Vitamin D3 suppresses morphological evolution of the cribriform cancerous phenotype

Development of cribriform morphology (CM) heralds malignant change in human colon but lack of mechanistic understanding hampers preventive therapy. This study investigated CM pathobiology in three-dimensional (3D) Caco-2 culture models of colorectal glandular architecture, assessed translational relevance and tested effects of 1,25(OH)2D3, the active form of vitamin D. CM evolution was driven by oncogenic perturbation of the apical polarity (AP) complex comprising PTEN, CDC42 and PRKCZ (phosphatase and tensin homolog, cell division cycle 42 and protein kinase C zeta). Suppression of AP genes initiated a spatiotemporal cascade of mitotic spindle misorientation, apical membrane misalignment and aberrant epithelial configuration. Collectively, these events promoted “Swiss cheese-like” cribriform morphology (CM) comprising multiple abnormal “back to back” lumens surrounded by atypical stratified epithelium, in 3D colorectal gland models. Intestinal cancer driven purely by PTEN-deficiency in transgenic mice developed CM and in human CRC, CM associated with PTEN and PRKCZ readouts. Treatment of PTEN-deficient 3D cultures with 1,25(OH)2D3 upregulated PTEN, rapidly activated CDC42 and PRKCZ, corrected mitotic spindle alignment and suppressed CM development. Conversely, mutationally-activated KRAS blocked 1,25(OH)2D3 rescue of glandular architecture. We conclude that 1,25(OH)2D3 upregulates AP signalling to reverse CM in a KRAS wild type (wt), clinically predictive CRC model system. Vitamin D could be developed as therapy to suppress inception or progression of a subset of colorectal tumors.

the CDC42-PRKCZ-PARD complex [11,12] and regulates spindle orientation in nonpolarized cultured cells [13]. PRKCZ spatially regulates PARD3 that cooperates with the heterotrimeric G protein subunit GNAI3 (guanine nucleotide binding protein alpha inhibiting activity polypeptide 3; also known as Gαi3) to localize the spindle orientation protein, G-protein signalling modulator 2 (GPSM2; also known as LGN) [14]. PARD3 directs the orientation of pulling forces linked through GPSM2 to spindle microtubules for appropriate spindle alignment [14]. Perturbation of this machinery drives transition to dysplasia in Drosophila [15] but effects on colorectal glandular architecture remain unclear.
In this study, we investigated CM pathobiology using three-dimensional (3D) organotypic CRC culture model systems. We tested 1,25(OH) 2 D 3 treatment and investigated effects of mutationally-activated KRAS. To investigate translational relevance of our experimental findings, we conducted histologic, immunohistochemical and/or RNA in situ hybridization assays in murine and human tumors.

