Among others, Ushijima et al. considered high peripheral MVD to correlate with more advanced tumor stages and a higher probability of nodal metastasis. 35 Nevertheless, CD34+ fibrocytes and fibroblasts have been described in wound healing 36 and, focusing on the respiratory system, especially in fibrotic remodeling in asthma, 37 chronic obstructive airway disease, emphysema 38 and possibly in idiopathic pulmonary fibrosis. 39 SMA is a marker of CAFs 40,41 and promotes exosomal cancer proliferation in NSCLC, 42 and was first described by Gabbiani et al. in 1975. 43 Since then, SMA has been used to analyze stromal features in different tumor entities, such as pancreas carcinoma, 26 ductal adenocarcinoma of the mammary gland, 44 cervical cancer, 28 head-and neck squamous cell carcinoma, 45 colon carcinoma, 31 cholangiocarcinoma 30 and squamous cell carcinoma of the lung. 46 In conclusion, SMA+ CAFs can more frequently be found in malignant tissue than in benign alterations of the respective tissue origin and, in addition, SMA+ CAFs are more abundant in the primary tumor and in macrometastases compared to micrometastases and tumor cell-poor lymphatic vessel spread. 30,46 Against this background, we investigated the impact of CAFs in NSCLC. We focused on CD34 and SMA as immunohistochemical biomarkers to evaluate the clinical and prognostic influence on stromal cells in stage I-III NSCLC.

Study population
Retrospectively, we studied the expression of CD34 and SMA in patients suffering from non-small cell lung cancer (NSCLC). All investigated patients were surgically treated in the Department of Thoracic Surgery, Niels-Stensen-Kliniken, Ostercappeln (Germany) during December 1998 and November 2004 and had stage I to stage III tumors. We identified n = 379 patients. Due to tissue loss on TMAs and uncertain primary histology, tissue samples from n = 304 were available for evaluation. Since the patients were treated between December 1998 to November 2004, Tumor Nodule Metastasis (TNM)-classification based on the proposed Union Internationale Contre le Cancer (UICC) sixth edition 47 was applied.

Immunohistochemistry
All surgically resected tissue samples were analyzed using 4 μm thick formalin-fixed paraffin-embedded (FFPE) tissue microarrays (TMA). Each NSCLC patient was represented by three punch cores. 48 Core positions were chosen from the original hematoxylin-eosin stained diagnostic material and transferred to the paraffin embedded tissue specimen. Each core was chosen to represent an adequate amount of representative tumor and stromal tissue. Immunohistochemistry was performed via peroxidase-conjugated avidinbiotin method. Primary antibodies included Roche/Ventana QBEnd/10 (CD34) and DAKO HHF35 (SMA). Briefly, after deparaffinization of TMA samples via xylene, rehydration was performed by application of graduated ethanolsolutions at indoor temperature. Primary antibodies were incubated for 30 minutes at indoor temperature. Washing was then followed by incubation of the preparated sections with biotinylated secondary antibodies. Here, 3-amino-9-ethylcarbazole was used as a substrate, and immunoreactions were detected (Ventana Optiview DAB IHC detection KIT, Germany). Healthy lung tissue served as a negative control, whereas positive controls were performed on tonsil tissue (CD34) and appendix vermicularis tissue (SMA). All samples were analyzed manually using an Olympus BX51 microscope at 200x magnification by ABS, LHS, BH and JR, respectively. The proportional amount of CD34+ and SMA+ cells from overall tumor stroma was recorded. To count CD34+ and SMA+ fibrocytes and fibroblasts, CD34+ and SMA+ endothelial cells were visually excluded from positivity in tumor stroma (c.f. Fig 1). Before statistical evaluation, the results were gathered and compared. Deviations were discussed interdisciplinary. PD-1 and PD-L1 stain in this collective were performed beforehand. 49

Statistical analysis
To investigate the NSCLC cohort, we used mean and standard deviation (SD), median and interquartile range (IQR; Q1-Q3), as well as raw count and frequencies. Bivariate associations between categorical variables were analyzed via Fisher's exact test or Chi-square test, if applicable. Continuous and ordinal variables were tested using either unpaired t-test or Mann-Whitney-U test depending on the normality of the data. Associations of more than two groups (eg, histology variable) with categorical and continuous outcomes were analyzed using Chi-square test or Kruskal-Wallis test. The area under the receiver operating characteristic curve (ROC) was used to determine cutoff values for CD34 and SMA-stain via Youden Index regarding median overall survival.
