The ovarian carcinoma cell line A2780 was obtained from the American Type Culture Collection (ATCC) and was authenticated by the authors within the last six months. Cell lines were maintained in RPMI culture medium supplemented with 10% FBS and 1% glutamine, at 37°C with 5% CO₂. A2780 cells were initially transfected with CRISPR/Cas9 plasmid directed against exon 1 of CDK12 (U6gRNA-Cas9-2A-GFP, Sigma-Aldrich). After 48 hours, cells were seeded at very low density in order to isolate, expand and analyze single clones by western blot and DNA sequencing. We obtained only a CDK12 heterozygous KO clone, out of about 150 clones screened. On this clone, we ran a second round of transfection and selection and obtained an A2780 homozygous KO clone (A2780 KO).

Total DNA was purified from cells (Maxwell Total DNA Purification Kit, Promega) and the selected CDK12 locus was amplified with a polymerase chain reaction (PCR) ( Supplementary Table 1 ). Amplified DNAs were separated through electrophoresis, and Sanger-sequenced.

Growth curves were obtained seeding the cells at 15000 cells/mL in six-well plates and counting them at different time points with a cell counter (Multisizer 3, Beckman Coulter). For clonogenic assay in six-well plates, cells were seeded at 125 cells/mL; colonies were left to grow for about ten days then stained with Gram’s Crystal Violet solution (Merck). For the limiting dilution assay, cells were seeded at 0.5 cells/well in 96-well plates and colonies were left to grow for about 30 days.

Exponentially growing cells were washed twice in ice-cold PBS, fixed in ice-cold 70% ethanol, washed in PBS, re-suspended in 2 mL of a solution containing 25 μg/mL of propidium iodide in PBS and 25 μL of RNase 1 mg/mL in water, and stained overnight at 4°C in the dark. We used the FACS Calibur (Becton Dickinson) for cell cycle analysis.

Exponentially growing A2780, and A2780 CDK12 KO cells were treated for 6 hours with colcemid (Roche) 0.2 μg/mL in order to block them in metaphase, then treated with hypotonic solution (75 mM KCl) and fixed with 3:1 (vol/vol) methanol:acetic acid. We karyotyped 19 metaphases for A2780 and fifteen for A2780 CDK12 KO cells lines. Chromosome preparations were analyzed by QFQ banding (Q-Bands by Fluorescence using Quinacrine) according to routine procedures and the karyotype was described using the International System for Chromosome Nomenclature 2016.

To investigate apoptosis in the cells, we used a caspase 3/7 luminescence-based assay (Promega) and the β-galactosidase staining assay (Cell Signaling Technology) for the detection of senescent cells.

Proteins were extracted and processed as already described (24). The primary antibodies used are listed in Supplementary Table 2 . All the protein blots for western analysis have been cropped before antibody hybridization to be able to detect in the same filter different proteins.

Total RNA from cell lines was purified with the Maxwell 16 Total RNA Purification Kit (Promega) and reverse transcribed with the cDNA Archive Kit (Applied Biosystems). mRNA expression of the genes was detected using Sybr Green assays (Applied Biosystems). Reactions were run in a total volume of 20 μL containing 10 ng of cDNA with Sybr Green and the forward and reverse primers of the gene, in triplicate ( Supplementary Table 1 ). All the data were normalized to the levels of the cyclophillin gene and expressed as the -fold increase over the wild type CDK12 cells.

As already described (25), NEBNext Ultra Directional RNA Library Prep Kit for Illumina (New England BioLabs Inc., Ipswich, MA, USA) was used for RNA-seq experiments and the NEBNext Multiplex Oligos for Illumina (New England BioLabs Inc.) for cDNA synthesis. The pre-pool sequencing was done using the NextSeq 500 (Illumina, San Diego, CA, USA) with the NextSeq 500/550 High Output Kit v2.5 (75 cycles; Illumina) and for all the samples we used stranded, single-ended 75bp-long sequencing reads.