Translational and human studies
To investigate relationships between polarity signalling and tumor morphology, we conducted histologic, immunohistochemical or RNAscope in situ hybridization studies in murine or human intestinal tumors. We used an intestinal epithelial-specific PTEN knockdown murine model [34] to investigate morphology of intestinal cancers driven purely by PTEN-deficiency. Two small intestinal cancers developed after a long latency and showed cribriform morphology ( Figure 6A). In whole sections of 35 human CRCs in cohort (i), CM was heterogeneously distributed and affected > 50% glandular structures in 11.3 ± 8.5/40 fields per tumor. CM involved < 20% CRC surface area and was detected at low power (LP) (x3) magnification in 19/35 CRC whole tumor sections (54%). In TMA studies of cohort (ii), CM was detected at LP microscopy in 131/306 CRC cores (43%) from 92 CRCs. CM was more frequent in grade I and II CRCs in both study cohorts (Supplementary Figure S6A). Thirty six CRCs were mutated at KRAS exons 12 or 13 while 56 had wt KRAS. PTEN mRNA and protein expression were assayed by RNAScope [35] and immunohistochemistry (IHC) respectively. SLC9A3R1 IHC assays were also conducted as readout of apical PRKCZ activity [36]. We found that log-transformed PTEN RNAscope values (Supplementary Figure S6B), PTEN IHC and SLC9A3R1 IHC scores all correlated in human CRC (PTEN RNA vs PTEN IHC, r = 0.33; p < 0.01; PTEN RNA vs SLC9A3R1 apical intensity r = 0.36; p < 0.01; PTEN IHC vs SLC9A3R1 apical intensity r = 0.28; p < 0.01 Figure 6B). Apical SLC9A3R1 intensity directly associated with CM in all CRCs ( Figure 6C) but PTEN RNA expression associated with CM only in the KRAS wt subset ( Figure 6D). Apical SLC9A3R1 intensity had prognostic significance and inversely associated with histological grade ( Figure 6E) and lymph node metastasis ( Figure 6F).
Collectively, the above translational studies support the utility of 3D organotypic models for investigation of multiscale development of cancer morphology. As a manifestation of well-and moderately-differentiated CRC, CM associates with defective AP signalling and may represent an early or intermediate stage in a trajectory of cancer dedifferentiation (Supplementary Figure S6C). dIscussIon PTEN modulates the highly conserved apical CDC42-PRKCZ-PARD polarity complex [11,12] that has a pivotal role in mitotic spindle orientation [7,10,29,37], organization of epithelial architecture and tissue homeostasis [38]. PTEN regulates spindle orientation in nonpolarized cultured cells [13] and in this study we show PTEN regulation of spindle alignment in polarized Caco-2 cells during formation of simple colorectal glandular structures. Conversely, PTEN-deficiency induced spindle misorientation, epithelial stratification, apical membrane misalignment and formation of multiple abnormal lumens characteristic of cribriform morphology (CM). Furthermore, development of CM in Caco-2 ShPTEN glands associated with increased gland cellularity and size, in accord with loss of PTEN antiproliferative activity [39]. SiRNA knockdown of PTEN also induced cribriform architecture in a different CRC cell type (SK-CO-15) that has the capacity for 3D organotypic growth [40]. Our findings thus provide a mechanistic template for PTEN regulation of mitotic spindle alignment, growth of simple or stratified epithelium, AM dynamics and organization of colorectal multicellular architecture.
Downstream of PTEN, the CDC42-PRKCZ-PARD apical complex [11] tightly orchestrates spindle dynamics [9] and cell polarization [10]. PARD3 is essential for recruitment of PRKCZ to the apical surface, formation of the PARD3-PRKCZ-PARD6 complex and for CDC42 activation of PRKCZ [38]. These molecular interactions are implicated in multiple processes of epithelial organization [10]. Robust functional readouts of apical PRKCZ activity are provided by signal intensities of phospho-PRKCZ [9] or SLC9A3R1 [36] at the apical domain. In this study, high apical p-PRKCZ or SLC9A3R1 intensities in wt Caco-2 glands associated with appropriate subapical PARD3 localization, correct spindle alignment and regular gland morphology. Conversely in PTEN-deficient Caco-2 ShPTEN glands, we found reduced apical p-PRKCZ and SLC9A3R1 intensities, PARD3 mislocalization, spindle misalignment and aberrant multicellular glandular architecture. Hence, PTEN regulates components of the CDC42/PRKCZ/PARD apical polarity complex to control spindle orientation and 3D gland morphology.
To explore the translational relevance of our findings, we investigated tumor formation in an intestinalepithelial specific PTEN-deficient murine model [34] and assessed polarity signalling against CM in 2 human CRC cohorts. In the murine model, small intestinal cancers driven purely by PTEN-deficiency [34] developed CM. These findings accord with previous reports of CM in various cancers of PTEN-haploinsufficient mice [50]. In human studies of cohort (i), we found CM on low power microscopy in 54% CRCs and heterogenous CM distribution whole tumor sections. Because PTEN deficiency and mutationally-activated KRAS can synergistically co-regulate tumor morphology in transgenic mice [51], we studied a larger series of 92 KRAS genotyped CRCs (cohort ii) and found CM on low power microscopy in 43% CRCs. CM associated with grade I and II CRCs in both cohorts, consistent with an early transition state during a trajectory of CRC dedifferentiation. In accord with previous findings [52], we found no relationship between PTEN expression and KRAS mutational status in human CRC. However, PTEN RNA directly associated with CM in KRAS wt tumors but not in the KRAS mutant CRC subgroup, nor in the total CRC series of cohort (ii). These findings suggest that PTEN-KRAS epistatic interactions may influence human CRC morphology.
Downstream of PTEN, apical PRKCZ represents a central morphogenic effector within the apical polarity complex [11,36]. Apical SLC9A3R1 intensity provides a robust readout of apical PRKCZ activity in 3D models [36] and can be reliably assessed in human formalin-fixed paraffin embedded (FFPE) colorectal specimens [36,53]. In cohort (ii) human CRCs, we found positive correlations between PTEN RNAscope, PTEN IHC and apical SLC9A3R1 IHC intensity. We and others have shown that apical SLC9A3R1 intensity [36,53] and PTEN expression [54,55] are substantively higher in normal colonic mucosa than in CRC [36,[53][54][55]. In CRCs of the present study, we found higher expression of apical SLC9A3R1 intensity in cribriform as opposed to non-cribriform CRCs, consistent with CM as an early or intermediate stage in a trajectory of cancer dedifferentiation. A similar rationale may explain the higher PTEN RNA expression in cribriform CRC than in non-cribriform KRAS wt CRCs.
Compelling experimental, epidemiological and clinical data show that Vit-D controls inception and progression of CRC [25,26]. However, there is a fundamental gap between discovery of Vit-D anticancer activity and identification of mechanistic biomarkers needed to exploit its full clinical potential. Strikingly, our findings show that 1,25(OH) 2 D 3 controls subcellular, cellular and multicellular scales of tissue assembly to suppress CM. Hence, 1,25(OH) 2 D 3 anticancer effects may be mediated in part by Ca 2+ /CaM-KII-dependent reprogramming of polarization machinery to suppress oncogenic disruption of homeostatic multicellular architecture. Our study shows that KRAS mutation indicates Vit-D-resistance. Conversely, apical SLC9A3R1 intensity provides readout of PRKCZ [36] a key morphogenic effector of the PTEN/CDC42/PRKCZ pathway [11], has prognostic relevance in human CRC and predicts 1,25(OH) 2 D 3 control of gland morphology. Apical SLC9A3R1 intensity is suppressed by mutationallyactivated KRAS in 3D models and associates with CM in both KRAS mutant and wt human CRCs.
Globally, cancer affects over 12 million new patients each year [56]. Cancer morphology has been a gold-standard for diagnosis and outcome prediction since the time of Virchow [57] but has remained a mechanistic "black box" with few advances and almost no literature exploring its pathobiology. Our MS now untangles the molecular framework of cribriform morphology in 3D CRC models, shows Vit-D suppression of CM evolution via core polarization machinery and conducts translational and clinical studies that support model predictions.
We also identify biomarkers of Vit-D resistance (KRAS mutation) and promorphogenic effects (apical SLC9A3R1) for use in future clinical trials.

