Doublecortin-like kinase 1 expression associates with breast cancer with neuroendocrine differentiation

Doublecortin-like kinase 1 (DCLK1), a microtubule associated kinase, has recently been proposed to be a putative marker for stemness and adverse prognosis in gastrointestinal cancers. However, it is not clear whether the protein also plays similar roles in breast cancer. Here, the expression of DCLK1 was analyzed in a large cohort of invasive breast cancers (IBC) by immunohistochemistry. DCKL1 was associated with favorable clinico-pathologic features, namely lower histologic grade, absence of lymphovascular invasion, fibrotic focus, necrosis and lower pN stage (p≤0.045). Additionally, independent significant correlations were found with estrogen receptor and neuroendocrine markers (p ≤0.019), implicating its relationship with IBC with neuroendocrine differentiation (IBC-NED). In the current cohort, IBC-NED showed worse outcome than luminal cancers without NED (hazard ratio=1.756, p=0.041). Interestingly, within the IBC-NED group, DCLK1 was found to be a good prognostic factor (hazard ratio =0.288, p=0.011). These findings were in contrast to those in gastrointestinal cancers, suggesting different functional roles of DCLK1 in different types of cancers. In clinical practice, NED is not routinely assessed; thus IBC-NED are not well studied. Its poor outcome and significant heterogeneity warrants more attention. DCLK1 expression could aid in the prognostication and management of this special cancer subtype.


INTRODUCTION
Doublecortin-like kinase 1 (DCLK1) is a microtubule associated kinase, containing two doublecortin domains in the N-terminus for regulation of microtubule polymerization and a serine/threonine protein kinase domain in the C-terminus.It also shows substantial homology to Ca2+/calmodulin-dependent protein kinase.In between the N and C termini, there is a serine/prolinerich domain for mediating multiple protein-protein interactions [1].Its function was first described in neuronal migration and development [2].Recently, its expression has been found in colon cancers [3][4][5], pancreatic [6] and esophageal cancers [7].
DCLK1 has been proposed as a cancer stem cell marker for gastrointestinal cancers.DCLK1 specifically marked cancer stem cells (CSC) that self-renew and generate tumor progeny in Apc Min/+ mice [8].DCLK1 positive differentiated tuft cells can be activated by tissue injury and initiate colon cancer [9].In pancreatic cancers, DCLK1 positive cells displayed increased sphere forming and tumor initiating capacity [10], and enhanced epithelial-mesenchymal transition (EMT), a process closely linked to the acquisition of stem cell properties, via regulation of microRNA biogenesis [4,6,11].Interestingly, colorectal cancer with high DCLK1 expression had increased cancer specific mortality [5].Upregulation of DCLK1 expression in blood circulation was found in chemoradiotherapytreated colorectal cancer patients [12].
Aberrant DCLK1 expression was detected in IBC [13].Knockdown of oncogenic miR-21 in IBC was accompanied by a decrease in DCLK1 expression [14].Apart from these, no other information is currently available.It is not clear whether the aberrant DCLK1 expression contributed to IBC tumor aggressiveness as in gastrointestinal cancers.In addition, CSC heterogeneity exists in different breast cancer subtypes [15], and the commonly used CSC markers did not identify all CSC populations.It will be interesting to explore using DCLK1 as a CSC marker in breast cancer subtypes.
In this study, we evaluated the expression of DCLK1 in a large cohort of breast cancer by immunohistochemistry (IHC), its association with clinico-pathological features and other biomarkers (including CSC markers) expression, as well as the relationship with breast cancer outcome.

RESULTS
A total of 1132 cases were included in this cohort.The mean patients' age was 54.6±12.7 (range 22-97) years.The mean tumor size was 2.67±1.52(range 0.2-13.9)cm.One hundred and seventy three cases (15.3%) were grade I, 457 cases (40.4%) were grade II and 502 cases (44.3%) were grade III.Nine hundred and eighty seven cases were IBC of no special type (IBC-NST).There were 35 cases of invasive lobular cancers (ILC), 48 cases of breast cancers with medullary features, nine cases of mucinous cancers and eight cases of neuroendocrine cancers.The remaining 45 cases were of other miscellaneous histologic types, including micropapillary carcinoma, papillary carcinoma, tubular carcinomas, tubulo-lobular carcinoma and metaplastic carcinomas.Details of the clinico-pathologic features are summarized in Table 2. Overall, 418 cases (36.9%) were DCLK1 high and 513 cases (63.1%) were DCLK1 low.Representative staining is shown in Figure 1.

