Loss of SATB2 Occurs More Frequently Than CDX2 Loss in Colorectal Carcinoma and Identifies Particularly Aggressive Cancers in High-Risk Subgroups

Simple Summary The immunohistochemical analysis of Special AT-rich sequence-binding protein 2 (SATB2) is increasingly being used to detect colorectal differentiation. Our study aimed to investigate SATB2 expression levels and the prognostic relevance of SATB2 loss in colorectal carcinoma (CRC), especially in comparison with CDX2, the standard marker of colorectal differentiation. We tested SATB2 expression in 1039 CRCs and identified SATB2 as a strong prognosticator in the overall cohort as well as in specific subcohorts, including high-risk subgroups. Compared to CDX2, SATB2 showed a higher prognostic power but was lost at a much higher frequency, generally rendering SATB2 as the less sensitive marker for colorectal differentiation compared to CDX2. Abstract Background: Special AT-rich sequence-binding protein 2 (SATB2) has emerged as an alternative immunohistochemical marker to CDX2 for colorectal differentiation. However, the distribution and prognostic relevance of SATB2 expression in colorectal carcinoma (CRC) have to be further elucidated. Methods: SATB2 expression was analysed in 1039 CRCs and correlated with clinicopathological and morphological factors, CDX2 expression as well as survival parameters within the overall cohort and in clinicopathological subgroups. Results: SATB2 loss was a strong prognosticator in univariate analyses of the overall cohort (p < 0.001 for all survival comparisons) and in numerous subcohorts including high-risk scenarios (UICC stage III/high tumour budding). SATB2 retained its prognostic relevance in multivariate analyses of these high-risk scenarios (e.g., UICC stage III: DSS: p = 0.007, HR: 1.95), but not in the overall cohort (DSS: p = 0.1, HR: 1.25). SATB2 loss was more frequent than CDX2 loss (22.2% vs. 10.2%, p < 0.001) and of higher prognostic relevance with only moderate overlap between SATB2/CDX2 expression groups. Conclusions: SATB2 loss is able to identify especially aggressive CRCs in high-risk subgroups. While SATB2 is the prognostically superior immunohistochemical parameter compared to CDX2 in univariate analyses, it appears to be the less sensitive marker for colorectal differentiation as it is lost more frequently.


Introduction
Considering that colorectal carcinoma (CRC) currently ranks among the three most common cancers in humans concerning incidence and mortality worldwide [1,2], further explorations on potentially relevant biomarkers are warranted in order to characterise these tumours precisely and improve prognostic predictions.
Special AT-rich sequence-binding protein 2 (SATB2), a transcription factor interacting with nuclear matrix attachment regions which is highly expressed in the non-neoplastic colorectal mucosa [3,4], attracted increasing scientific notice for the identification of the colorectal origin of cancers of unknown primary and of CRC metastases [5][6][7][8][9][10], delineating SATB2 as a valid addition to CDX2, which is still the most established immunohistochemical marker associated with colorectal differentiation [5,11].
Previous immunohistochemical assessments of SATB2 in CRC showed a general association of a diminished SATB2 expression with poorer survival characteristics and microsatellite status [12][13][14][15][16]. However, it remains unclear whether SATB2 is differently expressed within purely morphological (adenocarcinoma NOS vs. specific CRC subtypes, tumour budding subcategories (Bd1/2/3), WHO low-vs. high-grade carcinomas), immunohistochemical (CDX2 expression) and pTNM/UICC stage subgroups. Furthermore, it remains to be elucidated, how frequently the loss of SATB2 occurs in comparison to the loss of CDX2 and whether SATB2 can identify distinct prognostic groups within these colorectal cancer subsets.
To address these questions, we investigated SATB2 expression in a large cohort comprising 1039 CRCs and correlated the results with histomorphologic and immunohistochemical (CRC subtypes, tumour budding activity, WHO grade, CDX2 expression) as well as clinicopathological parameters (pTNM/UICC staging, tumour localisation) and explored the prognostic relevance of SATB2 expression in uni-and multivariate survival analyses in the overall cohort as well as in specific subgroups.

