Lipid kinase PIP5K1A regulates let-7 microRNA biogenesis through interacting with nuclear export protein XPO5

Abstract MicroRNAs (miRNAs) are small non-coding RNAs first discovered in Caenorhabditis elegans. The let-7 miRNA is highly conserved in sequence, biogenesis and function from C. elegans to humans. During miRNA biogenesis, XPO5-mediated nuclear export of pre-miRNAs is a rate-limiting step and, therefore, might be critical for the quantitative control of miRNA levels, yet little is known about how this is regulated. Here we show a novel role for lipid kinase PPK-1/PIP5K1A (phosphatidylinositol-4-phosphate 5-kinase) in regulating miRNA levels. We found that C. elegans PPK-1 functions in the lin-28/let-7 heterochronic pathway, which regulates the strict developmental timing of seam cells. In C. elegans and human cells, PPK-1/PIP5K1A regulates let-7 miRNA levels. We investigated the mechanism further in human cells and show that PIP5K1A interacts with nuclear export protein XPO5 in the nucleus to regulate mature miRNA levels by blocking the binding of XPO5 to pre-let-7 miRNA. Furthermore, we demonstrate that this role for PIP5K1A is kinase-independent. Our study uncovers the novel finding of a direct connection between PIP5K1A and miRNA biogenesis. Given that miRNAs are implicated in multiple diseases, including cancer, this new finding might lead to a novel therapeutic opportunity.


INTRODUCTION
MicroRN As (miRN As) such as lin-4 and let-7 are small non-coding RN As, w hich wer e first discover ed through developmental studies in Caenorhabditis elegans ( 1 , 2 ).The let-7 miRNA is highly conserved among species including humans ( 2 , 3 ).In the canonical miRN A bio genesis pathway, primary miRNAs (pri-miRNAs) are transcribed from their genetic locus by RN A pol ymerase II (Pol II) and processed into precursor miRNAs (pre-miRNAs) by the micr opr ocessor complex (consisting of Drosha and DGCR8) and exported to the cytoplasm via an exportin 5 (XPO5) / Ran-GTP complex ( 4-6 ).In the cytoplasm, the pre-miRNA is further processed by the RNase III endonuclease Dicer into small RNAs of a pproximatel y 21 nucleotides and then loaded into Argonaute proteins in the RNA-induced silencing complex (RISC) ( 4 , 5 ).Finall y, miRN As interact with the 3 untranslated region (3 UTR) of target mRNAs to in-duce mRNA degradation and translational r epr ession ( 4 , 5 ).During miRN A bio genesis, XPO5-mediated nuclear export of pre-miRNAs is a rate-limiting step for miRN A bio genesis ( 7 , 8 ) and, ther efor e, might be critical for the quantitati v e control of global miRNA le v els.
The C. elegans heter ochr onic gene pa thway regula tes the timing of de v elopmental e v ents during post-embryonic development ( 9 ).This pathway consists of genes that encode RNA-binding proteins, such as LIN-28 and LIN-41, mi-croRNAs, such as lin-4 and let-7 and transcription factors, such as lin-14, hbl-1 and lin-29 ( 1 , 2 , 10 , 11 ).C. elegans seam cells are lateral epidermal cells that divide during each larval stage with a stem-cell like pattern of cell fate and cell division ( 9 , 11 ).At each larval stage, one daughter of the cell division dif ferentia tes while the other daughter retains the ability to divide again.Around the beginning of the adult stage seam cells exit the cell cycle, fuse together and secrete a cuticular structure called alae ( 11 ).Mutations in heter ochr onic genes lead to temporal alterations to stagespecific patterns of cellular fate ( 11 ).For example, in precocious mutants, cells inappropriately express later cell fates during earl y stages, w hile in retarded m utants, cells reiterate earlier stage fates in place of later cell fates ( 12 ).These types of defects in de v elopmental timing result in easily scorable phenotypes, for example by using the cell fusion marker ajm-1::gfp (apical junction marker), adult cell fate marker col-19::gfp (adult specific marker) and cell terminal differentiation markers such as alae formation.Two major components of the heter ochr onic gene pathway, LIN-28 and let-7 miRNA, are e volutionarily conserv ed in animals where they have been shown to regulate each other's expression and have pivotal roles in pluripotency and dif ferentia tion ( 3 , 13 , 14 ).Howe v er, the mechanisms of action and downstream effectors of the lin-28 / let-7 pathway are poorly understood.Ther efor e, the identity of additional genes in the lin-28 / let-7 pathway in C. elegans will be important to better understand seam cell de v elopment and miRN A bio genesis.
In this study, we found that C. elegans PPK-1 regulates miRNA le v els and functions in the lin-28 / let-7 heter ochr onic pathway.Interestingly, we also found PIP5K1A (the ortholog of PPK-1) regulates miRNA le v els through interactions with nuclear export protein XPO5 in the nucleus, independent of its kinase activity.Furthermore, we demonstra ted tha t PIP5K1A blocks the binding of XPO5 to pre-miRNA.Ther efor e, this study describes the novel finding of a direct connection between the lipid kinase PIP5K1A and miRN A bio genesis.

