Until end of follow-up 31 December 2008, 626 incident cases of CRC had been registered in the prospective, population-based cohort study Malmö Diet and Cancer Study (MDCS) 16 . Cases were identified from the Swedish Cancer Registry up until 31 Dec 2007, and from The Southern Swedish Regional Tumour Registry for the period of 1 Jan - 31 Dec 2008. All tumours with available slides or paraffin blocks were histopathologically re-evaluated on haematoxylin and eosin stained slides. Histopathological, clinical and treatment data were obtained from the clinical and/or pathology records. TNM staging was performed according to the American Joint Committee on Cancer (AJCC). Information on vital status and cause of death was obtained from the Swedish Cause of Death Registry up until 31 Dec 2009. Follow-up started at date of diagnosis and ended at death, emigration or 31 Dec 2009, whichever came first. Median follow-up time was 3.35 years (range 0–17.69) for the full cohort (n = 626) and 6.05 years (range 1.03-17.69) for patients alive (n = 344). Patient and tumour characteristics of the cohort have been described in detail previously 4 14 15 . Ethical permission was obtained from the Ethics Committee at Lund University for the MDCS (Ref. 51/90), and the present study (Ref. 530/2008).

Cases with an insufficient amount of tumour material were excluded, whereby a total number of 557 (89.0%) tumours were suitable for tissue microarray (TMA) construction. Areas representative of cancer were marked on haematoxylin & eosin stained slides and TMAs were constructed as previously described 17 . In brief, two 1.0 mm cores were taken from each tumour and mounted in a new recipient block using a semi-automated arraying device (TMArrayer, Pathology Devices, Westminster, MD, USA). As demonstrated previously, there was no selection bias regarding the distribution of clinicopathological characteristics between the TMA cohort and the full cohort 14 .

For immunohistochemical analysis, 4 μm TMA-sections were automatically pre-treated using the PT-link system (DAKO, Glostrup, Denmark) and then stained in an Autostainer Plus (DAKO, Glostrup, Denmark). Immunohistochemical staining for SATB1 was performed using a monoclonal anti-SATB1 antibody (Clone EPR3895, Epitomics, Burlingame, CA, USA) diluted 1:100. The estimated fraction of cells with nuclear SATB1 expression was denoted as 0 (0-1%), 1 (2-25%), 2 (26-50%), 3 (51-75%) and 4 (>75%). Nuclear intensity was denoted as negative, weak, moderate or strong, with corresponding scores from 0–3, referring to the predominant intensity. A combined nuclear score (NS) was constructed by multiplying fraction and intensity. SATB1 expression was also evaluated in samples of normal colorectal mucosa from the surgical resection margins (n = 20). Stromal lymphocytes served as positive internal controls.

Immunohistochemical MSI screening status was assessed as previously described 4 18 , whereby tumour samples lacking nuclear staining of MLH1, PMS2, MSH2 or MSH6 were considered to have a positive MSI screening status. Hereafter, tumours with a positive MSI screening status are referred to as MSI and tumours with negative MSI screening status are referred to as MSS.

Immunohistochemical staining of beta-catenin was performed with a monoclonal anti-beta-catenin antibody (# 610153 BD Pharmingen, San Diego, Ca, USA), diluted 1:5000. The staining was evaluated as previously described 19 whereby membranous staining was denoted as 0 (present) or 1 (absent), cytoplasmic staining intensity as 0–2 and nuclear staining intensity as 0–2. The total score ranging from 0 (corresponding to membrane staining only, as in normal colonic mucosa) to 5 (tumours with strong nuclear and cytoplasmic staining) was then divided into three categories; 1 = score 0–1, 2 = score 2–3 and 3 = score 4–5). Sample IHC images representing different beta-catenin grades are shown in Additional file 1.

The immunohistochemical stainings was evaluated by two independent observers (BN and KJ), who were blinded to clinical and outcome data. Scoring differences were discussed in order to reach consensus.