Overall survival was defined as time (days) between histopathological diagnosis and death or censoring. Progression free survival was defined as time (days) between histopathological diagnosis and first relapse, progress after initial treatment, death or censoring, depending on the first chronological appearance. In univariate analyses, survival time was compared between groups using Log rank tests and visualized by Kaplan Meier plots. Multivariate survival analyses were performed via Cox proportional hazards model. Forward variable selection was applied using likelihood ratio tests. Inclusion criterion was set to 0.05. Hazard ratios (HR) are presented with 95% confidence intervals (95% CI). Data collection as well
To rule out the possibility that neoadjuvant treatment might have altered stromal CD34 and SMA levels we performed correlative analyses via contingency tables and Fisher's exact test. Neither CD34-positivity nor SMApositivity was associated with an applied neoadjuvant treatment regimen (P = 0.308 and P = 0.147). Yet, positive PD-L1 expression was associated with stromal SMA-negativity (P = 0.008) but not with CD34-positivity (P = 0.894). Although there was a significant association with tumor stage (P < 0.001), we focused on the impact of adjuvant therapies in the light of CD34-positive CAFs and SMApositive CAFs in stromal tumor tissue. Here, only surgicallytreated patients revealed SMA-dependent outcomes in PFS (P = 0.038) and OS (P = 0.038), whereas for the smaller subgroup of n = 66 adjuvant treated patients statistical significance could not be shown (PFS and OS both P > 0.05, data not shown). For cases with CD34+ CAFs, PFS was superior in pure surgically treated patients (P = 0.05) and adjuvant treated patients (P = 0.048). However, these data could not be reproduced for OS (both P > 0.05).

Discussion
Cytotoxic chemotherapies are directed towards rapidly dividing cells and therefore primarily cause damage to malignant tissue. In solid tumors the microenvironment plays a crucial role in cancer development and progression. Consequently, targeted therapies against the microenvironment might enhance therapeutic efficacy. 50 Here, we analyzed the effect of CD34+ and SMA+ CAFs on prognosis in NSCLC.
CD34 expression is a marker for fibrocytes and fibroblasts as well as neovascular endothelial cells, accounting for the evaluation of tumors' microvessel density (MVD). Evaluating MVD, high CD34+ levels are associated with greater tumor vascularization, larger tumors and a higher proportion of lymphonodal and distant metastatic spread. 35,46,51,52 In contrast, high CD34 levels in EMTpartaking fibroblasts and fibrocytes were considered to be predictors of healthy tissue and benign tissue alterations in pancreatic tissue, 26 head-and neck mucosa tissue, 45 ductal gland tissue of the breast 44 as well as tissue of the uterine cervix. 27 In particular, the loss of CD34+ fibrocytes was associated with malignancy. 26,27,44 Interestingly, in the present study we were able to correlate the grade of dedifferentiation of malignant NSCLC tissue to the proportional appearance of CD34+ fibroblasts. Whereas low graded tumors (G1) were all CD34+, lower stromal CD34+ fibroblast expression correlated with higher grading (G3 40.4% and G4 15.8% positivity, respectively). Moreover, stromal CD34+ CAF count was significantly associated with PFS and OS independently of age, sex, TNM-stage, histology, or tumor grading in stage I-III NSCLC patients. Furthermore, CD34+ CAFs occurred more often in pulmonary adenocarcinoma (ADC) than in squamous-cell carcinoma (SCC) or large cell carcinoma (LCC) of the lung (P = 0.006). However, tumor grading was significantly lower for ADC than for SCC or LCC (P < 0.001). In this context, further investigations are needed to determine if the predominant histological growth pattern or the degree of dedifferentiation is the major cause of altered stromal CD34+ fibroblast levels. SMA+ CAFs have already been identified as a predictor for malignancy and higher tumor stages in various tumor entities such as pancreas, 26 head-and neck squamous cell carcinoma, 45 ductal adenocarcinoma of the breast, 44 colon, 31 cholangiocarcinoma 30 or squamous cell NSCLC. 