RNA interference (RNAi)
RNAi experiments were performed at 20 • C using E. coli HT115 bacteria expressing RNAi constructs from the Ahringer libr ary ( 26 ).Tr ansformed bacteria were over laid on a nematode growth medium (NGM) plate containing 1mM IPTG and 50ug / ml Carbenicillin.Bacteria containing L4440 vector was used as a control.

Microscopic analysis
L4 animals (P0) were placed on ppk-1 (RNAi) plates and the F1 progeny was scored for seam cell number ( scm-1::gfp) , seam cell fusion ( ajm-1;;gfp ), col-19::gfp expression and alae formation.scm-1::gfp, ajm-1::gfp , col-19::gfp expression and alae formation were observed using an upright Zeiss Axioplan microscope under 40x or 63x magnification.Animals were immobilized using 1mM levamisole on 2% agarose pads.De v elopmental stage was assessed by vulval and gonad de v elopment using DIC microscopy.Confocal images were acquired with a Carl Zeiss LSM 880 microscope using the Zen black software version SP2.3.The tile scan and z-stack images were acquired with a Plan-Apochromat 63x / NA 1.4 objecti v e lens with 10-16 images per condition.Cross-sections of a 3D volume reconstruction were generated using the Imaris image analysis software (Bitplane).

RNA isolation and qRT-PCR
Animals were bleached (5% 5 N NaOH, 10% bleach in M9 solution) starved in M9 (without E. coli ) overnight to get synchronized L1s, which were put on plates for the experiments.qRT-PCR (Figures 1 B and 2A-H) was performed at the peak time point of lin-42 mRNA expression levels, as lin-42 mRNA is dynamically expressed during development and peaks during the L4 stage ( 27 ).
Total RNA was collected in TRIzol (Invitrogen) and isolated using PureLink ™ RNA Mini Kit (Invitrogen) or with Dir ect-zol Minipr ep Plus spin columns (Zymo Research) according to the manufacturer's protocol, including the on-column DNase I treatment.cDNA synthesis was performed using High-Capacity RNA-to-cDNA ™ Kit (Applied Biosystems) or High-Capacity cDNA Re v erse Transcription Kit (Applied Biosystems).Total RNA concentration was measured using the NanoDrop Spectrophotometer (ND-1000 Spectrophotometer).mRNA expression levels were determined by quantitati v e RT-PCR using SYBR Green (Applied Biosystems) according to manufacturer protocols (Roche).MiRNA expression levels were determined by RT-qPCR using TaqMan MicroRNA Assays (A pplied Biosystems).MiRN A le v els in animals were normalized to U18 snoRNA, while mRNA or pri-miRNA levels were normalized with pmp-3 or actin ( act-1) mRNA levels and 5.8S rRNA served as control for normalization of pre-miRNA le v el.For mammalian RNA samples, miRNA or pre-miRNA and mRNA or pri-miRNA le v els were normalized to U6 snRNA and GAPDH mRNA, respecti v ely.
Nuclear and cytoplasmic RNAs were isolated following the protocol by Gagnon et al ( 28 ) and Choudhury et al ( 29 ).Briefly, RKO cells were l ysed in ice-cold HLB (Hypotonic lysis buffer: 10 mM Tris, pH 7.5, 10 mM NaCl, 3 mM MgCl 2 , 0.3% NP-40 and 10% glycerol), supplemented with 100 U of Ribolock RNase inhibitor and incubated the mixture on ice for 10 min followed by centrifugation at 1000 g at 4 • C. for 3 min.We carefully transferred the supernatant (cytoplasmic fraction) to a new tube and kept the pellet on ice.The nuclear pellet w as w ashed with icecold HLB buffer three times and centrifuged at 300 g at 4 • C. for 2 min.TRIzol was then added to both nuclear and cytoplasmic fractions and proceeded for RNA extraction.RT-qPCR was performed using TaqMan MicroRNA assays (Applied Biosystems), as described earlier.The purity of cytoplasmic and nuclear fractions was determined using GAPDH mRNA and MALAT1 , respecti v ely.Experiments were done in triplicate.For validation of the percent of prelet-7a-1 / prilet-7a-1 in the cytoplasmic-nuclear fractionation experiment, first we calculated the C / N ratio of prelet-7a-1 or prilet-7a-1 by using the 2 − Ct [2 −(Cyt(ct)-Nuc(ct)) ] method ( Ct method).From this result, we calculated the C / N ratio of prelet-7a-1 / prilet-7a-1 , to get the percent of prelet-7a-1 / prilet-7a-1 in the cytoplasmic or nuclear fractions.
Primers (SYBR Green-based qPCR) for used in this study are: TaqMan probes (Thermo Fisher Scientific) are shown below.

