As lnc-FANCI-2 expression is up-regulated along with cervical lesion progression by high-risk HPV infections [8], we further verified the increased expression of lnc-FANCI-2 in cervical cancer tissues by RNAscope single molecule RNA in situ hybridization (RNA-ISH) using a lnc-FANCI-2 antisense probe spanning over the major isoform of lnc-FANCI-2 nt 359-1713 region (GenBank MT669800.1). We found both cytoplasmic and nuclear locations of the increased lnc-FANCI-2 ISH signals within the tumor nest in HPV16-infected cervical squamous cell carcinoma tissues, but not much so in its adjacent tissue areas (Fig. 1A/B). The increased expression of lnc-FANCI-2 became obvious in human foreskin keratinocytes with HPV16 or HPV18 infection when compared to the HFK cells without HPV infection (Fig. 1C).

However, we subsequently demonstrated the high expression of lnc-FANCI-2 in two HPV16-infected cervical cancer cell lines (SiHa and CaSki) and two HPV16-infected low-grade cervical lesion-derived cervical keratinocyte subclone lines (W12 20861 and W12 20863 cells), but not in two HPV18-infected cervical cancer cell lines (HeLa and C4II cells), nor other HPV-negative cell lines of HCT116 (colorectal cancer cells), BCBL-1 (body cavity B lymphoma cells), HEK293 (Ad5 E1/E2-immortalized human kidney cells), and HaCaT (spontaneously immortalized human epidermal cells) (Fig. S1A). Interestingly, the high expression of lnc-FANCI-2 was observed in a HPV-negative cervical cancer cell line C33A cells with mutations of both p53 and RB genes [41] (Fig. S1A). Northern blot confirmed the increased expression of lnc-FANCI-2 in C33A cells but no expression in HeLa cells (Fig. S1B).

Subcellular distribution of lnc-FANCI-2 was characterized by cell fractionation and Western blot assays using nuclear SRSF3 and cytoplasmic GAPDH protein as an indication of fractionation efficiency (Fig. S1C). By RT-qPCR analysis of the fractionated total RNA, we demonstrated that lnc-FANCI-2 RNA is mainly cytoplasmic in HPV16-positive CaSki cells, but nuclear in HPV16-positive SiHa and HPV-negative C33A cells (Fig. S1C). The differential subcellular distributions of lnc-FANCI-2 RNA from CaSki to SiHa and C33A cells (Fig. S1D) were further confirmed by RNAscope RNA-ISH using the lnc-FANCI-2 antisense probe described above.

lnc-FANCI-2 RNA is transcribed mainly from a proximal promoter TSS2, but also from an alternative minor, distal promoter TSS1. Two highly conserved YY1 binding motifs upstream of the TSS2 are essential for the TSS2 transcriptional activity [8] (Fig. 2A). To elucidate the function of lnc-FANCI-2 in high-risk HPV-infected cervical cancer cells, we knocked out (KO) lnc-FANCI-2 expression in HPV16-positive CaSki cells using CRISPR/Cas9 deletion of either a 3-kb promoter region encompassing both TSS1 and TSS2 or an 86-bp region containing two YY1-binding motifs in the TSS2 promoter (Fig. 2A). Two tested gRNAs with high KO efficiency were selected and cloned into a modified CRISPR/Cas9 expression vector to express both 5ʹ-specific gRNA and 3’-specific gRNA simultaneously for efficient genome editing [42]. CaSki cells stably transfected with the dual gRNA expression vector were generated. Through a serial dilution, several single cell clones were isolated and verified for homozygous deletion by PCR screening. PCR genotyping indicated successful homozygous deletion of the promoter or YY1-binding motifs, respectively (Fig. 2B).

Loss of lnc-FANCI-2 expression in the single-cell clones was examined by RT-qPCR (Fig. 2C). When compared to all control clones transfected and selected from an empty vector containing no gRNA, all single cell clones with deletion of YY1-binding motifs (ΔYY1) showed 70-80% reduction of lnc-FANCI-2 expression, whereas the single cell clones with deletion of both TSS1 and TSS2 promoters (ΔPr) showed >90% reduction (Fig. 2C). Northern blot analysis confirmed the decrease in levels of the two major lnc-FANCI-2 isoforms (Fig. 2D), with an abundant 4-kb lnc-FANCI-2 RNA derived from a distal polyadenylation site pA2 and a less abundant 2-kb of lnc-FANCI-2 RNA derived from a proximal polyadenylation site pA1[8] (Fig. 2A and 2D). Reduced lnc-FANCI-2 RNA expression was also confirmed by RNAscope RNA-ISH (Fig. 2E and 2F). The parental (WT) CaSki cells expressed ∼10 copies of lnc-FANCI-2 RNA per cell [8], whereas the copy number of lnc-FANCI-2 RNA in the ΔYY1-D5 cells dropped to ∼5 copies and to ∼2 copies per cell in the ΔPr-A9 cells (each averaged from 200 cells).

Although the WT CaSki cells preferentially grow as distinct cluster cell islands, the lnc-FANCI-2 KO cells displayed a dispersed cell growth pattern and some cells exhibited an irregular or spindle-like cell morphology (Fig. S2). The KO cells also grew slower than the parental CaSki cells (Fig. 1G), formed fewer colonies with reduced size (Fig. 1H), and showed decreased migration capacities (Fig. 1I), with ΔPr-A9 cells having a more severe phenotype than ΔYY1-D5 cells. Subsequently, the ΔPr-A9 and ΔPr-B3 cells were further examined for their HPV16 E6 and E7 expression and the ΔPr-A9 for cell senescence. We found both ΔPr-A9 and ΔPr-B3 cells, when compared with parental WT CaSki cells, exhibited some small, variable changes in the expression of E6, E7, p53 (E6 downstream target) and E2F1 (E7 downstream target) proteins (Fig. S3A), most likely from sampling in the assays, but the ΔPr-A9 cells displayed significant increase in cell senescence (Fig. S3B). Altogether, the data indicate that lnc-FANCI-2 RNA is important for cell growth, colony formation, and cell migration.

