Cytoplasmic LXR expression is an independent marker of poor prognosis for patients with early stage primary breast cancer

Immunohistochemical staining for LXR was performed as previously described (Jeschke et al. 2019; Müller et al. 2019; Shao et al. 2020a, 2020b; Sixou et al. 2018). Briefly, whole tissue sections were cut and prepared from paraffin-embedded BC samples using standard protocols. After deparaffinizing in xylol for 20 min, endogenous peroxidase reaction was blocked with 3% hydrogen peroxide in methanol. Next, slides were rehydrated with a series of descending alcohol dilution and then boiled in a pressure cooker for 5 min, and immersed in sodium citrate buffer. Phosphate-buffered saline (PBS) was used for all washes and sections were incubated in blocking solution (ZytoChem Plus HRP Polymer System Kit, ZYTOMED Systems GmbH, Berlin, Germany) before incubation with primary antibody against LXR (LS-B262, Lifespan Biosciences, WA, USA) with a 1:200 dilution for 16 h at 4 °C. The antibody used was raised against A synthetic peptide made to an internal portion of the human LXR protein sequence (between residues 50–150), resulting in specificity against both LXRα and LXRβ isoforms. After incubation with a biotinylated secondary anti-rabbit IgG antibody, and with the associated avidin–biotin–peroxidase complex (both Vectastain Elite ABC Kit; Vector Laboratories, Burlingame, CA, USA), visualization was performed with substrate and chromogen 3, 3-diamino-benzidine (DAB; Dako, Glostrup, Denmark). Negative and positive controls were used to assess the specificity of the immunoreactions. Negative controls (colored in blue) were performed in BC tissue by replacement of the primary antibodies by species-specific (rabbit/mouse) isotype control antibodies (Dako, Glostrup, Denmark). Appropriate positive controls (placenta samples) were included in each experiment. Sections were counterstained with acidic hematoxylin, dehydrated and immediately mounted with Eukitt (Merck, Darmstadt, Germany) before manual analysis with a Diaplan light microscope (Leitz, Wetzlar, Germany) with 25× magnification. Pictures were obtained with a digital CCD camera system (JVC, Tokyo, Japan).

The staining of LXR was assessed according to a semiquantitative immunoreactive score (IRS), determined by multiplication of the positive cell proportion score (0 = 0%, 1 = 1–10%, 2 = 11–50%, 3 = 51–80%, and 4 = 81–100% stained cells) and the staining intensity score (0 = negative, 1 = weak, 2 = moderate, and 3 = strong). As previously described for RIP140, PPARγ, and THRβ1 (Shao et al. 2020a, 2020b), LXR cytoplasmic and nuclear staining results were evaluated in parallel, with a separate determination of cytoplasmic IRS and nuclear IRS. Total IRS was calculated by the sum of cytoplasmic and nuclear IRS. For all other markers, staining and IRS were determined in the whole cells, without differentiation of nuclear and cytoplasmic staining. A total of hundred cells (three spots with around 30 cells each) were analyzed for each sample and the IRS corresponded to the mean of the IRS determined on the three spots by two independent blinded observers. Discordant cases were re-evaluated by both observers together. After re-evaluation, both observers agreed on the result.

Receiver-operating characteristic (ROC) curve analyses were performed to calculate the optimal cut-off values between low and high LXR expression, based on the maximum differences of sensitivity and specificity. The threshold determined regarding OS was an IRS > 2.5 for nuclear LXR, > 5 for cytoplasmic LXR and > 8.5 for total LXR, which were used to determine the percentages of tumors expressing low or high LXR levels described in Table 2. Correlation analyses presented in Tables 3 and 4 were performed by calculating the Spearman’s-Rho correlation coefficient (p values of Spearman’s-Rho test presented).

Survival times were compared by Kaplan–Meier graphics and OS differences were tested for significance using the chi-square statistics of the log-rank test. Data were assumed to be statistically significant in the case of p value < 0.05. Kaplan–Meier curves and estimates were then provided for each group and each marker. The p value and the number of patients analyzed in each group are given for each chart.

LXR expression was assessed by IHC in a cohort of 305 primary BC samples (with clinical characteristics summarized in Table 1). We used an antibody directed to both LXRα and LXRβ isoforms, to focus on the importance of the subcellular expression of both isoforms. Among these samples, one bilateral primary BC was regarded as two individual cases. During follow-up, 40 patients have experienced a local recurrence, 58 have developed distant metastases, and 88 have died. Median age at initial diagnosis was 57.8 ± 0.7 years (range 26.7–94.6 years) and median follow-up time was 125 ± 38.6 months (range 1–153 months).