Northern blotting
Total RNA (20-40 ug) was extracted from nematodes or cells using TRIzol reagent (Invitrogen, USA) and northern blots performed using biotin-labeled probes following the protocol by Gagnon et al. ( 28 ).Briefly, total RNA samples were run on 15% PAGE-urea gels and transferred to Hybond-N+ positi v ely charged nylon membranes (GE healthcar e, USA) by electrophor esis.Next, the membranes were further UV-cross-linked and dried at 60 • C for 1 h to improve binding.Before hybridization, the membranes wer e pr e-hybridized for at least 1 h at 40 • C in pre-hybridization buffer (#AM8677, Thermo Scientific).Next, hybridiza tion buf fer containing 50 pmol / ml biotin labeled single-stranded DNA oligonucleotide (See below) was added and the membranes were hybridized for overnight at 40 • C with gentle shaking and subsequently rinsed with low stringency wash buffer (#AM8673, Thermo Scientific) 3 times and High stringency wash buffer 2 times (#AM8674, Thermo Scientific) at room temperature.The biotin-labeled probes were detected using a Chemiluminescent Nucleic Acid Detection Module Kit (#89880, Thermo Scientific) following the manufacturer's instructions.The bands were quantified using the ImageJ software.

Western blotting
Animals were lysed by boiling in SDS sample buffer while cells were harvested and lysed in RIPA lysis buffer (Thermo Scientific ™) supplemented with protease inhibitor cocktail (Thermo Scientific).Total protein concentration was determined using Pierce BCA protein assay kit (Thermo Scientific).Proteins resolved by SDS page were transferred to PVDF membr ane.Membr anes were blocked with 5% milk protein in 1 × TBST and incubated with primary antibodies overnight.Membranes were washed three times with 1 × TBST and probed with a HRP-conjugated secondary antibody for 1 h followed by three additional washes.Specific antibody binding onto the membranes was detected using the SuperSignal ™ West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific).

Cell lines and transfection
HEK293T (obtained from the American Type Culture Collection (ATCC)) and RKO (Provided by Dr Kevin Haigis from Dana-Farber Cancer Institute of Harvard Medical School) cell lines wer e cultur ed in DMEM (high glucose) (Gibco, Cat #11995-065) supplemented with 10% FBS (Sigma-Aldrich) and 1% penicillin-streptomycin.Each cell line was maintained in a 5% CO 2 atmosphere at 37 • C.
To generate stable knockdown cell lines, PIP5K1A ShRNA (TRCN0000231477, Millipore Sigma) and control ShRNA (SHC016, Millipore Sigma) constructs were first transfected into HEK293T cells with the VSVG envelope vector and psPAX2 packaging vector using TransIT-Lenti reagent according to manufacturer's protocol (MirusBio).After 48 h, media was collected, centrifuged, and filtered through a 0.45 m nitrocellulose filter.The virus was then added to RKO cells, and stably transduced cells were selected with puromycin and were maintained in 1.0 g / ml puromycin media.To generate PIP5K1A -WT OE or PIP5K1A -D309N OE cells, pcDNA3.1-PIP5K1A-FLAGor pcDNA3.1-PIP5K1A-D309N-FLAGwere transfected to RKO cells respecti v ely , by using T ransIT-X2 reagent (Mirus Bio, Madison, WI, USA) following the manufacturer's protocol.The plasmids and small interfering RNAs (siRNAs) were transfected to HEK293T or RKO cell lines using the TransIT-X2 reagent (Mirus Bio, Madison, WI, USA) following the manufacturer's protocol.Silencer select siRNAs wer e pur chased from Santa Cruz Biotechnology: XPO5 siRNA (sc-45569) and Control siRNA (sc-37007).

