Keratin 19 regulates cell cycle pathway and sensitivity of breast cancer cells to CDK inhibitors

Keratin 19 (K19) belongs to the keratin family of proteins, which maintains structural integrity of epithelia. In cancer, K19 is highly expressed in several types where it serves as a diagnostic marker. Despite the positive correlation between higher expression of K19 in tumor and worse patient survival, the role of K19 in breast cancer remains unclear. Therefore, we ablated K19 expression in MCF7 breast cancer cells and found that K19 was required for cell proliferation. Transcriptome analyses of KRT19 knockout cells identified defects in cell cycle progression and levels of target genes of E2F1, a key transcriptional factor for the transition into S phase. Furthermore, proper levels of cyclin dependent kinases (CDKs) and cyclins, including D-type cyclins critical for E2F1 activation, were dependent on K19 expression, and K19-cyclin D co-expression was observed in human breast cancer tissues. Importantly, K19 interacts with cyclin D3, and a loss of K19 resulted in decreased protein stability of cyclin D3 and sensitivity of cells towards CDK inhibitor-induced cell death. Overall, these findings reveal a novel function of K19 in the regulation of cell cycle program and suggest that K19 may be used to predict the efficacy of CDK inhibitors for treatments of breast cancer.


Results
K19 is required for cell proliferation. MCF7 cells were genetically engineered to ablate K19 expression using the CRISPR/Cas-9 system to ensure complete loss of K19 expression. Experiments were carried out using two different KRT19 KO clones (KO1 and KO2) to assess the effects of K19 ablation. Both western blotting (Fig. 1a) and quantitative RT-PCR (qRT-PCR) (Fig. 1b) confirmed the loss of K19 expression in MCF7 KRT19 KO cell lines. These losses were specific to K19 as expression of K8 and K18, two other keratins expressed in MCF7 cells 4 remained unaffected compared to the wild type parental control (Fig. 1a).
While growing cells, we observed that KRT19 KO cells exhibited consistent decreases in cell proliferation compared to that of the parental control. To quantify our observation and determine cell proliferation, we counted cell numbers (Fig. 1c) and performed MTT assays (Fig. 1d) each day following cell passaging. Although the same number of cells were plated initially, both KRT19 KO clones showed modest but statistically significant decreases in cell number and metabolic activity. Of note, although both KRT19 KO clones showed same trends, we noticed that KO2 cells showed greater decreases in the cell proliferation rate compared to KO1 cells, likely due to the well-documented heterogeneity of the MCF7 cell line 26 from which these clones were derived. For an added measure, we decided to re-express K19 and thereby rescue K19 expression in KRT19 KO cells by generating KO2 cells stably expressing K19 through lentiviral transduction. Consistent with our findings in Fig. 1c,d, cell proliferation of KRT19 KO cells expressing K19 was increased compared to those expressing vector control (Fig. S1). Overall, our data indicates that K19 is required for cell proliferation. K19 is required for the expression of E2F1 target genes. Next, we sought out a mechanism underlying delayed cell cycle progression in KRT19 KO cells. Reactome had identified 43 cell cycle-related genes that were downregulated in KRT19 KO cells (Supplementary Table S4). We then used g:Profiler (http://biit.cs.ut.ee/gprofiler/) 29 and GSEA to identify potential transcriptional regulator(s) for those 43 genes based on their upstream cis-regulatory motifs (Fig. 4a). Using these methods, a key cell cycle transcriptional regulator E2F1 was identified as either an only transcriptional regulator according to g:Profiler, or a top regulator according to GSEA. In total, g:Profiler and GSEA analyses found 15 out of 43 genes to have putative E2F1 regulatory motifs (Fig. 4b). Among the 15 identified, expressions of genes such as TUBG1 30 , TERT 31 , and MYBL2 32 have already been shown to be E2F1-dependent in various tumor settings. Interestingly, E2F1 itself was one of the 43 cell cycle-related genes downregulated in KRT19 KO cells. Next, we sought out to confirm the RNA-sequencing result using qRT-PCR. 13 out of 15 cell cycle-related E2F1-dependent genes were downregulated in both KO1 and KO2 cells, compared to the parental control in statistically significant manners (Fig. 4c). These genes include ARPP19, TUBG1, TERT,  SETD8, RFC5, MYBL2, ESPL1, H2AFX, E2F1, YWHAB, TMPO, AKT2, and POLE. Overall, our bioinformatics analyses and qRT-PCR results point to E2F1 as a potential regulator mediating delayed cell cycle progression in KRT19 KO cells.