Considering that the altered expression of cell membrane proteins and secreted factors in the lnc-FANCI-2 KO cells might contribute to the observed different cell morphology and growth properties, we performed a Proteome Profiler Human sReceptor Array analysis to examine possible changes in expression of 105 well-characterized soluble protein receptors being involved in cell signaling using total cell lysates and cell culture supernatants from ΔPr-A9 and WT CaSki cells.

Using 30% cut-off (FC −/+0.3), we identified from the ΔPr-A9 cell lysate nine proteins with increased expression compared to the WT CaSki cells, including PODXL2, ECM1, NECTIN2, MCAM, ADAM9, ADAM10, CDH5, ITGA5 and NOTCH1, and six proteins with decreased expression including ITGB6, CDH13, LGALS3BP, TIMP2, ADAM8 and SCARF2 (Fig. 3A and 3B). We also found from the ΔPr-A9 cell culture supernatant five proteins with increased expression, including ADAM9, NECTIN2, ADAM10, CDH5 and ECM1, and the decreased expression of four proteins, including CRELD2, SDC1, SDC4 and TIMP2 (Fig. 3A and 3B). By immunoblot assays, we verified selectively the increase in PODXL2, MCAM, and ECM1 and the decrease in ADAM8 and TIMP2 from the ΔPr-A9 cell lysates and the increase in ECM1 and the decrease in TIMP2 in the ΔPr-A9 cell culture supernatant (Fig. 3C). As ADAM8 is proteolytically processed into two protein isoforms for cell adhesion [43] and migration[44], we confirmed by immunoblot the decreased expression of all three sizes of ADAM8 protein in the ΔPr-A9 cell lysate over the WT CaSki cells (Fig. 3C).

We next conducted genome-wide RNA-seq analyses of WT and lnc-FANCI-2 KO cells (four samples/group) to determine the transcriptomic consequences of lnc-FANCI-2 deficiency. We obtained ∼120 million mappable RNA reads to the human reference genome hg38 from each sample. Analysis of RNA-seq reads-coverage map by Integrative Genomics Viewer (IGV) showed ∼90% reduction of lnc-FANCI-2 expression in ΔPr-A9 cells and ∼72% decrease in ΔYY1-D5 cells compared to the WT CaSki cells (Fig. 4A), consistent with the results shown in Fig. 2C-F.

Hierarchical clustering analysis showed more global transcriptional similarity between ΔYY1-D5 and ΔPr-A9 cells (Fig. 4B). By applying a threshold of fold change (FC) ≥ 1.8 or FC ≤ −1.8 with FDR ≤ 0.01, we found 1230 genes in ΔPr-A9 and 797 genes in ΔYY1-D5 cells with significantly differential expression relative to the WT CaSki cells expressing ∼15890 genes. Nine (Italic) of 15 (60%) soluble receptor proteins in cell lysates (PODXL2, ECM1, NECTIN2, MCAM, ADAM9, CDH5, ADAM10, ITGA5, NOTCH1, SCARF2, ADAM8, TIMP2, LGALS3BP, CDH13, and ITGB6) exhibited consistent changes in expression by both RNA-seq and protein array assays. The most significantly affected genes are shown in Volcano plots (Fig. 4C and Table S1). Among these, 211 were upregulated and 189 were downregulated in both ΔPr-A9 and Δ YY1-D5 cells (Fig. 4D and Table S2). By applying more stringent criteria with FPKM ≥7 in at least one of four RNA-seq samples as a cutoff, we profiled the genes with large expression differences in both ΔPr-A9 and ΔYY1-D5 cells. The results displayed in the heatmap of Fig. 4E included 52 upregulated, and 47 downregulated genes excluding lnc-FANCI-2. Subsequently, we selectively verified by RT-PCR the decreased expression of ITGB6, NLRP2, PLAC8, PSG4, PSMB9, and SERPINB1 and the increased expression of CFH, CNTN5, EMP3, IGFBP3, PTGS2, and SERPINB2 in ΔPr-A9 cells (Fig. 4F), as well as the increased expression of IGFBP3 protein, a RAS signaling driver [30–32], in ΔPr-A9 cells by immunoblot (Fig. 4G).

By performing the Gene Set Enrichment Analysis (GSEA) on the Hallmark gene sets, which provide more refined and concise inputs for GSEA [45], we found the most significantly upregulated pathways in both ΔPr-A9 and ΔYY1-D5 cells, when compared with the WT CaSki cells, were KRAS signaling and epithelial mesenchymal transition (EMT). The most significantly downregulated pathways in both ΔPr-A9 and ΔYY1-D5 cells over the WT CaSki cells were interferon gamma (IFN-γ) and interferon alpha (IFN-α) responses (Fig. 5A). GSEA plots for ΔPr-A9 shows 29 of 111 genes, including IGFBP3 [30–32], in RAS signaling and 39 of 130 genes in EMT (Fig. 5B and 5C, Table S3) were significantly upregulated, whereas 37 of 139 genes in IFN-γ response and 24 of 71 genes in IFN-α response were significantly downregulated (Fig. 5D and 5E, Table S3). Similar enriched gene sets with increased RAS signaling and EMT and decreased IFN-γ and IFN-α responses were observed in ΔYY1-D5 cells (Fig. S4, Table S3).

Since ΔPr-A9 cells exhibit higher KO efficiency of lnc-FANCI-2 and a higher number of differentially expressed genes (DEGs) and show more severe phenotypic changes in cell proliferation, migration, and colony formation than ΔYY1-D5 cells, the ΔPr-A9 cells were primarily used for further focused studies on RAS signaling in this report.

RAS activation triggers two major downstream signal transduction pathways, Raf/Mek/Erk and PI3K/Akt (Fig. S5), to transduce signals from extracellular stimuli to the cell nucleus where specific effector genes are activated for their corresponding functions [30–32, 46, 47]. To further confirm the GSEA data and verify that two major signaling pathways of RAS are activated, we first examined and compared the RAS GTPase activities in ΔPr-A9 cells and the WT CaSki cells. We found a significant increase in RAS GTPase activity in ΔPr-A9 cells (Fig. 6A, top bar graphs), indicating that the endogenous lnc-FANCI-2 RNA in CaSki cells suppresses not only the expression of IGFBP3 (Fig. 4F and 4G), the most abundant IGFBP, but also RAS activation (Fig. 6A, top bar graphs). Since ΔPr-A9 cells had undergone long term selection and adaption during single cell screening, the increased RAS GTPase activity might result from selection pressure for cell survival. However, by transient siRNA knockdown (KD) of lnc-FANCI-2 expression in the WT CaSki cells, we obtained a similar, albeit weaker, increase of RAS GTPase activity (Fig. 6A, lower bar graphs). This result suggests that the increased RAS GTPase activity in ΔPr-A9 cells was unrelated to the persistent selection pressure, but rather the deficiency of lnc-FANCI-2.