Immunofluorescent staining
Samples wer e fix ed in 4% PFA for 10 min in phosphatebuffered saline (PBS) and washed three times for 5 min in PBS at room temperature, then incubated with 3% BSA in 0.3% Triton X-100 in PBS for 30min at room temperature.Samples were incubated with primary antibodies overnight a t 4 • C , and washed three times.Later they were incubated with secondary antibodies for 1 hour at room temperature, washed and mounted with ProLong ™ Gold Antifade Mountant with DAPI ((#P36931, Thermo Fisher).Primary antibodies: PIP5K1A Polyclonal Antibody (# 15713-1-AP, Thermo Fisher scientific), XPO5 purified MaxPab mouse polyclonal antibody (B01P) (# H00057510-B01P, Abnova), Fluorescein conjugated Anti-PI ( 4 , 5 ) P2 IgM (#Z-G045, Echelon Biosciences), DYKDDDDK Tag (D6W5B) Rabbit mAb (Alexa Fluor ® 594 Conjugate) (#20861, Cell signaling).Secondary antibodies: Goat anti-Rabbit IgG Secondary Antibody Alexa Fluor 488 (# A27034, Thermo Fisher scientific), Goat anti-Mouse IgG Secondary Antibody Alexa Fluor 568 (# A-11004, Thermo Fisher scientific).Confocal images were acquired with a Carl Zeiss LSM 880 upright confocal microscope, as described above.We define the C / N ratio of subcellular localization of XPO5 as the ratio of cytoplasmic to nuclear immunofluorescence intensity of XPO5, assessed by the mean fluorescence intensity in the cytoplasmic and nucleus volumes, which was performed by ImageJ software.

RNA immunoprecipitation assay (RIP)
RKO control and PIP5K1A knockdown cells were transfected with plasmids expressing HA-XPO5.Fortyeight hours after transfection, the plates were rinsed twice in ice-cold PBS, cross-linked at 254 nm (120 mj / cm 2 ), scraped and lysed in IP lysis buffer supplemented with Protease Inhibitor Cocktail and Ribolock RNase Inhibitor (40 U / l; Thermo Fisher Scientific) on ice for 20 min and then centrifuged at 15 000 rpm speed at 4 • C for 10 min.After centrifugation, 10% of the supernatant was separated as the input for qRT-PCR analysis.The rest of the supernatants were incubated with anti-HA magnetic beads or control magnetic beads with constant rotation at 4 • C for overnight.The beads were washed 4 times with wash buffer.After washing, the beads were treated with proteinase K, and RNA was purified as described above.cDNA synthesis and PCR was performed as described above.After qRT-PCR analysis, the levels of prelet-7 and U6 in the IP samples were divided by the levels of them in the input calculated by a relati v e enrichment method, 2 (-(IP(ct)) -(input(ct) .Fold enrichment of this ratio for HA IP normalized to the negati v e control IgG IP was determined.

Statistical analysis
All experiments were performed with at least three biological replica tes.Sta tistical analysis was performed by Student's two-tailed t test in GraphPad Prism 9. Data points ar e pr esented as the mean ± SEM.P -values are: * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 and NS = not significant.