Correlation with clinico-pathologic features, biomarkers and molecular subtypes
High DCLK1 expression correlated with lower grade (p < 0.001), the absence of FF (p =0.045), the absence of necrosis (p = 0.005), the absence of LVI (p = 0.004) and lower pN stage (p = 0.002) but not age, EIC and pT stage (Table 2).

Relationship with outcome
Follow up data were available for 987 cases with a mean follow-up duration of 65.6 months (range 1-210 months).One hundred and forty nine cases (15.1%) had It appears that the association of DCLK1 with outcome in luminal cancers is mainly related to the relationship with IBC-NED, which are clustered within the luminal group of cancers.Analysis of the prognostic impact of DCLK1 in the IBC-NED cases and non-NED luminal cases found that DCLK1 expression was associated with better DFS (log-rank= 12.187, p < 0.001) and OS (log-rank=7.222,p = 0.007) in IBC-NED but not in non-NED luminal cancers (Figure 2).Recent investigations suggested adverse prognosis of IBC-NED [16,19].This study also showed worse OS (log-rank=4.658,p = 0.031) and DFS (log-rank=9.294,p = 0.002) in IBC-NED than luminal cancers without NED.The prognostic impact on DFS of IBC-NED was independent of ER, PR, HER2, Ki67, grade, age, tumor size and nodal involvement (HR=1.756,p = 0.041 with reference to non-NED luminal).Interestingly, those IBD-NED with worse outcome showed focal but not diffuse NED expression (supplementary figure 1S).As DCLK1 was associated with diffuse NED expression, we then compared the patients' parameters and outcome based on groupings stratified by different levels of neuroendocrine and DCLK1 expression.IBC-NED with low DCLK1 expression showed the lowest DFS rate, when compared to non-NED luminal cancers regardless of the neuroendocrine expression pattern (DCLK1 low/ neuroendocrine focal: log-rank=8.861,p = 0.003; DCLK1 low/ neuroendocrine diffuse: log-rank=7.211,p = 0.007) (Figure 3).Of note, DCLK1 was found also to have independent favorable prognostic impact on DFS of IBC-NED (HR=0.288,p = 0.011, 95%CI= 0.111-0.748)after adjustment of grade, age, tumor size, LVI, pN stage, HER2, Ki67, PR and neuroendocrine markers expression (Supplementary Table S4).