Next, we examined activation of the PI3K/Akt and Raf/Mek/Erk pathways [48–50]. We observed a 3-fold increase in phosphorylated Akt (p-Akt) and 2.5-fold increase of p-Erk1/2 (p44/p42 MAPK), after being normalized to total Akt and Erk levels, in ΔPr-A9 cells over the WT CaSki cells (Fig. 6B). These results were confirmed in ΔYY1-D5 cells (Fig. S6A). The increased p-Akt and p-Erk (mostly p-Erk2/p42) was accompanied by elevated expression of MCAM, VIM, and CCND2, and decreased expression of RAC3 (Fig. 6B and Fig. S6A). These randomly chosen potential RAS effector proteins facilitate soluble receptor functions, cell proliferation, and EMT. Moreover, transient siRNA knockdown of lnc-FANCI-2 in the WT CaSki cells also led to the increased levels of p-Akt and p-Erk (Fig. 6C) and increased expression of MCAM and VIM at 48 h and 96 h post-transfection (Fig. 6C). These observations indicate that a transient loss of lnc-FANCI-2 in CaSki cells is sufficient to trigger RAS signaling. The same siRNA transfection in HeLa cells, a lnc-FANCI-2 negative cell line (Fig. S1A and S1B), exhibited no effect on p-Akt or p-Erk1/2 levels (Fig. S6B), but in SiHa cells containing mainly nuclear lnc-FANCI-2 (Fig. S1C and S1D) enhanced the expression of p-Akt and p-Erk1/2 (Fig. S6B), verifying the specific effect of lnc-FANCI-2 depletion on RAS signaling pathways independently of predominant cellular distribution of lnc-FANCI-2. The lnc-FANCI-2 KO-mediated increase in p-Akt and p-Erk was sensitive to PI3K inhibitor LY294002 [51] (Fig. 6D, lanes 4-6) and MEK1/2 inhibitor U0126 [52] (Fig. 6D, lanes 7-9), respectively. The inhibitors also decreased the expression of VIM and MCAM and cell proliferation, in particular by MEK1/2 inhibitor U0126 (Fig. 6D and 6E). The effects of LY294002 on cell proliferation were similar from ΔPr-A9 to the WT CaSki cells (Fig. 6E).

Given that IGFBP-3 is the most abundant IGFBP among all six IGFBP members in potentiation of IGF action and PI3K/AKT activities [30] and a reduced expression of IGFBP-3 mRNA level is associated with progression to cervical cancer [53], we further investigated the effect of lnc-FANCI-2 on the expression of IGFBP3 as lnc-FANCI-2 effector gene. As shown in Fig. 4F-G and Fig. S7A, the increased RNA and protein expression of IGFBP3 appears in the lnc-FANCI-2 KO ΔPr-A9 cells over the parental WT CaSki cells, indicating a suppressive effect of lnc-FANCI-2 on IGFBP3 expression. This suppressive function of lnc-FANCI-2 was further confirmed in lnc-FANCI-2 rescue experiments in the ΔPr-A9 cells (Fig. 7A and 7B). By transient expression of one major isoform of lnc-FANCI-2 RNA (a-PA2, GenBank: MT669800.1) [8] in the ΔPr-A9 cells, lnc-FANCI-2 (red) and IGFBP3 RNA (green) at 24 h post transfection were detected by RNAscope RNA-ISH using each specific antisense RNA probe (Fig. 7A). We demonstrated that the cells with rescued expression of cytoplasmic lnc-FANCI-2 RNA displayed much reduced expression of cytoplasmic IGFBP3 RNA (Fig. 7B).

As shown in Fig. 4E and Fig. S7B, the increased RNA expression of MCAM (CD146 or MUC18) [54] appears to be transcriptional or posttranscriptional in both ΔPr-A9 and ΔYY1-D5 cells. We confirmed by RT-PCR the lnc-FANCI-2 KO-mediated increase of MCAM RNA expression in ΔPr-A9 cells (Fig. 7C). A rescue experiment was further performed by transient expression of one major isoform of lnc-FANCI-2 RNA (a-PA2, GenBank: MT669800.1) [8] in ΔPr-A9 cells to determine both nuclear and cytoplasmic MCAM expression [55, 56] in individual lnc-FANCI-2 expressing cells. ΔPr-A9 cells ectopically expressing diffused cytoplasmic lnc-FANCI-2 RNA, as detected by RNAscope RNA-ISH, exhibited a marked reduction in nuclear MCAM protein (Fig. 7D-7E). Quantitative analyses of the ΔPr-A9 cells with or without transient lnc-FANCI-2 RNA expression showed significant reduction of nuclear MCAM protein in the 35 cells with rescued lnc-FANCI-2 expression when compared to the 54 cells expressing no lnc-FANCI-2 (Fig. 7F). Due to RNAscope procedures, the membrane-bound and cytoplasmic MCAM protein was mostly removed by protease III treatment when RNA-ISH was performed and thus only the nuclear signal of MCAM protein remained was detectable by IF staining. Using cell fractionation, we detected a cleaved MCAM of ∼46 kDa mainly in the nuclear fraction by Western blot (Fig. 7G). The increased nuclear MCAM expression in ΔPr-A9 cells was proportional to the cytoplasmic and total MCAM levels when compared with the WT CaSki cells (Fig. 7G). The nuclear MCAM signal appeared as a proteolytic cleavage product presumably by the increased metalloprotease ADAM9 and ADAM10 proteins and decreased metalloprotease inhibitor TIMP2 (Fig. 3), as observed in other studies [56].