ppk-1 is a heter ochr onic gene
C. elegans heter ochr onic genes in the lin-28 / let-7 pathway regulate the strict de v elopment timing of seam cells ( 2 , 11 ).We pr eviously r eported the use of CLIP-Seq to identify 2000 mRNAs interacting directly with the LIN-28 RNA binding protein ( 30 ), some of which were known heter ochr onic genes.To enrich for additional genes in the lin-28 / let-7 pathway, we determined the overlap among our set of LIN-28 CLIP hits ( 30 ), the set of 201 known let-7 suppressors ( 31 ), 213 known let-7 enhancers ( 32 ), and an analysis of lin-28 dependent genes from our cursory RNA-seq data (Supplementary Table S1) from staged late larval stage 1 (L1) WT and lin-28 ( n719 ) mutant animals.We found 16 candida te genes tha t are a t the intersection of these groups (Supplementary Figure S1).Of these genes, we focused on ppk-1 , an ortholog of human PIP5K1A (phospha tid ylinositol-4-phospha te 5-kinase type 1 alpha).While ppk-1 was identified in the 213 known let-7 enhancers ( 32 ) group, in that study, it did not show a strict positi v e relationship with let-7 , as the vulval burst score was only 1.1 ( 32 ).PPK-1 / PIP5K1A is an enzyme which produces phospha tid ylinositol-4-phospha te (PI4P) to phospha tid ylinositol-4,5-bisphospha te (PIP2), a lipid second messenger tha t regula tes se v eral cellular processes, such as signal transduction and vesicle trafficking ( 33 ).
Since PPK-1 was identified from LIN-28 CLIP-seq analysis, we tested if LIN-28 potentially regulates ppk-1 mRNA le v els.In lin-28 ( n719 ) mutants, ppk-1 mRNA le v els were decr eased compar ed to wild type animals at the L2 stage (Figure 1 A) (Supplementary Table S1).Surprisingly, lin-28 mRNA le v els also decreased in ppk-1 RNAi compared to control RNAi animals (Figure 1 B), suggesting that they function on each other in a feed forward loop.
To determine whether ppk-1 is a heter ochr onic gene, we examined the well-characterized cell fusion marker ajm-1::gfp (Apical junction marker), alae formation (cell terminal dif ferentia tion marker) and adult cell fate marker col-19::gfp (adult specific marker) in temporally-staged ppk-1 (RNAi) and control RNAi animals (the ppk-1 null mutant exhibits lethality ( 34 )).We confirmed efficient RNAi of ppk-1 mRNA by qRT-PCR (Figure 1 B).In control RNAi animals, seam cell fusion -as detected by ajm-1::gfp -occurs at the middle (mid) L4 sta ge, b ut is not observed at early stages (Figure 1 C-E).Howe v er, 15% of ppk-1 (RNAi) animals showed the ajm-1::gfp fusion phenotype at the mid-L3 stage, 44.4% of them at late-L3 stage and 42.9% of them at early-L4 stage (Figure 1 C-E).Similarly, 36.4% of ppk-1 (RNAi) animals showed precocious col-19::gfp expression at the L4 sta ge, b ut control RNAi animals did not (Figur e 1 F).Furthermor e, we also found that ppk-1 (RNAi) animals showed precocious alae forma tion a t the early L4 stage (14.3%) and mid-L4 stage (45%) compared to the wild type w hich onl y showed this a t the la te L4 stage (Figure 1 G, H).These results indicate that depletion of ppk-1 causes a weak precocious heter ochr onic phenotype.Furthermore, as presented in Figure 1 I, ppk-1 (RNAi) animals showed a reduced  seam cell number from L2 to L4 stages compared to control RNAi animals.Since these phenotypes are characteristic of precocious seam cell patterning, these results suggest that ppk-1 is a new heter ochr onic gene.

PPK-1 regulates miRNA levels and functions in the lin-28 / let-7 heter ochr onic pathwa y
One feature of precocious heter ochr onic mutants such as lin-28 is altered expression of miRNAs involved in de v elopmental timing, such as let-7 ( 35 ).In order to determine whether PPK-1 regulates let-7 miRNA expression, we performed qRT-PCR experiments over developmental time using lin-42 as a r efer ence gene, as its expression peaks once during each larval stage ( 27 ).ppk-1 (RNAi) animals showed the peak time point of lin-42 expression 3 hours earlier compared to control RNAi animals at 20 • C (Supplementary Figure S2A-B).
At the peak time point of lin-42 mRNA le v els, we observed that ppk-1 (RNAi) animals showed increased mature let-7 le v els compared to control RNAi animals by northern blotting and qRT-PCR analysis (Figure 2 A-C).We also found that mature let-7 le v els increased in ppk-1 (RNAi) animals compared to control RNAi animals at early stages such as m-L2 and e-L3 stage by a qRT-PCR assay, which was consistent with the precocious phenotypes of ppk-1 (RNAi) animals (Supplementary Figure S3).
We examined whether over-expression (OE) of PPK-1 was sufficient to alter let-7 expression, we created PPK-1 (OE) animals by integrating a ppk-1 multi-copy array.When dri v en fr om its own pr omoter, the PPK-1 (OE) animals showed sterility and embryonic lethality phenotypes (not sho wn).Ho we v er, as ppk-1 was known to be expressed in seam cells from previous report ( 34 ), we created PPK-1 OE integrated animals driving ppk-1 from a seam cell specific promoter.PPK-1 (OE) animals showed a 4 hour developmental delay compared to wild type animals (Supplementary Figure S2C, D), and showed increased ppk-1 mRNA le v els (Figure 2 D).PPK-1 OE animals showed decr eased matur e let-7 le v els compared to wild type animals at the peak time point of lin-42 le v els, through northern blotting and qRT-PCR analysis (Figure 2 E-G).These results suggest that PPK-1 negati v ely regulates the expression of mature let-7.
To test if ppk-1 has a genetic interaction with let-7 in the heter ochr onic pathway, we knocked down ppk-1 in let-7(n2853) mutants.We found that the let-7(n2853) mutation suppresses the precious col-19::gfp phenotypes and decreased seam cell number of ppk-1 (RNAi) animals (Figure 1 F, J). let-7 controls seam cell divisions by negati v ely regulating its targets lin-41 and hbl-1 .We found that the mRNA le v els of lin-41 significantly decreased or increased under depletion or ov ere xpression of PPK-1 compared to control animals, respecti v ely, howe v er hbl-1 mRNA le v els were not significantly changed in both conditions (Figure 2 C, G). lin-29 , encodes a zinc-finger transcription factor that is r equir ed for hypodermal seam cell terminal differentiation, and functions downstream of let-7 ( 36 ).We found that the lin-29(n546) mutation suppressed the precious phenotypes and decreased seam cell number of ppk-1 (RNAi) animals, similar to the let-7 ( n2853 ) mutation (Figure 1 F,  J).These results indica te tha t PPK-1 functions upstream of let-7 / lin-41 and lin-29 .We also tested the le v els of other miRNAs, such as lin-4 , miR-75 , miR-77 and miR-237 , in ppk-1 (RN Ai) animals.Interestingl y, most of those miRN A le v els were also significantly increased in ppk-1( RNAi) animals (Figure 2 H).Taken together, these results demonstra te tha t PPK-1 regula tes miRNA le v els and acts in the lin-28 / let-7 heter ochr onic pathway.
Since most of the C. elegans factors shown abov e hav e orthologs in mammalian cells, to further explore the detailed mechanism of how PPK-1 regulates mature let-7 le v els, we utilized mammalian cells going forward.