During G1 phase, E2F1 is bound to the retinoblastoma protein (Rb) which prevents gene transcription by E2F1 for transition from G1 to S phase 33 . The inhibition of E2F1 activity is relieved by phosphorylation of Rb, as   www.nature.com/scientificreports www.nature.com/scientificreports/ K19 is required for the proper expression of select CDKs and cyclins. The cell cycle is regulated by cyclins which activate CDKs to promote different stages along the cell cycle 37 . In particular, dimerization with D-type cyclins, such as cyclins D1 and D3, activates CDK4 and CDK6 to hyper-phosphorylate Rb 38 . Since Rb phosphorylation was decreased in KRT19 KO cells (Figs 5 and S2), we decided to examine levels of cyclins D1 and D3 along with cyclin E which complexes with CDK2 to also phosphorylate Rb 39 . In addition, since KRT19 KO cell lines showed a decreased number of cells in G2/M phase in Fig. 2, the level of cyclin B1, which binds to CDK1 and regulates the progression in G2/M phase was also examined 40 . Examining levels of cyclins, we found that cyclins D1, D3 and B1 were decreased in KRT19 KO cells compared to the parental control ( Fig. 6a,b). In contrast, cyclin E level was not decreased in KRT19 KO cells and in fact, increased in the KO2 clone. Overall, these data suggest that the decrease in Rb phosphorylation is due to decreased D-type cyclin availability. Interestingly, when examining levels of CDKs, levels of CDK4 and CDK1 were decreased in KRT19 KO cells unlike that of CDK-activating kinase CDK7 (Figs 6a and S3). We then confirmed K19-dependent expression of cell cycle regulators using KRT19 KO cells stably expressing vector or K19 (Fig. 6c,d). Rescue of K19 expression increased E2F1, cyclins D1, D3, B1 and CDK1 levels while cyclin E level was decreased. Thus, these data illustrate the requirement of K19 in expression levels and thus activities of D-type cyclins and their partner CDK as well as cyclin B1-CDK1 complex.
We then decided to confirm the correlation between expression of K19 and D-type cyclins in human tumors. Tissue sections including both invasive tumor and adjacent benign components from 21 breast cancer patients with high grade tumors were immunostained for K19, cyclin D1 and cyclin D3 and semiquantitatively scored for their immunoreactivity (Fig. S4, Supplementary Tables S6-8). While 19 of 21 (>90%) of tumors that were scored as high positive for K19 were also high positive for cyclin D1, only 9 of 21 (<43%) of benign components that were high positive for K19 were high positive for cyclin D1 (Fig. 6e). Likewise, while 19 of 21 of tumors scored as high positive for both K19 and cyclin D3, a fewer number (13 of 21) of benign components scored as high positive for both K19 and cyclin D3 (Fig. 6f). Thus, there was a higher correlation between high K19 and D-type cyclins co-expression in tumor compared to benign tissues. These findings further support the correlation between expression of K19 and D-type cyclins in breast cancer.
K19 interacts with cyclin D3 and regulates its stability. As mentioned, D-type cyclins are critical for the phosphorylation of Rb and subsequent activation of E2F1. Although their protein levels are decreased in KRT19 KO cells (Fig. 6), these decreases were not observed at the mRNA level (data not shown). These results suggest that K19 may potentially regulate cyclins at the protein level through a protein-protein interaction. Indeed, cyclin D3 has been shown to interact with actin cytoskeleton protein in MCF7 cells 41 . Thus, we decided to look for a potential interaction between K19 and cyclin D3. For this, we performed co-immunoprecipitation experiments, and pull down of cyclin D3 showed K19 and vice versa (Fig. 7a,b), suggesting that K19 binds to cyclin D3.