The significance of an inverse correlation of lnc-FANCI-2 and MCAM in CaSki cells was further investigated by analyzing their expression in 304 cervical cancer samples from the TCGA dataset. There was a significant negative correlation in lnc-FANCI-2 (being wrongly assigned as LINC00925 by NCBI) [8] and MCAM RNA levels (Fig. 7H). The opposite effects of lnc-FANCI-2 and MCAM on cervical cancer survival (Fig. 7I and Fig. S8) show that the cervical cancer patients with a higher level of lnc-FANCI-2 [8] but a lower level of MCAM in the cervical tissue (Fig. S8 and Fig. 7I) exhibited a better survival prognosis.

To further dissect the lnc-FANCI-2-associated expression of MCAM and IGFBP3 on RAS signaling, we examined MCAM and IGFBP3 on phosphorylation of Akt and Erk1/Erk2 in CaSki cells in the presence or absence of lnc-FANCI-2. As shown in Figure 7J, we found, although KD or KO of lnc-FANCI-2 promotes phosphorylation of Akt and Erk (Fig. 7J, lane 2), that KD of MCAM expression in the parental WT CaSki cells in the absence of lnc-FANCI-2 (Fig. 7J, lane 3) or in ΔPr-A9 cells (Fig. 7J, lane 5) led to reduction of Akt and Erk phosphorylation. These data suggest that MCAM is a lnc-FANCI-2 effector but also could be a trigger of signal transduction as reported [54, 57].

IGFBP3 protein has been viewed as a RAS signaling regulator [30–32], but recent studies show that IGFBP3 has a variety of intracellular ligands involved in many unexpected functions [58, 59]. Thus, KD or KO of lnc-FANCI-2-mediated IGFBP3 expression on RAS signaling in the parental WT CaSki cells or ΔPr-A9 cells was examined by Western blot after siRNA KD of IGFBP3 expression. As shown in Figure 7K, KD of IGFBP3 expression was found to increase phosphorylation of Erk1/2 by ∼70% (lane 4) to ∼60% (lane 6), but not much so for Akt phosphorylation (lanes 4 and 6). Instead, KD of IGFBP3 expression could prevent lnc-FANCI-2 KD-enhanced Akt phosphorylation in the parental WT CaSki cells (Fig. 7K, compare lane 2 and lane 3). These data suggest a separate role of IGFBP3 on phosphorylation of Erk1/2 from Akt in the presence or absence of lnc-FANCI-2.

We speculated lnc-FANCI-2 restriction on RAS signaling through interactions with cellular proteins. Subsequently, comprehensive identification of lnc-FANCI-2-binding proteins in the parental WT CaSki cells was performed by an IRPCRP protocol (modified from published ChIRP [60]) in combination with mass spectrometry (Fig. 8A). In this protocol, the lnc-FANCI-2-binding proteins in the parental WT CaSki cells were covalently crosslinked to lnc-FANCI-2 RNA via UV irradiation and pulled down by pooled 30 antisense oligos crossing over the entire lnc-FANCI-2. The RNA extracted from the pulldowns were verified to be lnc-FANCI-2 specific by RT-PCR in the absence (-) or presence (+) of reverse transcriptase (RT) (Fig. 8B). The lnc-FANCI-2-binding proteins in the pulldowns were then subjected to LC-MS/MS analyses. We identified 32 specific lnc-FANCI-2-binding proteins, including H13, HNRH1, K1H1, MAP4K4, and RNPS1 (Fig. 8C, Table S4).

MAP4K4, a serine/threonine protein kinase, stimulates cancer cell proliferation, invasion and migration, and is recently characterized as a novel MAPK/ERK pathway regulator [61–63] and negatively regulates RAS signaling by binding to Ras p21 protein activator 1 or RASA1 [64, 65]. Although the exact MAK4K4 binding site on lnc-FANCI-2 was not explored, many enzymes turn out an RNA-binding enzyme [66, 67]. Thus, whether MAP4K4 association with lnc-FANCI-2 could regulate PI3K/Akt or MAPK/Erk signal transduction was investigated in the parental WT CaSki cells. We demonstrated that siRNA KD of MAP4K4 expression in the parental WT CaSki cells, as seen for KD of lnc-FANCI-2, led to ∼40% or more increase of phosphorylation of both Akt and Erk1/2 and increased expression of MCAM and IGFBP3 (Fig. 8D, compare lane 2 to lane 3). Interestingly, siRNA KD of MAP4K4 expression in the lnc-FANCI-2 KO ΔPr-A9 cells had no effect on expression of IGFBP3, only a minimal effect on MCAM, but strongly on pErk1/2 (Fig. S7C). The data suggest that, through RNA-protein interactions, the increased lnc-FANCI-2 RNA in cells and their association with MAP4K4 and other cellular proteins could be one arm to control RAS signaling and gene expression of its effectors (Fig. 8E).

CaSki, a HPV16-positive (HPV16⁺) human cervical cancer cell line with wild-type p53, pRb, HRAS, KRAS, and NRAS (https://depmap.org/portal/cell_line/ACH-001336?tab=mutation) was obtained from the American Type Culture Collection (ATCC, Manassas, VA), maintained in Dulbecco’s modified Eagle medium (DMEM) (Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin, and grown in a cell culture incubator at 37°C with 5% CO₂.

SiHa (HPV16⁺), HeLa (HPV18⁺), C4II (HPV18⁺), C33A (no HPV), and other HPV-negative cell lines, HEK293T (Ad5 E1/E2-immortalized human kidney cells), and HaCaT (spontaneously immortalized human epidermal cells), were all obtained from ATCC and grown in Dulbecco’s modified Eagle’s medium (DMEM) with 10% FBS at 37°C and 5% CO₂. HCT116 cell line, a colorectal cancer cell line, was a gift from Dr. Bert Vogelstein of Johns Hopkins University and was grown in McCoy’s 5A medium with 10% FBS at 37°C and in 5% CO₂. W12 subclones 20861 (HPV16⁺, integrated) and 20863 (HPV16⁺, episomal) were a gift from Dr. Paul Lambert of University of Wisconsin, Madison, and grown on mitomycin C-pretreated NIH3T3 feeder cells in F12 medium at 37°C and in 5% CO₂. BCBL-1 (body cavity B lymphoma cells with KSHV infection) was obtained from the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH and cultured in RPMI 1640 containing 10% FBS at 37°C and in 5% CO₂.