PIP5K1A, the ortholog of C. elegans PPK-1 regulates miRN A e xpression
As PPK-1 is an ortholog of human PIP5K1A, we determined whether PIP5K1A regulates miRNA le v els similar to C. elegans PPK-1.Firstly, we generated PIP5K1A stable knockdown (KD) in RKO colon cancer cells, a wellestablished model in miRN A bio genesis r esear ch involving pre-miRNA / XPO5 ( 37 ), by a lenti virus-based inducib le shRNA system and confirmed knockdown of PIP5K1A by qRT-PCR (Figure 3 A).Next, we investigated whether PIP5K1A regulates let-7 miRNA expression.We found that knockdo wn of PIP5K1A sho wed incr eased matur e let-7a and unchanged prelet-7a-1 le v els compared to control cells by northern blotting and qRT-PCR assays (Figure 3 B-E).We also found that mature let-7b and let-7c le v els increased in PIP5K1A KD cells compared to control cells using a qRT-PCR assay (Figure 3 E).Furthermore, PIP5K1A KD cells showed decreased prilet-7a-1 and prilet-7b le v els compared to control cells (Figure 3 F).As prilet-7c expression le v els are v ery low in these cells, we did not investigate them.
We also tested the effect of ov ere xpression (OE) of PIP5K1A on let-7 le v els by northern blotting and qRT-PCR experiments.Firstly, we confirmed that PIP5K1A mRNA e xpression le v els incr eased in PIP5K1A OE cells compar ed to control cells (Figure 3 G).We found that PIP5K1A OE cells showed decreased mature let-7a and unaltered prelet-7a-1 le v els relati v e to wild type cells (Figure 3 H-K).Mature let-7b and let-7c le v els also decreased, as tested by qRT-PCR (Figure 3 K).We also found that PIP5K1A OE cells showed increased prilet-7a-1 and prilet-7b le v els (Figure 3 L).Gi v en that the change in le v els of pri-miRNA and matur e miRNA ar e not concordant, these results indica te tha t PIP5K1A regulates let-7 miRNA expression at least partially at the post-transcriptional le v el.
As PIP5K1A negati v ely regula tes ma ture let-7 le v els, we tested whether the knockdown of PIP5K1A effects expr ession of downstr eam targets of let-7 miRNAs, such as LIN28A , MYC , HMGA2 ( 38 ) .We found that those mRNA le v els decreased in PIP5K1A KD compared to with control cells (Figure 3 M).Furthermore, other miRNAs such as miR-122, miR-125a / miR-125b (orthologs of C. elegans lin-4 miRNA ( 39)) and miR-886 le v els increased in PIP5K1A KD cells, howe v er some not significantly (Figure 3 N).Taken together, these findings indicate that PIP5K1A negati v ely regulates mature miRNA le v els similar to its C. elegans ortholog, PPK-1.