Since SATB1 and SATB2 are very similar in sequence, the specificity of the anti-SATB1 antibody was analysed using ELISA and Western blot against purified recombinant SATB1 and SATB2 proteins. ELISA-plates (Costar) were coated with purified recombinant MYK/DDK-SATB1 produced in HEK293 cells (#TP300421, Origene Technologies, Rockville, MD, USA) or full-length SATB2 recombinant protein (#H00023314-P01, Abnova, Taiwan) at 1 μg/ml per well and incubated overnight at 4°C. Plates were blocked using 3% BSA in PBS for 1 hr at RT. Anti-SATB1 antibody (Clone EPR3895, Epitomics, Burlingame, CA, USA) was diluted 1:500 in 1% BSA in PBS and added to the wells. Following incubation for 1 hr at RT, plates were washed 3 times with wash buffer (1xPBS/0,05% Tween 20) and an HRP conjugated secondary antibody (#P0399, DakoCytomation) diluted to 1:2000 was added and incubated for 1 hour at RT. Plates were washed as before, substrate (Ready-to-use ABTS solution, Sigma Aldrich, #A3219) was added and allowed to develop for 30 minutes at RT. OD 405 was read on a microtiter plate reader.

For Western blot, 50 ng of purified recombinant MYK/DDK-SATB1 produced in HEK293 cells (#TP300421, Origene Technologies, Rockville, MD, USA) or 50 ng of a full-length SATB2 recombinant protein (#H00023314-P01, Abnova, Taiwan) were separated on precast 4–20% CriterionTGX SDS- PAGE gradient gels (Bio-Rad Laboratories, Hercules, CA) under reducing conditions, followed by blotting to PVDF membranes (Immobilon-P, Millipore), according to the manufacturer's recommendations. Membranes were blocked (5% dry milk, 0.5% Tween20, 1 × TBS) for 45 min at RT prior to addition of antibody (anti-SATB1, Clone EPR3895, Epitomics, Burlingame, CA, USA, diluted 1:200; anti-SATB2, HPA029543, Atlas Antibodies AB, Stockholm, Sweden, diluted 1:250). Following incubation for 1 h with the primary antibody, the membranes were washed 4x5 min in 1xTBS with 0.1% Tween20. An HRP-conjugated swine anti-rabbit antibody (#P0399, DakoCytomation), diluted 1/3000 in blocking buffer, was added to the membrane and incubated for 1 h followed by a final round of washing. Detection was carried out using Chemiluminescence HRP Substrate (Immobilon) according to the manufacturer's instructions.

As demonstrated in Additional file 2 A and B, the anti-SATB1 antibody binds specifically to SATB1 with no discernable cross-reactivity to SATB2. Thus, the antibody is highly specific to SATB1 and does not cross-react to SATB2, despite the extensive sequence similarity shared by the two proteins.

Chi Square and Spearman’s correlation (R) tests were used to explore the associations between SATB1 expression and clinicopathological and investigative parameters. Kaplan-Meier analysis and log rank test were used to illustrate differences in cancer specific survival (CSS) and overall survival (OS) according to three categories of SATB1 expression; negative (NS = 0), intermediate (NS 1–3) and high expression (NS >3), the latter including all tumours with >75% positive nuclei or various fractions of moderate-strong SATB1 staining. Cox regression proportional hazards models were used for estimation of hazard ratios (HRs) for death from CRC or overall causes according to negative and positive SATB1 expression in both uni- and multivariable analysis, adjusted for age (>/<=75 years), gender, T-stage (I-II, III, IV), N-stage (0,1,2), M-stage (0, 1), and differentiation grade (high-intermediate vs low). A backward conditional selection method was used for variable selection by the model. The interaction between SATB1 expression and SATB2 expression was explored by a Cox proportional hazards model including SATB1, SATB2 and an interaction variable. All tests were two-sided. A p-value of ≤ 0.05 was considered significant. All statistical analyses were performed using SPSS Statistics version 18 (SPSS Inc, Chicago, IL).