CaSki cells were treated with 20 µM phosphoinositide 3-kinase (PI3K) inhibitor, LY294002 (Cell Signaling Technology, Danvers, MA, #9901) or 10 µM mitogen-activated protein kinase kinase (MEK) inhibitor, U0126 (Cell Signaling Technology, Danvers, MA, #9903) and examined for their proliferation and viability by CCK-8 assay and RAS effector expression by Western blot.

Raft tissues with or without HPV16 or HPV18 infection and cervical cancer tissue sections from Zhejiang University Women’s hospital were described in our previous reports [8, 92].

Synthetic double-stranded siRNA targeting the lnc-FANCI-2 transcript is listed in a supplementary Table S5. Human MAP4K4 siRNA (#M-003971), human IGFBP3 siRNA (#M-004777) and non-targeting control siRNA(#D-001210-01) were purchased from Dharmacon, Inc (Lafayette, CO). Each siRNA transfection into CaSki or other type of cells was carried out by LipoJet in vitro transfection reagent (SignaGen, Frederick, MD, USA). Plasmid pHBL03 has a short isoform of lnc-FANCI-2 cDNA (lnc-FANCI-2a-PA1, GenBank Acc. No. MT669801). Overlapping PCR products generated from 5’ and 3’ RACE PCR products with a primer pair of oHBL 17 and oHBL 18 were digested and inserted into pcDNA3 at Hind III and Xba I sites. The insertion was verified by sequencing. Plasmid pHBL05 has a long isoform of lnc-FANCI-2 (lnc-FANCI-2a-PA2, GenBank Acc. No. MT669800). PCR products generated from genomic DNA with oHBL 19 and oHBL 23 were digested and swapped into pHBL03 to replace lnc-FANCI-2a-PA1 at Bsmb I and Xba I sites.

CRISPR/Cas9 system was used to knockout lnc-FANCI-2 gene from CaSki cells. Two lnc-FANCI-2 specific guide RNAs (gRNAs) were designed by an on-line CRISPR design tool developed by the Zhang lab (http://crispr.mit.edu). A modified cloning strategy was used to clone the two gRNA into pSpCas9(BB)-2A-Puro vector (Addgene, Watertown, MA, #62988) and described in our previous publication [42]. This strategy allows all three components (two gRNAs and Cas9) express from the same plasmid to ensure the transfected cells expressing two gRNAs simultaneously knocking out the targeted gene [42]. The sequences of primers to make guide RNAs were listed in Supplementary Table S5.

Two gRNAs each under control by a U6 promoter in a pSpCas9(BB)-2A-Puro vector were used to create deletion in CaSki cells. Plasmid pHBL25 which contains gRNA 1 and 6 was used to delete the entire promoter region of lnc-FANCI-2 and plasmid pHBL27 which contains gRNA 3 and 4 was used to delete only two YY1-binding motifs spanning over an 87-bp region in the lnc-FANCI-2 promoter. CaSki cells transfected with the gRNA-expressing vectors were then under 2 weeks of puromycin selection (0.1 µg/ml). The puromycin-resistant CaSki cells were diluted to a final concentration of 0.5 cells per 100 μl, plated 100 μl of diluted cells into each well of a 96-well plate, and grew under continuous purymycin selection. The colonies were inspected to ensure the clones were grown from one cell, and refed or re-plated as needed in 3 weeks of expansion.

A direct PCR was performed for number of single cell clones as described [42]. Briefly, several hundreds of cells expanded from a single cell clone were resuspended in 50 μl of phosphate-buffered saline (PBS) and then frozen-thawed on dry ice for three times. Cell lysates were treated with 4 μl of Qiagen protease (QIAGEN, Hilden, Germany, #1017782,) at 56°C for 10 min, and the protease was inactivated at 95°C for 5 mins. The resulting product was directly used for PCR screening. P1 + P5 were used for verification of promoter deletion and P3 + P4 for YY1 motifs deletion. To ensure the selected single cell clones displaying homozygous knockout of lnc-FANCI-2, total cell DNA was isolated using a QIAamp DNA Blood Minikit (QIAGEN, Hilden, Germany, #51106) and then the homozygous deletion was screened and the cells with homozygous KO were selected and confirmed by PCR. The primers used for genotyping are listed in supplementary Table S5. At least two single cell clones either with lnc-FANCI-2 promoter deletion or with YY1 motif deletion were selected. Three single cell clones from the cells transfected with an empty vector, pSpCas9(BB)-2A-Puro, were also selected and served as controls for the genotyping screening.

Cell proliferation and viability were determined by a CCK-8 assay (Dojindo Molecular Technologies, Rockville, MD). For proliferation assay, 500 µl of cell suspension (50,000 cells/well) were dispensed into individual wells in a 24-well plate. Cells were then treated with inhibitors (20 μM LY294002, or 10 μM U0126 respectively) after 24 h culture. At each time point of 24 h, 48 h and 72 h, 50 µl of CCK-8 solution was added in three wells of each group and incubated for 1 h in a cell culture incubator at 37°C. Cell proliferation was determined by absorbance at 450 nm, and normal medium was used to subtract background. Percentage of cell viability was calculated by the following formula: cell viability % = OD450 of inhibitors treated sample / OD450 of untreated sample × 100%.

CaSki Cells were seeded at 1 × 10⁴, 5 × 10³ or 1 × 10³ cells/well in 6-well plates. After 24 h of cell growth, the culture medium in each well was replaced by DMEM containing 1% methylcellulose and 10% FBS. The plates were fixed with 3.7% formaldehyde and stained with 1% crystal violet in 2 weeks. Transwell cell migration assay was performed as previously described [93]. Briefly, CaSki cells were re-suspended in serum free DMEM containing 0.1% BSA. 1 × 10⁵ cells in 100 ul were then plated on top of the filter membrane in a transwell insert (SARSTEDT, Nümbrecht, GERMANY, # 83.3932.800) and incubate for 10 min at 37 °C to allow the cells to settle down, then 600 μl of complete media containing 10% FBS were added into the bottom of the lower chamber in a 24-well plate. After 24 h in culture, the cells on the top of the insert membrane were removed, and cells on the bottom were fixed, stained by 0.2% crystal violet and then counted.