PIP5K1A physically interacts with and co-localizes with XPO5
From our results above, PIP5K1A seemed to participate in regulating miRNA processing or biogenesis.Mammalian XPO5 belongs to the importin-beta family, a Ran-GTPdependent dsRNA-binding protein that regulates the nuclear export of pre-miRNAs ( 6 , 40 ) and has a rate-limiting role in miRN A bio genesis.This led us to test whether there is a connection between PIP5K1A and XPO5.We confirmed that siRNA knockdown of XPO5 indeed results in decr eased matur e let-7a , let-7b and let-7c le v els compared to control cells using a qRT-PCR assay as previously reported ( 37 ) (Supplementary Figure S4).
Firstly, we tested the simple hypothesis that PIP5K1A and XPO5 have a physical interaction.We immunoprecipitated PIP5K1A from HEK293 cell lysates with anti-PIP5K1A.We found that XPO5 co-precipitated with PIP5K1A (Figure 4 A).Howe v er, immunoprecipitation of XPO5 with an anti-XPO5 antibody, did not show PIP5K1A co-precipitating with XPO5 (data not shown).As this anti-XPO5 antibody only recognizes the C-terminus of XPO5 (see Materials and Methods), we created an N-terminus tagged HA-XPO5 and performed the same experiments.We found that PIP5K1A co-precipitated with XPO5 when immunoprecipitation of XPO5 was with an anti-HA antibody (Figure 4 B).We also used HA-KAP1(KRAB-associated protein) as a negati v e control to test the special interaction between PIP5K1A and XPO5.When we used HA antibody to immunoprecipitated KAP1, we did not detect PIP5K1A co-precipitating with KAP1 (Supplementary Figure S5).These results indicate that PIP5K1A physically interacts with XPO5 and seems to do so through its C-terminus.
To explore this interaction in vivo , the subcellular localization of the PIP5K1A-XPO5 complex was examined by immunofluorescence.We found that PIP5K1A co-localized with XPO5 mainly in the nucleus (Figure 4 C).The presence of the colocalization signal within the nucleus was further examined by generation of a 3D cross-section recon-  B, H ) Northern blot analyses of mature let-7a and prelet-7a-1 expression from the indicated cells.5.8S rRNA was used as a loading control.All experiments were performed with at least thr ee biological r eplica tes.(C , D, I, J) Quantita tion of let-7a / 5.8S ( C , I ) or prelet-7a-1 / 5.8S ( D, J ) le v els using Image J from the indicated cells.( E, K ) qRT-PCR analysis of mature let-7a, let-7b and let-7c le v els from indicated RKO cell lines.U6 snRNA was used as an endogenous control.( F, L ) qRT-PCR analysis of prilet-7a-1 and prilet-7b le v els from indicated RKO cell lines.pri-miRNA le v els were normalized to GAPDH mRNA.( M ) qRT-PCR analysis of LIN-28A , MYC and HMGA2 mRNA le v els in control and PIP5K1A KD RKO cell lines.GAPDH mRNA was used as an endogenous control.( N ) qRT-PCR analysis of mature miR-122 , miR-125a , miR-125b and miR-886 le v els in control and PIP5K1A KD RKO cells.U6 snRNA was used as an endogenous control.All experiments wer e performed with at least thr ee biological r eplica tes.All da ta ar e r epr esented as mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001 and NS: not significant.
structed from a confocal z-stack, demonstrating that colocalization (yellow) of PIP5K1A (green) and XPO5 (red) occurs in the nucleus in three dimensions (3D), as shown in the x,y, x,z and y,z planes (Figure 4 D).As the nuclear fraction of XPO5 binds with pre-miRNA in a Ran-GTP dependent manner and then exports the pre-miRNA to cytoplasm ( 6 , 7 ), this suggests that PIP5K1A may has a function with XPO5 in nucleus.