Study Population
Our study cohort comprised one thousand and thirty-nine CRC patients that underwent surgical resection between 1997 and 2019 at the University Hospital Klinikum rechts der Isar of the Technical University of Munich, Germany. Only patients with colorectal carcinomas were included in this study. Patients suffering from other colorectal tumours (e.g., neuroendocrine tumours, non-epithelial tumours, etc.), insufficient tissue on the constructed tissue microarray or incomplete clinicopathological/survival data were excluded. The original cohort  was recently extended with cases from 2019 (n = 36) [17]. Clinicopathological characteristics as well as survival data were extracted from hospital records or from the Munich Cancer Registry. Definitions of survival parameters, survival endpoints and general treatment modalities were defined as described previously [18,19]. The local ethic committee of the Technical University of Munich approved this study (reference number: 252/16 s).

Evaluation of SATB2 Expression and Clinicopathological Parameters
SATB2 expression was analysed by SATB2 immunohistochemistry in 1039 CRCs on a tissue microarray with two tumour-carrying cores from each tumour. We used an automated immunostainer (BOND RXm System, Leica Biosystems, Germany) for the immunohistochemical staining of SATB2 (EP281, Cellmarque 384R, Ready-To-Use, Cell Marque, USA), which is the standard SATB2 antibody used in daily clinicopathological routine and which has been used by various previous studies [5,11,15,20]. We stained 2 µm thick sections from our tissue microarray. Antigen retrieval was performed with Epitope Retrieval 1 after deparaffinisation, (Leica Biosystems, Germany; equivalent to citrate buffer pH 6) for 20 min and antibody binding was detected using a Polymer Refine Detection Kit (Leica Biosystems, Nußloch, Germany) without a postprimary antibody and haematoxylin counterstain. Naturally, pretested positive/negative control-tissues were stained in parallel. Two independent observers (MJ, MS (Maxime Schmitt)) that were blinded to clinicopathological parameters manually performed the evaluation of SATB2 expression.
We assessed the number of positive carcinoma cells for each individual patient. Counting a minimum of 500 tumour cells, the resulting cumulative percentage score for both cores was then assigned for each CRC (range: 0-100%). Only a nuclear staining of SATB2 was considered specific. SATB2 expression patterns (combined from both cores) were classified into three separate groups: diffuse, if the tumours either showed a complete expression or only a very focal loss in singular cells; heterogeneous, if areas with a complete loss of staining were observed; absent, if the tumours were completely negative. A strong staining intensity was defined as an intensity comparable to normal colonic mucosa, a still easily identifiable but slightly weaker staining was rated as medium. A barely visible staining intensity was classified as weak. Cases without any detectable staining were classified as absent.
SATB2-grouping results from the TMA were compared with SATB2 staining from 20 randomly selected full block slides, interobserver variability was probed in 150 cases that were assessed by the two observers in a blinded fashion.
SATB2 expression was correlated with clinicopathological variables including staging data and with Haematoxylin and Eosin-based histopathological parameters defined by the recent WHO classification (CRC subtypes: Adenocarcinoma NOS, Mucinous adenocarcinoma, Signet-ring carcinoma, Medullary carcinoma, Serrated adenocarcinoma, Micropapillary adenocarcinoma, Adenoma-like adenocarcinoma, Adenosquamous carcinoma, Carcinoma with sarcomatoid components, Undifferentiated carcinoma, MANEC/NEC; WHO grade: low grade, formerly G1/G2 vs. high grade, formerly G3 and tumour budding: Bd1 (0-4 Buds in 20×), Bd2 (5-9 Buds in 20×), Bd3 (≥10 Buds in 20×)). The parameters were available from a previous study on the same cohort regarding the distribution and the prevalence of the essential morphologic criteria given in the 2019 WHO classification of colorectal carcinoma, from which also the microsatellite status was extracted [17] (cohort details; Table 1). Furthermore, SATB2 expression was correlated with CDX2 expression, which was analysed in a previous study [18], where a similar methodology regarding the finding of an optimised cutoff for CDX2 expression groups was used [21]. The cases from 2019 that were recently added to the collective (n = 36) were classified regarding the aforementioned parameters (histomorphology, CDX2 expression) as described previously [17,18].