Cell senescence was determined by Cellular Senescence Assay (MilliporeSigma, Burlington, MA, # QIA117-1KIT). Briefly, CaSki Cells were plated at 2 × 10⁵ cells/well in 6-well plates and cultured for 2 days to 50% confluence. Cells were washed once with 2 ml PBS and fixed with 0.5 ml of Fixative Solution at room temperature for 10 to 15 min. Then, cells were washed twice with 2 ml PBS, and 1 ml Staining Solution was added to each well. Cells were incubated at 37°C overnight without CO₂ and protected from the light. Blue stained cells were then counted under light microscopy.

CaSki and SiHa cells were fractionated by using Nuclei EZ Prep Kit (Sigma-Aldrich, #NUC-101) following the manufacturer’s protocol. The details of the method could be found in our previous publication [94]. Western blot analysis for nuclear protein SRSF3 and cytoplasmic protein GAPDH was used to determine the fractionation efficiency. The fractionated cytoplasmic and nuclear RNAs were used for detection of lnc-FANCI-2 RNA with human GAPDH RNA serving as a control for RNA fractionation efficiency of the cytoplasmic RNAs.

Detection of lnc-FANCI-2 by RT-qPCR was performed as described [8]. Pre-designed primers for lnc-FANCI-2 are listed in Table S5. Briefly, 2 µg total RNA was converted to cDNA using Superscript First-stand Synthesis kit (Thermo Fisher Scientific, Waltham, MA, #11904018).

qPCR was performed using TaqMan gene expression Master Mix (Applied Biosystems, Waltham, MA, #4369016) on a StepOne Plus Real-Time PCR system (Applied Biosystems). TaqMan Gene Expression Assay (Applied Biosystems, #4331182) for GADPH (Hs02758991_g1) was served as an internal control. Data were plotted as fold change over the control group using the 2^-ΔΔCt method by which the data was normalized first to the values for GAPDH and then to the median value for control samples. Data are presented as a bar graph with mean ± SD for each group.

To validate lnc-FANCI-2 KO in CaSki cells, Northern blot was performed using polyA⁺ mRNA isolated from 100 µg of total CaSki RNA with PolyATtract mRNA Isolation Systems (Promega, #Z5310) or 10 µg of total CaSki total RNA as described previously [8]. Briefly, RNA samples were denatured in NorthernMax Formaldehyde loading dye (Thermo Fisher Scientific, #AM8552) at 75°C for 15 min, then separated in an 1% formaldehyde-containing agarose gel and transferred onto a GeneScreen Plus hybridization transfer membrane (Perkin Elmer, Waltham, MA, #NEF987001PK). RNAs on the membrane were crosslinked by exposing to UV light, and then prehybridized with PerfectHyb Plus hybridization buffer (Sigma-Aldrich, St. Louis, MO, #H7003) for 2 h at 42°C. Specific oligos against lnc-FANCI-2 or GAPDH listed in Table S5 were labeled with [γ-³²P]-ATP using T4 PNK (Thermo Fisher Scientific, Waltham, MA, #18004–010) and added into hybridization buffer for overnight incubation at 42°C. The membrane was then washed with 2× SSPE/0.5% SDS solution for 5 mins, followed by twice washes with 0.5× SSPE/0.5% SDS each for 15 min at 42°C and then exposed to a PhosphorImager screen.

Total RNA was from WT CaSki cells, ΔYY1-D5 and ΔPr-A9 cells were extracted using TRIzol reagent (Thermo Fisher Scientific, Waltham, MA, #15596018). Total ribo-minus RNA-seq libraries, four samples in each group, were prepared with TruSeq Stranded Total RNA Library Kit and then subjected to pair-end sequencing using Illumina-HiSeq3000/4000 platform.

The obtained reads were processed using the CCBR Pipeliner utility (https://github.com/CCBR/Pipeliner). Briefly, reads were trimmed from adapters and low-quality bases using Cutadapt (version 1.18) (https://bioweb.pasteur.fr/packages/pack@cutadapt@1.18) before alignment to the custom reference genome described below. The transcripts were aligned using STAR v2.5.2b in two-pass mode [95]. Expression levels were quantified using RSEM (RNA-Seq by Expectation-Maximization) (version 1.3.0) [96] with a custom gene annotation described below.

The custom reference genome allowing quantification of both HPV16 and host expression used in this alignment consisted of the human reference genome (hg38/Dec. 2013/GRCh38) with a HPV16 FASTA (https://pave.niaid.nih.gov/) sequence added as an additional pseudochromosome. The custom gene annotation used for gene expression quantification consisted of a concatenation of the hg38 GENCODE annotation version 30 [97] and the HPV16 genome, with one other notable alteration. The hg38 v30 gene annotation is incorrect at the location of the lnc-FANCI-2 locus in the hg38 genome at chr15:89378104-89398487, which affects quantification of this gene. To correct this annotation and provide accurate quantification, therefore, we first performed BLAST against hg38 using the sequences of the 14 known isoform transcripts of lnc-FANCI-2 (MT669800 for lnc-FANCI-2a-PA2, MT669801 for lnc-FANCI-2a-PA1, MT669802 for lnc-FANCI-2b-PA2, MT669803 for lnc-FANCI-2b-PA1, MT669804 for lnc-FANCI-2c-PA2, MT669805 for lnc-FANCI-2c-PA1, MT669806 for lnc-FANCI-2d-PA2, MT669807 for lnc-FANCI-2d-PA1, MT669808 for lnc-FANCI-2e-PA2, MT669809 for lnc-FANCI-2e-PA1, MT669810 for lnc-FANCI-2f-PA2, MT669811 for lnc-FANCI-2f-PA1, MT669812 for lnc-FANCI-2g-PA2, and MT669813 for lnc-FANCI-2g-PA1). The results were used to determine the precise exon start and stop sites for each of the isoforms. These were then used to manually adjust the default hg38 v30 gene annotations at the locus indicated above to account for the correct structure of the lnc-FANCI-2 locus and its 14 isoforms.