Cisplatin was purchased from Sigma Aldrich; paclitaxel from ChemieTek; olaparib from LC Laboratories; VE822, MK1775 and KU55933 from Axon Medchem; THZ1 and THZ1 hydrochloride were from Insight Biotechnology Limited; ET-743 from PharmaMar and PF-00477736 from Pfizer. For cytotoxicity experiments, cell lines were seeded in 96-well plates and were treated after 48 hours with different concentrations of the drugs. After 72 hours cell viability was examined with the MTS assay (Promega) and absorbance was acquired using a plate reader (Infinite M200, TECAN). Drug concentrations inhibiting growth in 50% of the cells (IC50s) were calculated for each cell line, with the interpolation method.

The cell lines were tested for their ability to migrate through a Nucleopore Track-Etch Membrane (pore size 8 μm, Whatman International), coated with Matrigel Basement Membrane Matrix for chemo-invasion, using Boyden chambers. We added 05 to 2 x 10⁴ cells, re-suspended in DMEM culture medium with 0.1% FBS, to each well of the upper chamber for chemotaxis and chemo-invasion tests; DMEM+0.1% FBS (control wells) or 3T3 cell line supernatant (chemo-attractive agent) were added to the lower chamber. The chambers were incubated at 37°C overnight, then the cells from the upper side of the filter were removed and filters were fixed in methanol and stained for nuclear and cytoplasmic detection. Cells that had migrated the lower side of the filters were quantified by brightfield microscopy using a 40× objective, and the average number of cells per field was calculated.

NCr-nu/nu mice (five-week-old females) were from Harlan S.p.a Italy and maintained under standard pathogen-free conditions. The Istituto di Ricerche Farmacologiche Mario Negri IRCCS adheres to national and international laws, regulations, and policies on the maintenance, care and use of laboratory animals, as already reported (26). The in vivo experiments were approved by an institutional review board and the Italian Ministry of Health (Authorization no. 705/2016-PR). Exponentially growing A2780 and A2780 CDK12 KO cells (approximately 7.5 x 10⁶ cells per mouse) were injected subcutaneously in the flank of four mice per group. A Vernier caliper was used to measure tumor diameters, and tumor volumes were calculated following the formula: [(smallest diameter)² x biggest diameter]/2. Tumor and body weights were recorded at different time points from implant and mice were euthanized by carbon dioxide (CO2) overdose when tumors reached 10% of the animal’s body weight.

Statistical analyses were done with GraphPad Prism software. We used a t-test to compare cell growth, colony formation, IC50s of cytotoxic experiments and chemo-attraction and chemo-invasion data between the CDK12 wild type (WT) and CDK12 KO cell lines. To compare the activation of apoptosis between CDK12 WT and CDK12 KO cells we used two-way ANOVA followed by a Bonferroni test.

RNA-Seq was analyzed as previously described (27). Briefly, after quality control using fastqc and mapping against the human GRCh38 genome build, alignment was done with STAR, reads counted with HTSeq-Count, and differently expressed transcripts defined with the voom/limma R package. Genes were considered up-regulated or down-regulated if the adjusted p-value was <0.05 and log2 (KO vs WT) was respectively >1 or <-1. Genes deregulated in both cell lines were analyzed for enrichment in cancer hallmarks using the web-based tool of the Molecular Signaling Database (MsigDB, http://software.broadinstitute.org/gsea/msigdb), filtering for a false discovery rate (FDR) <0.05.

A2780 cells were selected to generate KO clones for CDK12 with the CRISPR/Cas9 gene editing tool. As already reported (28), this was not very successful as we obtained only one clone from A2780 (A2780 KO). Sequencing of genomic DNA showed the A2780 KO clone was homozygous in exon 1 of CDK12, carrying one allele harboring a deletion of 30 base pairs and one containing an insertion of 270 base pairs ( Figure 1A ). These data were confirmed by the cDNA sequencing of the CDK12 KO cells (data not shown).

We tested whether our guide RNAs could be directed against other genes besides CDK12 (off-target genes). We used the online CRISPR Design tool by Zhang laboratory (http://crispr.mit.edu/) and found no off-target sequence in other genes.