Statistics
Using SPSS version 26 (SPSS Institute, Chicago, IL, USA) statistical analyses were performed applying X 2 test as well as X 2 test for trends and Fisher's exact test. The Cutoff Finder, a publicly available biostatistical tool that represents a bundle of optimisation and visualisation methods for cutoff determination, was used to define the optimal cutoffs for SATB2 expression groups [21]. Where applicable, the Bonferroni method was used to correct for multiple testing. Univariate survival analyses were performed using the Kaplan-Meier method and significance of survival differences was tested by a log-rank test. The Cox proportional hazard model was used for multivariate analyses. All statistical tests were performed two-sided, p-values ≤ 0.05 were considered significant.

Distribution of SATB2 Expression and Biostatistical Generation of SATB2 Expression Groups
Most CRCs showed a diffuse SATB2 expression (61%, n = 639), a heterogeneous staining was noted for 340 cancers (33%), 60 tumours (6%) showed a complete absence of SATB2. A nuclear staining in ≥90% of tumour cells was observed in 65% (679/1039) of cases. In order to transform this continuous variable into dichotomous SATB2 groups (binary variable), we used the Cutoff Finder [21], a publicly available biostatistical tool for cutoff determination, to identify the best cutoff for SATB2 stratification. Following these initial statistical analyses, two SATB2-groups were formed: CRCs that showed an SATB2 expression above the 20th percentile (>70% tumour cells; n = 808, 78%) were categorised as SATB2-high, CRC on/below the 20th percentile (range:0-70% of tumour cells; n = 231, 22%) were categorised as SATB2-low/absent. Examples of the two SATB2 expression groups among certain CRC subtypes are given in Figure 1. SATB2-low/absent CRCs usually showed a reduced SATB2 staining intensity and a significantly higher rate of a heterogeneous/absent staining pattern (p < 0.001, details see Table S1). Only the number of positive tumour cells (regardless of staining pattern or intensity) were used to form the SATB2 expression groups. A comparison of the results of the SATB2-grouping with full block slides showed an excellent concordance with the results from the TMA (95%, p < 0.001, Kappa Cohens value: 0.88). Furthermore, an excellent interobserver variance was evident (p < 0.001, Kappa Cohens value: 0.95).  As illustrated in Figure 2 and depicted in detail in Table S2, SATB2-low/absent CRCs were significantly enriched in higher pT/pN/pM and combined UICC-stages, right-sided tumours, carcinomas with lymphatic and blood vessel invasion as well as in tumours with positive margins (p < 0.001, respectively). Compared to SATB2-high neoplasms, SATB2low/absent CRCs were significantly increased in CRCs with high (Bd3) tumour budding activity and in poorly differentiated carcinomas according to the WHO grade (p < 0.001, respectively). Furthermore, a low/absent SATB2 expression was significantly enriched in the mucinous, micropapillary, medullary and signet-ring CRC subtypes as well as in MANEC/NEC (p < 0.001). MSI-H CRCs were also associated with an absent or low SATB2 expression (p = 0.01).