Following antibody optimisation and staining, SATB1 expression was successfully annotated in 529 (95.0%) tumours and beta-catenin in 527(94.6%) tumours. Sample images of immunohistochemical SATB1 staining are shown in Figure 1. SATB1 expression could be evaluated in 16/20 (80%) samples of adjacent, benign-appearing colorectal mucosa, of which 14 (87.5%) were denoted as having negative expression (Figure 1A) and 2 (2.5%) displayed weak expression in < 10% of cells. In the evaluated CRC cohort, 307 (58.0%) of the tumours were SATB1 negative (Figure 1B), and in the remaining tumours, SATB1 was expressed in various fractions and intensities (Additional file 3). In a small subset of tumours, strong SATB1 expression was seen in the vast majority of tumour cells (NS = 12; n = 18, Figure 1H). Notably, SATB2 was also abundantly expressed in all these tumours, with a NS = 12 in 15/18 cases and a NS = 8 in 3 cases. There was no marked heterogeneity in SATB1 expression between duplicate cores.

Next, we examined the relationship between SATB1 expression and established clinicopathological and investigative parameters. Three categories of staining were compared, i.e. SATB1 negative tumours (NS = 0), an intermediate group (NS = 1-3), and SATB1 high tumours (NS > 3). In the full cohort, SATB1 expression was significantly associated with female gender (p = 0.021), but not with any other conventional clinicopathological parameters (Table 1). A strong, signficant association was however seen between SATB1 expression and MSS tumours (p = <0.001), beta-catenin overexpression (p = <0.001) and SATB2 expression (p = <0.001). In light of the significant correlation between SATB1 and beta-catenin overexpression, we also assessed the association of SATB2 expression with beta-catenin grades, using the same categories as in Table 1. In line with the findings for SATB1, this revealed a strong positive correlation between SATB2 expression and beta-catenin overexpression (p < 0.001). Similar correlations of SATB1 expression with clinicopathological and molecular parameters were seen when the full range of nuclear scores was used in the analyses (data not shown).

Category denoted as negative refers to tumours with SATB1 nuclear score (NS) = 0, intermediate to NS 1–3 and strong to NS > 3. § Chi square test for linear trend. R = Spearmans correlation coefficient.The categories marked as not done and unknown were not included in the analysis. *Significant at the 0.05 level. ** Significant at the 0.01 level. N1 = 1-3 positive nodes, N2= > 4 positive nodes. MSS = Microsatellite stable, MSI = Microsatellite unstable. Beta-catenin grade: 0 = score 0–1, 1 = score 2–3, 2 = score 4–5.

Next, we examined the prognostic impact of SATB1 expression in strata of negative, intermediate and high SATB1 expression (Figure 2). Kaplan Meier analysis revealed no prognostic significance of SATB1 expression for CSS, neither in the full cohort (Figure 2A), nor in separate analyses for colon and rectal cancers (data not shown). Similar findings were seen for OS (data not shown).

The significant association of SATB1 expression with MSI status, beta-catenin grades and SATB2 expression led us to examine the modifying effect of these factors on the prognostic value of SATB1 expression. While no such effect was seen for MSI status and beta-catenin overexpression, SATB1 positivity was significantly associated with a shorter CSS in SATB2 negative tumours (p = 0.001, Figure 2B), in contrast to SATB2 positive tumours, where SATB1 was not significantly associated with survival (Figure 2C). Notably, none of the SATB2 negative tumours displayed high SATB1 staining (Figure 3B). The adverse prognosis for SATB1 positive/SATB2 negative tumours was confirmed in univariable Cox regression analysis (HR = 2.63; 95% CI 1.46-4.71, p = 0.001 for CSS and HR = 2.31; 95% CI 1.32-4.04, p = 0.003 for OS), remaining significant in multivariable analysis, adjusted for age, gender, TNM stage and differentiation grade (HR = 2.07; 95% CI 1.06-4.05, p = 0.034 for CSS and HR = 2.05; 95% CI 1.09-3.88, p = 0.026 for OS) (Table 2), with a significant interaction between SATB1 and SATB2 expression (p_interaction =0.011 for CSS and 0.015 for OS) (Table 2). The independent prognostic value of SATB1 was lost when vascular invasion was included in the multivariable analysis (data not shown). The prognostic value of SATB2 did not differ according to SATB1 expression (Figure 3).

Multivariable analysis included adjustment for age (>/<=75 years), gender, T-stage (I-II, III, IV), N-stage (0,1,2), M-stage (0, 1) and differentiation grade (high-intermediate vs low). p†: P-value for term of interaction by Cox multivariate analysis including SATB1 expression, SATB2 expression, and a term of interaction.