Alterations in CDK12 sequence are predicted to code for probably non-functional proteins. In fact, deletion in one allele (30 bp deletion) would produce a protein lacking 10 AA; the other allele has a 270 bp insertion that causes a number of stop codons, implying that no full protein should be formed. No expression of CDK12 protein in A2780 KO cells was observed, using both a C-terminal and a N-terminal epitope directed CDK12 antibody ( Figure 1B ). We carefully checked for extra bands (possibly CDK12) with different sizes, but we never found any. Considering that one allele is predicted to produce a protein lacking 10 AA, the fact that we could not find any CDK12 protein by western blot suggests that the encoded protein is very unstable.

We checked the levels of a number of proteins that have been associated and regulated by CDK12 (7, 29). We found that the basal level of RNA polymerase II phosphorylation was similar in WT and KO CDK12, while slightly lower levels of the CDK12 cyclin partner cyclin K ( Figures 1B, C ). The levels of CDK9, the other transcriptional CDK involved in Ser2 RNA polymerase II phosphorylation, were not changed, nor were those of CDK13 ( Figure 1B ). p53 protein levels were similar in A2780 and the corresponding CDK12 KO clone ( Figure 1B ).

CDK12 KO cells were morphologically indistinguishable from their parental counterparts (data not shown). We characterized the clone for its proliferation rate and clonogenicity. KO clones had slower growth than their parental cells ( Figure 2A ). Colony and limiting dilution assays suggested that A2780 CDK12 KO cells had lower clonogenic ability than parental cells. CDK12 KO cells did have less colony formation ability ( Figure 2B ) and less single cell growth in the limiting dilution assay ( Figure 2C ). We wondered whether the absence of CDK12 led to an increase in cell death that could partly explain the differences in cell growth. As shown in Figure 2D , the percentage of cells undergoing apoptosis was higher in CDK12 KO cells than in their parental cell line at different times. There was no activation of autophagy, as demonstrated by the absence of a cleaved form of LC3 protein and the increase in p62 ( Supplementary Figure 1A ), or senescence, as indicated by the absence of β-galactosidase staining ( Supplementary Figure 1B ) in CDK12 KO cells.

A2780 and CDK12 KO cells were analyzed for their DNA content and distribution in the different phases of the cell cycle. A2780 and A2780 CDK12 KO cells were regularly distributed in the different phases of the cell cycle ( Figure 3A ). However, A2780 CDK12 KO cells had a twice DNA content as compared to parental cells.

Considering the recent report of an association of bi-allelic inactivation of CDK12 with tandem duplication (17), we analyzed the karyotype of the cells to investigate the number and structure of the chromosomes. The composite reconstructions of WT and KO karyotypes ( Figure 3B ) showed that structural rearrangements of A2780 were maintained in A2780 CDK12 KO cells, but duplicated. The A2780 cell line (upper panel) showed a modal number of 46 chromosomes, ranging from 44 to 47 corresponding to a diploid DNA content. Its composite karyotype is 44~47,XX,-1x2,+der (1) dup (1q),+del(1)(p36.3p32),+der(1)t(1);?(q11);?x2,-4,der(6)t(1;6)(q21.2;q22),7, der(7)add(7)(q32), der (12)t(?;12),del(14)(q23),der(21)add(21)(q22). The A2780 CDK12 KO cells line had a modal number of 90 chromosomes, ranging from 88 to 95 (lower panel), in line with the doubled DNA content shown by FACS analysis. Its composite karyotype is 88~95,XXXX,-1x4, +der(1)dup(1q)x3, +del(1)(p36.3p32)x2,+der(1)t(1);?(q11);?x4,+3,-4,+4,+5,der(6)t(1;6)(q21.2;q22)x2,-7x2,der(7) add (7) (q32) x 2, -10, der (12) t (?;12) x 2, -13x2, der (13;14)(q10;q10),-14,del(14)(q23)x2,-17,-18,-19,-21, der (21)add(21)(q22)x2.