46 Interestingly, the latter two studies reported high levels of SMA+ CAFs in both the primary tumor and lymphonodal macrometastases, whereas SMA levels showed low expression in lymphonodal micrometastases. 30,46 In our study, only primary tumor's tissue was evaluated towards the CAF infiltration, whereas further analyses on lymph node tissue were not performed. Yet, we were able to demonstrate, that in the primary lesion of the tumor, SMA+ CAFs were associated with higher TNM stages (stage I 29%, stage III 44% positivity) and lymphonodal spread (pN0 29%, pN2 42% positivity). However, in our analysis the absence of SMA+ CAFs positively influenced PFS of NSCLC patients, but not OS, and hence was not found to be an independent prognostic marker. Interestingly, tumoral positive PD-L1 stain 49 was associated with SMA+ CAFs. Further investigations are needed to understand the possible immune-modulatory effect 26 of SMA+ CAFs on tumor cells.
While the cohort size of n = 304 cases is relatively large, the present study shows limitations. On the one hand, the cohort was sampled 15 years ago, resulting in a partly outdated TNM staging system. Current data security measures and ethical standards prevent an update towards the current eighth IASLC TNM staging system. Furthermore, useful data on EGFR-, KRAS-, ALK-and ROS-mutational status of the adenocarcinoma subcohort is not present and fresh frozen tissue is no more attainable. Therefore tumor mutational analysis cannot be subsequently performed. On the other hand, our analysis provides insight into patients treated before the era of antiangiogenic treatment, specific mutation targeting or even immune-checkpoint inhibition. Despite the awareness, that neoadjuvant treatment might change tumor biology and behavior, the exclusion of those cases would have led to limited explanatory power, especially regarding the associations of SMA+ CAFs and TNMas well as pN-stage. Hence, contingency tables were used to show an independence of stromal cell expression patterns from a therapy induced before surgical treatment and need careful consideration and validation.
It has previously been shown that CD34+ fibroblasts are preferentially expressed in healthy tissue and benign tissue alterations, 26,28,44,45 unlike SMA+ fibroblasts, that were associated with a higher probability of malignant venous infiltration in the study by Nishishita et al. 31 and showed higher presence in primary tumors and macrometastases. 30,46 Here, we were able to demonstrate that the supposed grade of invasiveness, 45 characterized by the TNM tumor stage and its lymphonodal spread, is related to the level of SMA+ CAFs in stage I-III NSCLC. As already mentioned, these findings need further investigation in stage IV NSCLC and stage II-III tissue without neoadjuvant applied systemic or local treatment measure. Moreover, our study revealed that CD34+ fibroblasts are not exclusively present in healthy tissue and benign tissue alterations, but may also occur in malignant NSCLC tissue. Yet, CD34+ CAFs are inversely associated with G3/4 differentiated tumors. When focusing on the independent predictive effect of CD34+ CAFs in stage I-III NSCLC, whether CAFs influence EMT along with lymphatic and metastatic spread, as proposed in vitro by Kim et al. 25 should be investigated, In summary, the findings are consistent with available data for CAFs in malignant tissues [26][27][28]30,31,[44][45][46] and contribute to our understanding of CAFs in NSCLC prognosis.