LXR immunoreactivity was present in both nucleus and cytoplasm of cancer cells. Examples of LXR staining from 5 patients are displayed in Fig. 1, with nucleo:cytoplasmic immunoreactive score (IRS) ratio indicated in each panel. Extreme (0:0 or 12:8) and intermediate (6:6) IRS were shown in panels A–C all showing equivalent distribution of LXR staining in the nuclear and cytoplasmic compartments. Besides, tumors with low nuclear and high cytoplasmic LXR expression (0:6) or with high nuclear and low cytoplasmic IRS (12:0) were also observed and are exemplified in panels D and E, respectively.

Distribution and correlation of LXR expression were analyzed in the whole cohort (Table 2). The mean IRS values of nuclear and cytoplasmic LXR expression were 5.11 and 3.05, respectively, demonstrating that LXR expression was stronger in the nucleus than in the cytoplasm. Both nuclear and cytoplasmic LXR expression was detected in a majority of BC samples, since only 15.41% and 28.52% of the tumors expressed no detectable nuclear and cytoplasmic LXR, respectively (Table 2).

Based on IRS cut-offs defined by performing an ROC-curve analysis for OS, all patients were divided into low and high expression subgroups. Based on these nuclear and cytoplasmic LXR cut-off values, an inverse distribution of BC was noticed, with a majority of samples exhibiting a high nuclear IRS (73.77% of the tumors) and a low cytoplasmic one (77.70%). Total LXR was positively correlated with nuclear and cytoplasmic (p = 4.23 × 10⁻⁷⁵ and 1.84 × 10⁻³⁸, respectively), and a positive correlation (p = 0.012) was also detected between nuclear and cytoplasmic LXR expression.

Using Spearman’s-Rho test, we analyzed the correlation between LXR expression and known clinicopathological features, including age, tumor size (pT), lymph-node status (pN), metastasis (cM), histology, stage, tumor grade, ER, PR, and HER2 status. As shown in Table 3, total and nuclear LXR expression were positively correlated with ER and PR status, whereas they were not correlated with other aggressive markers (pT, pN, cM, histology, stage, grade, and HER2 status). It should be noted that at the time of diagnosis, grading was not available for 143 noninvasive lobular carcinoma (classified in the NST group in Table 1). No significant associations were observed between cytoplasmic LXR and all parameters mentioned in Table 3.

As described in Table 2, we optimized the IRS cut-off values of LXR by performing ROC-curve analysis and divided the patient cohort into low and high expressing subgroups for all survival analyses. Comparisons of patient OS according to LXR expression levels were subsequently calculated by Kaplan–Meier analyses. Considering the whole cohort, no statistically significant correlations of nuclear, cytoplasmic, or total LXR expression were found with patient outcome (Fig. 2). Nonetheless, the trend existed with a possible favorable survival for the patients with low cytoplasmic LXR expression (p = 0.068).

We then then stratified patients in two subgroups, according to staging (stage I vs stage II–III). As shown in Fig. 3B, no significant difference was observed in the subgroups of patients with stage II–III (p = 0.977), whereas in stage I subgroup (Fig. 3A), patients with tumors showing a low level of cytoplasmic LXR expression had a significantly better outcome compared to those with a high level of cytoplasmic LXR (p = 0.001).

To better decipher this correlation with survival, we also stratify patients according to lymph-node status (negative vs positive) and tumor size (pT1 vs pT2-3). The correlation of cytoplasmic LXR with patient OS was significant in lymph-node negative tumors (Fig. 3C p = 0.040) but not in the subgroup of patients with tumors presenting lymph-node invasion (p = 0.359). The correlation was even more significant in the subgroup of patient with pT1 tumors (p = 0.00032, Fig. 3E) and lost in the pT2-3 subgroup (p = 0.198, Fig. 3F).

As a control, we also analyzed OS according to the expression of RXR, the heterodimerization partner of LXR. No significant correlations with OS were found in the whole cohort or in the subgroups according to staging, lymph-node status, or tumor size (Supplementary Fig. 1). Altogether, these data demonstrated that cytoplasmic LXR expression is associated with poor prognosis in patients with early stage BC (i.e., with small and noninvasive tumors).

Multivariate analyses were performed for the whole cohort and for the two subgroups according to staging (stage I vs stage II-III), using a Cox regression model with cytoplasmic LXR expression and other relevant clinicopathological characteristics, namely age at diagnosis, metastasis status, ER, and HER2 (Table 4). All parameters were demonstrated to be independent prognostic markers of OS for the whole cohort.

Interestingly, cytoplasmic LXR was regarded as an independent prognostic marker in the whole cohort, but this correlation was much stronger in the stage I subgroup, with hazard ratio of 1.871 (95% CI 1.156–3.030; p = 0.011) and 7.172 (95% CI 2.764–18.611; p = 5.1 × 10⁻⁵), respectively. In contrast, cytoplasmic LXR did not exhibit an independent prognostic value in the stage II–III subgroup. Similarly, cytoplasmic LXR was a strong independent prognostic marker in the pT1 subgroup (data not shown), but not in the pT2-3 subgroup, nor in the lymph-node negative or positive subgroups.