Since K19 interacts with cyclin D3 and cyclin D3 level is dependent on K19 ( Fig. 6a-d), we next decided to examine protein stability of cyclin D3 by examining its level following inhibition of protein synthesis using cycloheximide. When normalized to untreated controls, cyclin D3 level was significantly decreased in www.nature.com/scientificreports www.nature.com/scientificreports/ , and KO2 cells were harvested, and immunoblotting was performed with antibodies against the indicated proteins. (b) Signal intensities of cyclins from (a) were quantified and normalized to those of the GAPDH loading control. Data from at least three experimental repeats normalized to that of the parental control are shown as mean ± SEM. Differences are not statistically significant unless denoted by *p < 0.05; **p < 0.001. (c) Whole cell lysates of KRT19 KO cells expressing vector (V) or K19 (K19) were harvested, and immunoblotting was performed with antibodies against the indicated proteins. (d) Signal intensities of bands from (c) were quantified and normalized to those of GAPDH loading control. Data from at least three experimental repeats normalized to that of the parental control are shown as mean ± SEM. Differences are not statistically significant unless denoted by *p < 0.05; **p < 0.001. Tissue sections from 21 differerent breast cancer patients with most aggressive tumors were immunostained for (e) K19 and cyclin D1 or (f) K19 and cyclin D3. The immunoreactivity of cells in both the invasive tumor and adjacent benign epithelium in each case were scored and categorized as shown in Supplementary Tables S6 and 7. Those that were strongly positive in both K19 and cyclin D1 or cyclin D3 were separated from cases that were not (others).  www.nature.com/scientificreports www.nature.com/scientificreports/

Conclusions
In this study, we discovered that K19 plays a critical role in the proliferation of MCF7 breast cancer cells. Studies over the years have shown that keratins can promote tumor growth, especially when their expression levels are altered 3 . The requirement of K19 in MCF7 cell proliferation further supports the active participation by this family of cytoskeletal proteins in cancer cell growth.
As mentioned in the introduction, although many studies have shown a correlation between higher levels of K19 expression to worse breast cancer patient prognosis, in vitro as well as in vivo studies have shown mixed results on the role of K19 in cancer cell proliferation and tumor growth. Intermediate filament proteins are known to function in a highly context-dependent fashion 1 . This is due in large part to the presence of other intermediate filament proteins co-expressed in tested cells, as they can play either complementary or opposing functions. Thus, levels of other intermediate filament proteins in the tested system must be examined to ensure the exact contribution by K19. In addition, the choice of cell line, methods of modulating K19 levels, and/or contexts under which assays were performed may have contributed to discrepancies from different studies.
Our findings uncovered the role of K19 in a key molecular mechanism governing cell cycle progression. Specifically, we identified that KRT19 KO cells harbor decreased levels of a transcription factor critical for an exit out of G0 phase, E2F1 43 , and its downstream targets. While decreased levels of E2F1 targets may be due in major part to decreased Rb phosphorylation, it is interesting to note that decreased E2F1 level was observed at both mRNA and protein levels. While it is unclear how K19 would regulate E2F1 expression, keratins have been shown to contribute towards gene expression both transcriptionally 44 and post-transcriptionally 45 . Further studies will be required to define the exact mechanism of K19-dependent E2F1 expression and determine whether K19-dependent regulation of E2F1 targets is due more to the decreased E2F1 level or decreased activity involving phosphorylated Rb.
Besides E2F1, levels of cyclins and CDKs were also found to be K19-dependent. Since CDK4 undergoes proteasome-dependent degradation during growth arrest 46 , K19 may be involved in maintaining the stability of CDKs and cyclins. Indeed, our data demonstrate that K19 interacts with cyclin D3 and regulates its protein stability. However, many proteins governing the G1/S phase transition including cyclin D1 47 and cyclin D3 48 are themselves E2F1 target genes. Thus, decreased levels of these proteins may also be contributed by decreased E2F1 activity. Future experiments will be needed to decipher the exact mechanism by which K19 contributes to the expression of cyclins and CDKs. Nevertheless, our findings suggest that K19 serves to regulate CDK activation and subsequent regulation of Rb/E2F1 pathway for the transition to S phase.