We transplanted parental and CDK12 KO cells in nude mice and examined their tumor take and tumor growth. The A2780 KO tumor lag was much longer than A2780, all the mice having a palpable tumor by day 135, while all the animals transplanted with A2780 cells were dead with tumors by day 25 ( Table 1 ). Tumor growth rates of A2780 CDK12 KO cells were slightly lower than A2780 cells ( Supplementary Figure 2 ).

CDK12 was reported to modulate the alternative splicing of specific genes, increasing the tumorigenicity and aggressiveness of breast cancer cells (30). We therefore investigated whether CDK12 KO cells showed a difference in migrating from their parental cells. In chemotaxis experiments ( Supplementary Figure 3 ) there was no difference in migration between parental and CDK12 KO clone. In wound healing assay, we found no differences in the ability to repair the wound in CDK12 KO cells compared to the parental cells (data not shown).

CDK12 down-regulation has been reported to be associated with lower mRNA and protein levels of DNA repair genes (4). We investigated the levels of a number of DNA damage genes in CDK12 KO cells compared to those in parental cells. Figure 4A shows the relative mRNA levels of some of the genes acting in DNA damage and repair, expressed as the fold of increase over the parental cells. There was a partial downregulation of BRCA1, CHK1, and WEE1 in A2780 KO cells, while PARP1 was downregulated. Protein levels of the same genes ( Figures 4B, D ) assessed by western blot, confirmed the downregulation of PARP1, while no significant change of BRCA1, CHK1, and WEE1 protein levels. We then checked for the level of endogenous DNA damage evaluating the activation of H2AX, a bone fide marker of this. No increase in p-Ser139 H2AX was detected in CDK12 KO cells compared to the WT cells, suggesting no greater DNA damage in CDK12 KO cells ( Figure 4C ).

Considering the reported role of the cyclin K/CDK12 complex in eukaryotic gene expression and to gain a broader picture of changes in gene expression, we compared the baseline gene expression profiles in KO versus WT cells, as described in Materials and Methods. In CDK12 KO cells, 6.8% of genes were up-regulated in A2780 KO and 7.6% were down-regulated ( Supplementary Table 3 ). We did not find the reported (4) enrichment in DNA repair genes in A2780 CDK12 KO cells. However, among the up-regulated genes there was significant enrichment of genes involved in apoptosis, in agreement with the experimental data ( Figure 2D ) ( Supplementary Table 4 ).

Recent reports suggest that cells with CDK12 mutations and/or transient downregulation with small interference RNA are more sensitive to alkylating agents, PARP inhibitors and irinotecan (11, 12, 31). We pharmacologically characterized A2780 CDK12 KO cells compared to parental cells using agents with different mechanisms of action; DNA interfering agents: cisplatin (DDP), ET-743, olaparib (OLA); antitubulin agents: paclitaxel (PTX), and agents targeting the DNA damage response pathway: ATR inhibitor (VE822), ATM inhibitor (KU55933), CHK1 inhibitor (PF477736), WEE1 inhibitor (MK1775) and CDK7/CDK12 inhibitors (THZ1 and THZ1 HYDRO). Contrary to some reports, KO clones were not more sensitive to DDP and OLA; the patterns of sensitivity were also similar with ET-743, PTX, KU55933, PF477736 and MK1775. CDK12 KO cells showed the same sensitivity to THZ1 and THZ1 HYDRO as WT cells ( Figure 5 ). Interestingly, CDK12 KO cells were five fold more resistant to the ATR inhibitor VE822 with an IC50 value of 1.99 + 0.81μM versus 0.39 + 0.07 μM ( Figure 5 ; Supplementary Table 5 ).

To see whether the cells’ sensitivities to DDP and VE822 were due to some changes in apoptosis, we ran an apoptosis test 24, 48 and 72 hours after treatment with different drug concentrations. Supplementary Figure 4 shows the apoptotic levels of treated and untreated A2780 and A2780 CDK12 KO cells. There were no differences in DDP-induced apoptosis between the WT and CDK12 KO cells, while VE822 caused weaker induction of apoptosis in KO than in WT cells. These data suggest that the lower sensitivity of A2780 CDK12 KO to VE822 than A2780 may be partly due to a lower induction of apoptosis.