Discussion
In this study, we investigated the expression of Special AT-rich sequence-binding protein 2 (SATB2) in one thousand and thirty-nine resected CRCs, correlated the results with histomorphologic parameters (CRC subtypes, tumour budding activity, WHO grade) [17], expression of CDX2 [18] as well as clinicopathological parameters (pTNM/UICC staging, microsatellite status, localisation) and analysed the prognostic relevance of SATB2 in the overall cohort and in specific subcohorts.
In recent years, SATB2 has gained increasing attention as a relatively specific marker of colorectal differentiation [23][24][25][26] and functional studies have revealed the tumoursuppressive properties of SATB2 in experimental settings [12,[27][28][29][30], demonstrating that SATB2 is a complexly regulated tumour suppressor that represses CRC progression by inhibiting the transcription of SNAIL, a master regulator of epithelial-mesenchymal transition. In our cohort of more than one thousand tumours, SATB2 low/absent CRCs were significantly associated with higher UICC stages and massively enriched in tumours with high-risk histomorphological features such as high tumour budding (Bd3) or poorly differentiated carcinomas according to the WHO grade, which is in line with the functional studies postulating the tumour suppressive properties of SATB2 [12,27,28]. Consistent with previous findings, SATB2-low/absent CRCs were also associated with strongly reduced survival parameters in univariate analyses (log-rank test) of the overall cohort. Interestingly, although SATB2 expression retained its statistical significance in multivariate analyses when UICC stage, resection status, MSI-status, age and gender were incorporated, this independent prognostic power vanished when we added the most common histomorphological parameters of CRC, tumour budding, WHO grade and the different CRC subtypes to our multivariate analyses. These findings argue towards the fact that the massive enrichment of SATB2 low expressing tumours in high-grade categories of these parameters probably washes out the strong prognostic effect of SATB2 that is present in univariate analyses.
Nevertheless, we wanted to know whether SATB2 can identify any prognostic subgroup that is not identified by either UICC staging or classical histomorphological parameters and which is also retained in multivariate analyses incorporating all factors.
Interestingly, SATB2 low/absent CRCs were associated with an especially aggressive disease course in CRCs with high tumour budding activity and in UICC stage III cancers and showed highly reduced survival times in these high-risk subgroups. In subsequent multivariate analyses incorporating all prognostic factors, SATB2 retained its prognostic relevance in both UICC stage III carcinomas and dissociative cancers with a high tumour budding activity. These findings delineate SATB2 as a worthwhile immunohistochemical biomarker in CRC that can identify especially aggressive cancers in these high-risk subgroups of CRC and delivers valuable additional prognostic information in addition to standard histomorphological factors and UICC staging. Regarding the translation of these results into clinicopathological routine, we suggest that the responsible pathologist reports SATB2 loss by stating the percentage of SATB2 expressing cancer cells, ideally combined with the notion that a significantly reduced expression of SATB2 has been associated with a poorer clinical outcome.
As SATB2 has emerged as a considerable alternative to CDX2 to verify or rule out colorectal differentiation, another aim of our study was to compare the incidence and overlap of SATB2 loss with loss of CDX2 and also to compare the prognostic relevance of these two markers with one another [18]. The findings of these aspects of our study are also particularly interesting, because they highlight possible strengths and weaknesses of the most commonly used colorectal markers. Notably, the frequency of a reduced or completely lost SATB2 expression is much higher compared to CDX2 and especially a completely absent SATB2 expression was far more frequent than a complete negativity for CDX2. This implies, that when both CDX2 and SATB2 (as singular markers) are assessed regarding their ability to detect a colorectal origin in neoplastic tissues, SATB2 has to be ranked as the less sensitive marker compared to CDX2. However, as only two out of 1039 CRCs (0.2%) showed a complete loss of both SATB2 and CDX2, and the majority of SATB2 negative CRCs showed a strong and diffuse expression of CDX2 (and vice versa), a combined panel of both markers appears to be able to identify the overwhelming majority of colorectal cancers and is probably the most expedient approach for routine diagnostic settings.
When we then moved on to compare the prognostic impact of SATB2 and CDX2, we observed an opposing picture. Compared to CDX2, the loss of SATB2 showed a considerably higher prognostic impact in univariate analyses (log-rank test) of the overall cohort and in nearly all clinicopathological subscenarios of CRC, in which CDX2 mostly showed at best minimal prognostic impact in our cohort. A crucial difference between the prognostic difference of SATB2 and CDX2 was also that SATB2 retained its prognostic power in right-and left-sided CRCs, while CDX2 did not show any prognostic significance in right-sided tumours although it is more frequently lost in the right colon [18]. In line with these findings, we additionally observed that SATB2-low/absent CRCs were able to identify patients with a poor prognosis in both CDX2 expression groups (CDX2-low/absent vs. CDX2-high), while CDX2 showed no prognostic relevance in SATB2 expression subgroups, rendering SATB2 as the prognostically superior immunohistochemical biomarker in CRC compared to CDX2.