Nucleoshuttling of cell cycle regulators plays a key role in their functions 49 . Since K19 complexes with cyclin D3 (Fig. 7), K19 in the cytoplasm may provide a scaffold for cyclin D3, regulating its nucleoshuttling dynamics and cascade of downstream events. In support of this, keratins bind to and regulate subcellular localization of many nucleoshuttling proteins including 14-3-3 which plays a part in regulating cell cycle progression 1 . Thus, cytoplasmic and nuclear localization of cell cycle regulators including cyclin D3 may provide important clues toward how K19 regulates cell cycle progression.
Our data also suggest that K19 may be required for cell cycle progression during G2/M progression as decreased levels of cyclin B1 and CDK1 were observed. This may involve one or both of the following mechanisms: (1) decreased levels of cyclin B1 and CDK1 may simply be due to downregulated E2F1 activities as both cyclin B1 50 and CDK1 51 are E2F1 target genes. It should also be noted that CDK1 regulates other cell cycle processes including G1/S transition 52 . (2) Alternatively, K19 may play a role of regulating levels and activities of cyclin B1 and CDK1 through CDK7. Since inactivation of CDK7 leads to decreased levels of CDK1 53 , K19-dependent expression of CDK1 may involve CDK7 activity. Indeed, the fact that a loss of K19 resulted in decreased sensitivity towards CDK7 inhibitor supports this notion.
In the future, K19 may ultimately be used as a prognostic indicator for cancer patients. Compared with hormonal-based therapy or chemotherapy alone, the addition of palbociclib or ribociclib has shown better mean progression-free survival rates among ER+, advanced breast cancer patients 54,55 . Since ER+ breast cancer shows higher levels of K19 relative to other subtypes, and KRT19 KO cells exhibit decreased sensitivity to CDK inhibitors, decreased levels of K19 may underlie de novo resistance to CDK inhibition observed in clinic 56 . Given the fact that K19 is already one of the most sensitive biomarkers of breast cancer 9,13 , it may be used to develop therapeutic strategies whereby clinicians make more informed decisions on which patients should receive CDK inhibitor treatments for improving therapeutic efficacy.
MTT assay. 1000 cells were plated into each well of 96 well plate and grown in 37 °C with 5% of CO 2 and 95% of relative humidity condition. On the day of experiment, cells were incubated with 0.5 mg/ml of MTT [3-(4,5dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] (Alfa Aesar, Haverhill, MA) containing media for 3.5 h, and formed formazan crystals were dissolved with 150 µL of isopropyl alcohol at 4 mM HCl, 0.1% NP40. The absorbance of plate was then measured at 570 nm on SpectraMax microplate reader (Molecular Devices, San Jose, CA) and the data results were processed on a SoftMax Pro software (Molecular Devices). To determine the effect of CDK inhibitors on cell viability, cells were treated with DMSO control, 250 nM ribociclib, 500 nM palbociclib, or 25 nM THZ1 24 hrs after plating. Cells were then grown for 72 h and MTT assay performed. Absorbance readings from drug-treated cells were normalized to DMSO control to calculate cell viability.
MCF7 RNA-seq and bioinformatical analyses. RNA from three biological replicates each of parental and KRT19 KO cells were isolated using TRIzol reagent according to the manufacturer's instructions. Ribosomal RNA was depleted of using the NEBNext ® rRNA Depletion Kit and cDNA libraries were prepared using the NEBNext ® Ultra ™ Directional RNA Library Prep Kit for Illumina ® (NEB, Ipswich MA). RNA was barcoded using the NEBNext Multiplex Oligos for Illumina (NEB). All samples were multiplexed and sequenced on the Illumina HiSeq 3000 platform using 50 cycles single-end sequencing. Reads were aligned to human genome version hg19 using TopHat2 58 . Cufflinks and Cuffdiff were used to quantify transcripts and determine differential expression. PCA was performed using the PARTEK suit.
Cell cycle analysis. 10 5 cells were seeded into a 10 cm petri dish and grown to early log phase, after which growth media was replaced with 0.1% serum-containing media for 48 h. After serum starvation, cells were induced with 10% serum-containing media for 24 h. After exposure, cells were harvested and washed twice with PBS before fixation with ice-cold 70% EtOH and stored overnight at 4 °C. Cell aliquots were then washed twice with PBS before incubation with staining solution (0.1% Triton X-100 (Sigma-Aldrich), 0.2 μg/ml DNase-free RNase A (Sigma-Aldrich) and 20 µg/ml propidium iodide (Sigma-Aldrich) in PBS) for 15 min at 37 °C. DNA content of 10,000 events (cells)/sample was analyzed using a Becton Dickinson FACSCanto ™ flow cytometry system (Franklin Lakes, NJ) and the CELLQuest software version provided by the manufacturer. Cell cycle analysis was carried out using FlowJo V10 Software.