Downstream analysis and visualization were performed within the NIH Integrated Analysis Portal (NIDAP) using R programs developed on the Foundry platform (Palantir Technologies). Briefly, raw counts data produced by RSEM were imported into the NIDAP platform, genes were filtered for low counts (<1 cpm) and the voom algorithm [98] from the Limma R package (version 3.40.6) [99] was used for quantile normalization and calculation of differentially expressed genes. Pre-ranked GSEA was performed using the Molecular Signatures Database version 6.2 [100, 101] and the fgsea package [102]. Raw data and the analyzed RNA-seq data supporting the findings in this study have been deposited in the NCBI GEO database (the accession#: GSE190904).

To detect the changes of soluble receptors expressed and released from parental WT CaSki cells and lnc-FANCI-2 KO cells, Proteome Profiler Human sReceptor (Soluble Receptor) Array (R&D Systems, Minneapolis, MN, # ARY012) was performed according to a manufacturer protocol. Briefly, 7 × 10⁶ WT CaSki cells and ΔPr-A9 cells were cultured in 15 ml of DMEM supplemented with 10% FBS in a T75 flask for 24 h. Cell culture supernatants were then collected and centrifuged at 4,000 × g for 15 min to remove cell debris. Cells were rinsed by 10 ml of PBS and solubilized in Lysis Buffer 17 at 4°C for 30 min. The cell lysis was then centrifuged at 14,000 × g for 5 min to remove cell debris. Protein concentration of cell lysis was measured by Micro BCA Protein Assay (PIERCE, Waltham, MA, #23235). Next, two sets of N and C membranes were blocking by Buffer 8/1 in 4-Well Multi-dish for 1 h on a rocking platform. 500 μl of culture supernatant or 100 μg cell lysate generated from the WT CaSki cells or ΔPr-A9 cells were diluted in 3 ml of Buffer 8/1 and then applied to each membrane after three washes with 1× Wash Buffer. The membranes were incubated overnight at 4 °C on a rocking platform and washed three times to remove unbound materials followed by incubation with their specific cocktail of biotinylated detection antibodies for 2 h at room temperature on a rocking platform. 2 ml of diluted Streptavidin-HRP was then applied to each membrane for 30 min at room temperature on a rocking platform after three times of wash. 1 ml of the prepared Chemi Reagent Mix was added onto the membranes after three times of wash. The signal was then captured by ChemiDoc Touch Imaging System (Bio-Rad). Quantification of the relative protein expression in ΔPr-A9 cells compared to the WT cells was determined in Image Lab software (Bio-Rad). The average signal of duplicate spots subtracting background value from negative control spots represented the levels of each protein. The relative change in protein levels was then determined by comparing ΔPr-A9 cells to the WT cells.

Immunoblot analysis was performed for individual proteins by using the following antibodies: anti-MCAM (#17564-1-AP), anti-E2F1 (#12171-1-AP), anti-MAP4K4 (#55247-1-AP), and anti-p53 (#10442-1-AP) antibodies were from ProteinTech (Rosemont, IL). Anti-tubulin (#T5201) antibodies were from Sigma-Aldrich (St. Louis, MO). Anti-hnRNP C1/C2 (#Ab10294) and anti-RAC3 (#ab129062) antibodies were from Abcam (Cambridge, United Kingdom). Anti-CCND2 (Cyclin D2, #3741), anti-Akt (#9272), anti-phospho-Akt (#9271), anti-Erk1/2 (p44/42 MAPK, #4695), anti-phospho-Erk1/2 (#4370), and anti-IGFBP3 (D1U9C, #25864) were from Cell Signaling Technology (Danvers, MA). Anti-VIM (#MA5-11883) was from Thermo Fisher Scientific (Waltham, MA). Anti-PODXL2 (#AF1524), anti-ECM1 (#MAB39371), anti-TIMP2 (#AF971) and anti-ADAM8 (#AF1031) were from R&D Systems (Minneapolis, MN). Anti-SRSF3 (#NBP2-76892) was from Novus Biologicals (Centennial, CO). Rabbit polyclonal anti-HPV16 E7 (#GTX133411) was from GeneTex (Irvine, CA).

RAS GTPase activity in cell lysate of the WT CaSki or ΔPr-A9 cells was determined by RAS GTPase Chemi ELISA Kit (Active Motif, Carlsbad, CA, # 52097) according to the manufacturer protocol. Briefly, 2 × 10⁷ cells were solubilized in 500 µl of Complete Lysis/Binding buffer at 4°C for 15 min. Protein concentration of cell extract was then measured by Micro BCA Protein Assay (PIERCE, #23235) after cell debris was removed by centrifuged at 14,000 × g for 10 min. 2 µg of GST-Raf-RBD in 50 µl of Complete Lysis/Binding buffer was coated in each well of a 96-well plate by incubating for 1 h at 4°C with 100 rpm agitation. After three times of wash, 50 µg of extract in 50 µl Complete Lysis/Binding buffer was then added in each well. HeLa (EGF treated) extract was served as a positive control. The plate was incubated for 1 h at room temperature with 100 rpm agitation. The active RAS protein (GST-Raf-RBD binding) in each well was then detected by incubating with primary RAS antibody and three times of wash, and then, HRP-conjugated secondary antibody and three times of wash. Chemiluminescence in each well was read in luminometer by adding 50 µl room-temperature Chemiluminescent Working Solution.

Endogenous lnc-FANCI-2 in cervival cancer tissues and their derived cell lines was examined by RNAScope Multiplex Fluorescent V2 Assay (Advanced Cell Diagnostics, Minneapolis, MN, #323100) as described previously [8]. A custom-designed probe targeting to nt 359-1713 of GenBank Acc. No. MT669800.1 transcript for lnc-FANCI-2a-PA2 was utilized. Dual staining of lnc-FANCI-2 and IGFBP RNA was performed according to RNAscope Multplex Fluorescent v2 Manual part 2 (#323100), and the channel one probe for IGFBP3 (#310351) and channel three probe for lnc-FANCI-2 (#509061-C3) were applied.