Three-dimensional (3D) cultures
Development of multicellular architecture was assessed in PTEN-expressing Caco-2, PTEN-deficient Caco-2 ShPTEN cells, parental PTEN-expressing SK-CO-15 cells and a subclone rendered PTEN-deficient by SiRNA knockdown, in organotypic cultures. Cells were cultured and embedded in Matrigel matrix (BD Biosciences, Oxford, UK), then imaged by confocal microscopy during progressive development of multicellular glandular architecture, as previously described [12,36]. SK-CO-15 cells express apical membrane markers at low level [60] and apical SLC9A3R1, PRKCZ or p-PRKCZ were undetected in these cells, in this study. Cribriform morphology (CM) was defined as multiple aberrant lumens surrounded by abnormal stratified epithelium in 3D multicellular structures in culture and in tumors [6]. Effects of transfections or treatments on glandular morphology of 3D cultures were assessed against endpoints of CM or individual features of epithelial configuration (columnar or stratified) or single central lumen formation. www.impactjournals.com/oncotarget

Intestinal-epithelial specific PTEN-deficient murine model
All animal procedures were conducted in accordance with local and national regulations. Mice were generated, housed, and genotyped, and Cre activity was induced as previously described [34]. A total of 30 Ah::CreERT T+ / 0 ;Pten F/F mice and 29 Ah::CreERT T+/0 ;Pten +/+ mice were enrolled into cohorts for prolonged follow up. Tissues were harvested, fixed, and processed according to standard protocols, as previously described. [34].Animals were monitored closely for symptoms of disease, and were then necropsied as previously described [34]. The morphology of tumors arising in PTEN-deficient murine intestinal epithelium was assessed by H&E histology. Cribriform morphology was assayed as previously defined [6].

Human colorectal cancer studies
We conducted 2 separate studies of polarity signalling against cribriform morphology (CM) in human colorectal cancer (CRC). We used anonymised formalin fixed, paraffin embedded (FFPE) samples from (i) 35 patients with non-genotyped CRCs and (ii) 92 patients with KRAS-genotyped CRCs. We assessed CM at low power microscopy according to previously defined criteria [6] in both study cohorts. To assess CM heterogeneity, we scored CM in 40 fields per tumor at 20x magnification across whole tumor sections in cohort (i). Scores of 0, 1 and 2 were given for CM involvement of < 10%, 11-50% and > 50% CRC epithelium per field. In cohort (ii), specimens were arranged in tissue microarrays (TMAs). To assess polarity signalling in CRC FFPE specimens, we assessed PTEN RNA expression by RNAscope in situ hybridization [35]. We assessed apical SLC9A3R1 intensity by immunohistochemistry (IHC) as readout of apical PRKCZ activity as outlined previously [36], in FFPE specimens of both study cohorts. To assess PTEN protein expression, PTEN IHC was also assessed in cohort (ii) TMAs. Samples used in this research were released from the Northern Ireland Biobank (NIB13-0090), approved by the Office of Research Ethics Committees Northern Ireland (Reference number 11/NI/0013/-/NIB13-0090).

Data analysis
Descriptive statistics were expressed as the mean ± sem. Statistical analyses were by one or two-way ANOVA or Student's t test using SPSS for Windows release 22.0 (IBM Corp, NY, USA) or Graphpad Prism software (v4.02; Graphpad CA 92037 USA). Scatterplots and bar charts were used for display of quantitative numerical or categorical data. PTEN RNA values were log transformed to provide a normal distribution. www.impactjournals.com/oncotarget

Editorial note
This paper has been accepted based in part on peerreview conducted by another journal and the authors' response and revisions as well as expedited peer-review in Oncotarget.