Conclusions
Our study has five major findings: (1) a low/absent SATB2 expression is significantly enriched in advanced stage CRCs that have an aggressive histomorphological phenotype with high tumour budding activity and/or a poor differentiation according to the WHO grade. (2) Loss of SATB2 is of high prognostic relevance in uni-and multivariate analyses (including UICC stage) in the overall cohort, but shows no independent prognostic value in the overall cohort when the main histomorphological parameters of CRC (tumour budding, WHO grade, CRC subtypes) are added to the multivariate analyses. (3) SATB2 shows an especially high prognostic relevance in uni-and multivariate analyses of high-risk clinicopathological subgroups (high tumour budding/UICC stage III) and identifies CRCs with a particularly aggressive disease course in these high-risk scenarios. (4) SATB2 loss occurs much more frequently than loss of CDX2, with a substantial portion of SATB2negative CRCs showing a diffuse or at least heterogeneous CDX2 positivity, generally delineating CDX2 as the more sensitive marker of colorectal differentiation in carcinomas. (5) SATB2, in general, showed a vastly better prediction of survival outcome compared to CDX2, with SATB2 retaining its prognostic impact in CDX2 expression subgroups (CDX2 low/absent vs. high), rendering SATB2 as the superior prognostic biomarker compared to CDX2.
In conclusion, our study identifies SATB2 as a potentially valuable additional prognostic biomarker in CRC. Further studies are warranted to explore the possible therapeutic implications of a diminished or completely lost SATB2 expression. Both SATB2 and CDX2 can individually be completely lost in CRCs, while a total absence of both markers is almost never observed. Therefore, a combined panel of both markers appears to be the most solid approach to pinpoint or rule out colorectal differentiation.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/cancers13246177/s1. Figure S1: Prognostic relevance of SATB2 expression for diseasespecific survival in right-sided (A, coecum until splenic flexure) and left-sided (descending colon until rectum) CRCs, Table S1: Correlation between SATB2 expression groups with SATB2 staining pattern (A) and SATB2 staining intensity (B), Table S2: Distribution of SATB2 expression groups with clinicopathological and morphological parameters in the overall cohort, Table S3: Impact of SATB2 expression on overall, disease-specific and disease-free survival in tumour budding-(Bd1, Bd2, Bd3), WHO grade (low, high), microsatellite (MSS, MSI-H) and CDX2 expression subcohorts (low/absent, high), Table S4: Impact of SATB2 expression on overall, disease-specific and disease-free survival in the different UICC stage groups, Table S5: Multivariate overall survival analysis in the overall cohort under inclusion of SATB2 expression, age, gender, CRC subtype, tumour budding, WHO grade and microsatellite status, Table S6: Multivariate overall survival analysis in the overall cohort under exclusion of CRC subtype, tumour budding and WHO grade.

Informed Consent Statement:
Patients signed a general informed consent protocol during admission to the hospital. Data Availability Statement: All relevant data are within the paper and its supporting information files. The data underlying the results presented in the study are available from the study group upon reasonable request, some restrictions apply due to confidentiality of patient data. Since these data are derived from a research trial with ongoing follow up there are legal and ethical restrictions to share sensitive patient related data publicly. Data can be requested in the context of a translational research project by sending a request to the corresponding author.