DISCUSSION
In this study, the expression of DCLK1 in a large cohort of breast cancer was analyzed.In contrast to its cancer initiating roles in gastrointestinal tumors, DCLK1 expression in breast cancer did not appear to be related to stem cell features and aggressive behavior.No positive correlation was observed with other breast CSC markers.Nonetheless, DCLK1 was negatively correlated with grade, basal (c-kit and CK5/6) as well as EMT (p-cadherin) markers.DCLK1 was more frequently found in luminal cancers than basal-like and HER2-OE subtypes [15,20,21], associated with IBC-NED and potentially an independent favorable prognostic factor in these cancers.
The diagnosis of IBC-NED has been controversial.It was first defined by the presence of morphologic features similar to those of neuroendocrine tumors of the gastrointestinal tract or of the lung.However, classic neuroendocrine morphology is comparatively rare in breast cancer, thus its significance has been debatable for some time.Recently, a formal categorization of this tumor in WHO classification has been established.Apart from the morphologic features (including well-differentiated neuroendocrine tumors and poorly differentiated neuroendocrine carcinoma or small cell carcinoma), it has been defined with the expression of NED markers in over 50% of the tumor cell population in the 2003 WHO classification [22].In the latest 2012 classification [18], it was revised to include all tumors expressing NED markers to a greater or lesser degree.The prognostic significance of NED per se in IBC-NED has also been uncertain, with reports suggested no effects on prognosis [23][24][25], better [26,27] or worse [16,19] prognosis.In the current study, the latest WHO criteria were adopted for diagnosis of IBC-NED [18].In agreement with the others [16,19,28,29], IBC-NED was mainly associated with luminal cancers.However, IBC-NED was not associated with low tumor grade or favorable outcome as would be expected for luminal cancers.This result corroborated with heterogeneity in IBC-NED, which encompass both low grade special subtypes and aggressive high grade cancers [17].In contrast to the previous reports showing the poor prognosis of diffuse neuroendocrine markers expression [19] [16] compared to non-NED IBC cases.Of note, when the non-NED IBC were further classified into luminal and non-luminal cases, IBC-NED in fact showed an intermediate outcome, which was better than non-luminal but worse than non-NED luminal cancers.Possibly the heterogeneity in the cohort and variation in diagnostic criteria contributed to the variation in its prognostication.
As IBC-NED is heterogeneous, factors that further stratify IBC-NED into different prognostic groups could be useful for the management of these cancers.DCLK1 expression was found to be an independent favorable prognostic factor for DFS in IBC-NED regardless of the NED expression pattern.Currently, there are no specific recommendations for treatment of IBC-NED and these cases are treated as IBC-NOS.As IBC-NED are mostly luminal, they are treated with either endocrine therapy alone or together with adjuvant chemotherapy, with the latter depending on risk assessment.The additional independent prognostic impact of DCLK1 could be crucial for therapeutic decision to identify IBC-NED with favorable outcome, and sparing patients from chemotherapy.The mechanism of its good prognostic  A recent report suggested that DCLK1 can antagonize Runx2 [30].Runx2 has been shown to be a regulator of epithelial cell fate in mammary gland development and breast cancer [31].Overexpression of Runx2 drove EMT-like changes in normal mammary epithelial cells, whereas its deletion in basal breast cancer cells inhibited cellular phenotypes associated with tumorigenesis [32].
In fact, in our preliminary study on Runx2 expression, we observed a significantly higher expression of basal marker in DCLK1loRunx2hi cases compared to the others (data not shown).Interestingly, there are also interactions between them in IBC-NED patients' outcome.Cases with DCLK1loRunx2hi expression were found to have significantly worse DFS than other subgroups (data not shown).Our data may suggest that DCLK1 could at least partly act via Runx2.In summary, DCLK1 was found to be a good prognostic factor in breast cancer, particularly in IBC-NED.The result was in contrast to its tumor promoting roles in gastrointestinal cancers, suggesting different functional roles of DCLK1 in different type of cancers.In addition, using the current WHO classification, we found that IBC-NED showed a worse outcome, attributable in part to tumor heterogeneity.DCLK1 expression was shown to stratify IBC-NED into different prognostic groups.The findings could aid in the prognostication and management of this special type of IBC.

Patients data
The histologic files of the three involved institutions were searched for IBC over periods of 2 (03-04), 4 (2002-05) and 7 (2003-09) years.All consecutive cases with excision specimens were included.All the specimens were formalin fixed, paraffin embedded and routinely processed.The 4 micron slides were stained with H&E and reviewed by two of the authors.The tumors were graded using modified Bloom and Richardson grading [16] and the histologic diagnosis was confirmed (WHO [17]).Invasive breast cancers with neuroendocrine differentiation (IBC-NED) were defined by the presence of neuroendocrine morphological features (neuroendocrine carcinoma) and the presence of neuroendocrine differentiation (NED) (neuroendocrine marker positivity) [17].IBC was considered as having NED if ≥ 1% tumor cells showed any expression of GC and/or SYN or showing morphologic features of NED [18,19].In addition, lymphovascular invasion (LVI), the presence of extensive in situ components (EIC) and fibrotic focus (FF) were also evaluated as present or absent as previously reported [20].Patient details and clinical information were retrieved from the medical records including patients' age, tumor size, pN stage, pT stage and patient outcome data.For the outcome data, overall survival (OS) was defined as the time interval from the date of initial diagnosis to the date of breast cancer related death.Disease free survival (DFS) was defined as the duration from the date of initial diagnosis to the first detection of breast cancer specific relapse or death.The study was approved by Joint Chinese University of Hong Kong-New Territories East Cluster clinical research ethics committee.

Tissue microarray construction
Cellular areas of the tumors on H&E slides were chosen and the corresponding areas were taken from the paraffin blocks for tissue microarray (TMA) construction.Two 0.6 mm tissue cores were obtained from each case.One additional core was taken from available nodal metastases.The TMAs were assembled with a tissue arrayer (Beecher Instruments, Silver Springs, MD).Sixty two composite TMA blocks, each containing maximum 54 tissue cores, were constructed.Serial 4 micron sections were cut and transferred to Superfrost Plus glass slides (Menzel-Glaser, Germany).One section from each tissue array block was stained with H&E to confirm the presence of representative tumors in the TMA blocks.