RNA harvest, cDNA synthesis and qRT-PCR. RNA was harvested using Direct-Zol RNA MiniPrep Plus (Zymo Research, Irvine, CA) following the manufacturers' protocols. RNA was reverse-transcribed with the iScript cDNA Synthesis Kit (Bio-Rad Laboratories, Hercules, CA) using the manufacturer's protocol. qRT-PCR was performed on the first strand cDNA with primers and PerfeCTa ® SYBR ® Green FastMix ® , ROX ™ (Quanta bio, Beverly, MA) using the Applied Biosystems StepOne ™ Real-Time PCR Systems.
Antibodies and other reagents. Preparation of cell lysates, protein gel electrophoresis, and immunoblotting. Cells grown on tissue culture plates were washed with PBS and prepared in cold Triton lysis buffer (1% Triton X-100, 40 mM HEPES (pH 7.5), 120 mM sodium chloride, 1 mM EDTA, 1 mM phenyl methylsulfonyl fluoride, 10 mM sodium pyrophosphate, 1 μg/ml each of cymostatin, leupeptin and pepstatin, 10 μg/ml each of aprotinin and benzamidine, 2 μg/ml antipain, 1 mM sodium orthovanadate, 50 mM sodium fluoride). For immunoblotting, cell lysates were centrifuged to remove cell debris. Protein concentration was determined using the Bio-Rad Protein Assay (Bio-Rad) with BSA as standard then were prepared in Laemmli SDS-PAGE sample buffer. Aliquots of protein lysate were resolved by SDS-PAGE, transferred to nitrocellulose membranes (0.45 μm) (BioRad, Hercules, CA) and immunoblotted with the indicated antibodies, followed by horseradish peroxidase-conjugated goat anti-mouse or goat anti-rabbit IgG (Sigma-Aldrich) and Amersham ECL Select Western Blotting Detection Reagent or Pierce ECL Western Blotting Substrate (Thermo Scientific, Hudson, NH). Signals were detected using ChemiDoc Touch Imager (Bio-Rad). For Western blot signal quantitation, the Image Lab software (Bio-Rad) was used.

Immunohistochemistry. Tissue sections from 21 differerent breast cancer patients with tumors of
Nottingham scores 9/High grade 3 were obtained at MedStar Georgetown University Hospital. Formalin-fixed paraffin embedded tissue sections were immunostained on the DAKO Autostainers Link 48 (Dako/Agilent Technologies, Carpenteria, CA) using the DAKO Envision Flex -System HRP along with the mouse monoclonal K19 (RCK108) (Dako/Agilent Technologies) and the rabbit monoclonal cyclinD1(EP12) (Dako/Agilent Technologies) ready to use antibodies. The respective K19 and cyclin D1 membraneous and nuclear immunoreactivity was scored based on the distribution/percentage of positive cells in both the invasive tumor and adjacent benign epithelium in each case. Negative cases are cases with absolutely no immunoreactivity, whereas cases with a percentage of positive cells ≤50% are scored as low to moderate positive and 51-100% as high positive. MedStar Georgetown University Hospital approved the experiments and all experiments were performed in accordance with guidelines and regulations of Institutional Review Board (IRB) approval. Patients were not required to give informed consent in this study because the analysis used anonymized clinical data that were obtained through a retrospective review of charts (IRB approval: IRB-Exemption-Number 2017-0029). The authors would like to thank the patients and their families for providing the tissue samples with the IRB approval.
Graphs and statistics. All graphs in the manuscript are shown as mean ± standard error of means (SEM).
For comparisons between two data sets, Student's t-test (tails = 2, type = 1) was used, and statistically significant p-values are indicated in Figures and Figure legends with asterisks (*p < 0.05). For the comparison of human tissue staining in Fig. 6e,f, p values were calculated using Fischer's exact test with two tails.

Data Availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.