For RNA-ISH combined IF staining of MCAM protein, 3 × 10⁵ of CaSki cells were grown on a glass coverslip in a 6-well plate for 24 h before transfection. Plasmid pHBL05 containing a long isoform of lnc-FANCI-2a-PA2 (GenBank Acc. No. MT669800.1) were transfected into cells with LipoD 293 DNA in vitro transfection reagent. Cultured Adherent Cell Sample Preparation for the RNAscope Multplex Fluorescent v2 was performed according to a manufacturer protocol (Advanced Cell Diagnostics, Minneapolis, MN, MK-50-010). Briefly, the cells were washed with PBS and fixed by 10% neutral buffered formalin at room temperature for 30 min. The cells were washed with PBS three times and dehydrated by 50% ethanol for 5 min, 70% ethanol for 5 min, 100% ethanol for 5 min, and 100% ethanol for 10 min. The cells were then rehydrated by 70% ethanol for 2 min, 50% ethanol for 2 min, and PBS for 10 min. The cells were then applied to hydrogen peroxide at RT for 10 min and washed with distilled water for three times. The cells were then applied to Protease III (1:15 dilution with PBS) at RT for 10 min. The cells were ready for RNAscope Multplex Fluorescent v2 Manual part 2 (#323100). IF was performed after RNA-ISH but before counterstaining with DAPI. The cells were washed with PBS, and then were blocked with 2% bovine serum albumin (BSA) in PBS for 1 h at 37°C or overnight at 4°C. The cells were incubated with the MCAM primary antibody (ProteinTech, Rosemont, IL, #17564-1-AP, diluted 1:200 in 2% BSA blocking buffer) for 2 h at 37°C. An Alexa Fluor-conjugated secondary antibody (1: 500, ThermoFisher Scientific) was diluted in a blocking solution and incubated for 1 h at 37°C. The slides were washed with PBS for three times. DAPI was used for nuclei counterstaining before mounting in a Prolong Gold Antifade mounting medium (Thermo Fisher Scientific, #P36934).

Confocal images were collected using a Zeiss LSM710 laser-scanning microscope equipped with a 63 × Plan-Apochromat (N.A. 1.4) objective lens. Three dimensional distributions of lnc-FANCI-2 were generated by Z stacks using ZEN 2.3 software (Zeiss).

To identify the lnc-FANCI-2 RNA-associated proteins, we pulled down lnc-FANCI-2 RNA from WT CaSki cells using IRPCRP, a modified ChIRP protocol from the published method [103], followed by mass spectrometry [60]. Briefly, a total of 30 lnc-FANCI-2 anti-sense oligo probes (oLLY496-oLLY525), each with biotinTEG at 3’ end, across entire lnc-FANCI-2 RNA (GenBank Acc. No MT669800.1) were designed using the online probe designer (singlemoleculefish.com). Two probe pools were used in IRPCRP assays: the probe pool 1 was the mixed oligos in even numbers (oLLY496 /498/500/502/504/506/508/510/512/514/516/518/520/522/524) and the probe pool 2 mixed with the oligos in odd numbers (oLLY497/499/501/503/505/507/509/511/513/515/517/519/521/523/525) (Table S5). CaSki cells were seeded and harvested at 24 h and the cell lysates from 200 million cells were used for each IRPCRP reaction. Instead of formaldehyde crosslink in the standard protocol [103], UV cross-linking (254 nm, energy at 480 mJ/cm²) was applied in our study to minimize non-specific binding. After UV cross-linking, the cells were lysed on ice in 1× radioimmunoprecipitation assay (RIPA) buffer (Boston Bioproducts, Ashland, MA, #BP-115) (50 mM Tris base/Tris-HCl [pH 7.4], 150 mM NaCl, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate [SDS], 1% NP-40) supplemented with protease inhibitors (complete mini-EDTA-free protease inhibitor cocktail, Millipore Sigma, Burlington, MA, #469315900) and RNase inhibitor (Thermo Fisher Scientific, #AM2694) for 30 min, followed by a brief sonication. The obtained cell lysates were spin for 10 min of centrifugation at 20,000 × g at 4°C and the collected cell lysate supernatants were pre-absorbed with C-1 magnetic beads (Thermo Fisher Scientific, #65002) for 1 h at room temperature before proceeding to oligo probe hybridization.

Two different hybridization protocols were used to pull down lnc-FANCI-2 RNA. One had the pre-absorbed cell lysate incubated with two separate probe pools in hybridization buffer (750 mM NaCl, 0.1% SDS, 50 mM Tris-Cl pH 7.0, 1 mM EDTA, 15% formamide) at room temperature overnight and the pre-washed C-1 magnetic beads was then added and incubated for additional 4 h; the other had two separate oligo probe pools immobilized onto pre-washed C-1 magnetic beads first at room temperature for 1 h and then mixed with the pre-absorbed cell lysates in hybridization buffer overnight at room temperature, along with the negative control-beads only IRPCRP reaction without oligo probes serving as a control in the modified protocol. After hybridization, the beads were washed with 1× wash buffer (2× NaCl and sodium citrate, 0.1% SDS) for 5 times and resuspend in 1 ml lysis buffer. 100 µl beads from each IRPCRP were set aside for RNA isolation of lnc-FANCI-2 pulldown efficiency and the leftover beads were sent for mass spectrometry analysis. The input and IRPCRP RNA were isolated using a MIRNeasy mini kit (Qiagen, #217004) following the standard protocol after proteinase K (Millipore Sigma, #71049) digestion.

For mass spectrometry analysis, the IRPCRP beads were processed for trypsin/LysC digestion before submitted to Thermo Scientific Orbitrap Exploris 240 Mass Spectrometer and a Thermo Dionex UltiMate 3000 RSLCnano System for Proteinomics. Peptides from trypsin digestion were loaded onto a peptide trap cartridge at a flow rate of 5 µl/min. The trapped peptides were eluted onto a reversed-phase Easy-Spray Column PepMap RSLC, C18, 2 µM, 100A, 75 µm × 250 mm (Thermo Scientific) using a linear gradient of acetonitrile (3-36%) in 0.1% formic acid. The elution duration was 110 min at a flow rate of 0.3 µl/min. Eluted peptides from the Easy-Spray column were ionized and sprayed into the mass spectrometer, using a Nano Easy-Spray Ion Source (Thermo Fisher Scientific) under the following settings: spray voltage, 1.6 kV, Capillary temperature, 275 °C. Other settings were empirically determined. Raw data files were searched against human protein sequences database using the Proteome Discoverer 2.4 software (Thermo Fisher Scientific, San Jose, CA) based on the SEQUEST algorithm. All the protein peptides identified in the IRPCRP pulldowns were summarized in Table S4.