Immunohistochemistry and scoring
The TMA slides were assessed for the different groups of biomarkers, in addition to DCLK1.The first group were steroid hormone receptors (estrogen receptor (ER), progesterone receptor (PR), and androgen receptor (AR)).The second group were growth factor receptors (human epidermal growth factor receptor 2 (HER2), epidermal growth factor receptor (EGFR)) and a proliferation marker (Ki67).The other groups were neuroendocrine markers (chromogranin (CG) and synpatophysin (SYN)), basal markers (p63, c-kit, CK14 and CK5/6), cancer stem cell markers (CD44, SOX2 and ALDH1) and EMT markers (vimentin and p-cadherin).IHC of all markers was performed by BenchMark XT automated slide-staining instrument (Ventana, Arizona, USA) with Ultraview Universal DAB Detection Kit (Ventana, Arizona, USA) after deparaffinization, rehydration and antigen retrieval.After primary antibody incubation, the sections were incubated with anti-mouse horseradish peroxidase labeled polymer (Roche, Arizona, USA) for 30 min at room temperature, and then developed with diaminobenzidine.All slides were counterstained with hematoxylin.The TMA slides were scored for the intensity of staining in the nucleus, cytoplasm or membrane according to different antibodies by two of the authors blinded to the clinical information and the staining results of other markers.Details of the antibodies, antigen retrieval, staining conditions and scoring were listed in Table 1.For DCLK1 staining, the reactivity assessed was cytoplasmic.DCLK1 was assessed for both intensity and proportion of positively stained cells.The staining intensity was graded from 0 to 3 whereas the proportion of stained cells was scored on a scale of 0-4 (0= no detectable staining, 1= 1-25% positive cells, 2= 26-50% positive cells, 3=51-75% positive cells and 4= over 75% positive cells).An immunoscore was obtained by adding the intensity score and the percentage score.Positivity for DCLK1 was defined using the mean immunoscore as the cutoff.Immunoscore of 0-3 was regarded as negative and >3 as positive.Any discrepancies were resolved by reviewing at a multi-head microscope to reach a consensus.

Statistical analysis
The findings were analyzed using the statistical software SPSS for Windows, Version 18. Chi-square analysis or Fisher's exact test were used to test for the association of DCLK1 expression with tumor grade, FF, LVI, EIC, pN, pT, molecular subtypes and biomarker expression.Mann-Whitney U test was used for analyzing the differences in patient's age and tumor size with DCLK1 expression.The relationship of DCLK1 as a continuous variable with various clinico-pathologic features and biomarker expression was also analyzed with Mann-Whitney U test (for categorical data) and spearman correlation (rs) (for continuous variables).Survival data were evaluated with Kaplan Meier analysis and Cox regression analysis using the backward Wald method.Statistical significance was established at p < 0.05.

Figure 1 :
Figure 1: Representative immunohistochemical staining of DCLK1 (x400).High cytoplasmic immunoreactivity of DCLK1 in tumor cells but not in surrounding stroma.All micrographs were taken with a 40x objective, Nikon microscope equipped with a digital color camera and software.

Figure 2 :
Figure 2: Kaplan-meier analysis of DFS and OS on non-NED luminal and IBC-NED cancers according to DCKL1 expression.DFS in non-NED luminal cancers A. and IBC-NED B. with different level of DCLK1 expression was compared.OS in non-NED luminal cancers C. and IBC-NED D. with different level of DCLK1 expression was compared.DCKL1 expression was related to DFS and OS in IBC-NED but not non-NED luminal cancers.

Figure 3 :
Figure 3: Kaplan-meier analysis on DFS of non-NED luminal and IBC-NED according to NED marker and DCKL1 expression.DCKL1 low IBC-NED regardless of focal or diffused NED marker expression showed worse DFS compared to non-NED luminal cancers.

Table 4 : Association DCLK1 with clinico-pathological features and biomarkers in IBC-NED
, this study showed poorer outcome in patients with focal rather than diffuse NED expression in IBC-NED.One study applied similar criteria demonstrated