PIP5K1A r egulates the pr e-miRN A / XPO5 comple x during miRNA biogenesis
To address the question of how PIP5K1A functions in miRN A bio genesis, we tested whether PIP5K1A affects expression and / or function of XPO5.Because XPO5 is a known factor that has an important role in miRNA biogenesis and PIP5K1A associates with XPO5 from the above results, we tested the simple hypothesis that PIP5K1A regulates XPO5 le v els and that accounts for its contribution to miRN A bio genesis.Thus, we anal yzed the transcriptional and translational le v els of XPO5 in PIP5K1A KD cells.We found that PIP5K1A does not affect XPO5 mRNA and XPO5 protein le v els (Figure 4 E, F).This result indicates that PIP5K1A neither regulates XPO5 transcriptionally nor translationally.
As the C-terminal region of XPO5 is essential for the formation of the pre-miRNA / XPO5 / Ran-GTP complex ( 37 , 41 , 42 ) and it is also important to interact with PIP5K1A from our results, this led us to hypothesize that PIP5K1A may have a role in the process of the formation of the pre-miRNA / XPO5 comple x.For e xample, PIP5K1A binding may block pre-miRNA binding to the C-terminus of XPO5.If this is true, knockdown of PIP5K1A should increase the binding ability of pre-miRNA with XPO5.To test this hypothesis, we performed an RN A imm unoprecipitation (RIP) assay in RKO cells.We expressed HA-XPO5 in PIP5K1A KD and control cells and immunoprecipitated XPO5 with the anti-HA antibody.We checked for XPO5 protein le v els by western blot analysis (Figure 5 A).Interestingly, we found tha t XPO5-associa ted prelet-7a-1 le v els dramatically increased in cells with knockdown of PIP5K1A compared to control cells (Figure 5 B).As a negati v e control, U6 snRNA le v els were not changed (Figure 5 B).Moreover, we used 3 biotin labeled prelet-7a-1 and performed an RNA pull-down assay.We expressed HA-XPO5 in RKO cells and immunoprecipitated XPO5 with the anti-HA antibody.Eluted HA-XPO5 protein was incubated with RAN, GTP and 3 biotin labeled prelet-7a-1 with or without GST-PIP5K1A.The experiment showed tha t prelet-7a-1 -associa ted HA-XPO5 le v els were significantly reduced by addition of GST-PIP5K1A, compared to that without PIP5K1A (Figure 5 C, D).We also confirmed the RNA pulldown efficiency and did not see a difference under these conditions (Supplementary Figure S6).Furthermore, we also examined the cellular localization of XPO5 under depletion or ov ere xpression of PIP5K1A Figure 5. PIP5K1A regulates the binding ability of pre-miRNA / XPO5 complex .(A, B) RNA immunoprecipitation (RIP) assay.( A ) RIP assay was employed by using anti-HA antibody and normal IgG antibody in control and PIP5K1A KD RKO cells.HA::XPO5 was transfected in control and PIP5K1A KD RKO cells and then HA::XPO5 was immunoprecipitated with anti-HA antibody.Normal IgG was used as a negati v e control.XPO5 protein was analyzed with anti-HA antibody by immunoblotting.GAPDH was used as a loading control.( B ) Fold enrichment of pre-let-7a-1 detected by qRT-PCR.U6 snRNA was used as a negati v e control.n = 6 independent biological replicates.( C ) Western blot of XPO5 protein le v els in 3 biotin labeled prelet-7a-1 pull-down assay.Eluted HA-XPO5 protein was incubated with RAN, GTP and prelet-7a-1 in the absence or presence of the GST-PIP5K1A.HA-XPO5 and GST-PIP5K1A proteins were detected by western blotting with anti-HA and anti-GST antibodies, respecti v ely.( D ) Quantification of the enrichment of HA-XPO5.These results were normalized to the input of HA-XPO5 and calculated with ImageJ software.n = 3 independent biological replicates.Error bars r epr esented as mean ± SEM. **** P < 0.0001.( E, F ) The C / N (Cytoplasmic / Nuclear) ratio of prelet-7a-1 / prilet-7a-1 from the indicated cell lines by qRT-PCR analysis.All experiments were performed with at least three biological replicates.All data are represented as mean ± SEM. * P < 0.05, ** P < 0.01, **** P < 0.0001 and NS: not significant.by an immunofluorescence assay.We found a decreased C / N ratio of XPO5 localization in PIP5K1A OE compared to control cells (Supplementary Figure S7A-D) consistent with the idea that more PIP5K1A could bind with XPO5 to inhibit its function in the export of pre-miRNAs to cytoplasm.Howe v er, we did not observe a difference in the C / N ratio of XPO5 localization in PIP5K1A KD compared to control cells (Supplementary Figure S8A-C).Perhaps XPO5, after releasing its cargo in the cytoplasm, ra pidl y returns to the nucleus to mediate another round of transport.
These results indicate that PIP5K1A may function by blocking the binding of XPO5 to prelet-7a-1 .If this conclusion is correct, knockdown of PIP5K1A should increase prelet-7a-1 le v els in the cytoplasmic fraction, because more prelet-7a-1 can bind with XPO5 and then be exported from the nucleus to the cytoplasm.So next we performed a qPCR assay to test this possibility.We note here that we used prelet-7a-1 / prilet-7a-1 le v els to present the real prelet-7a-1 le v els as a qPCR assay could not distinguish between prelet-7a-1 and prilet-7a-1 .As expected, we found knockdown of PIP5K1A resulted in higher prelet-7a-1 / prilet-7a-1 le v els in cytoplasmic fraction compared to that in control cells (Figure 5 E).We also found PIP5K1A OE resulted in lower prelet-7a-1 / prilet-7a-1 le v els in the cytoplasmic fraction (Figure 5 F), consistent with the decreased C / N ratio of XPO5 localization by immunofluorescent assay (Supplementary Figure S7D).Howe v er, U6 snRNA le v els (a negati v e control) were not changed (Supplementary Figure S9).We confirmed the similar purity of cytoplasmic and nuclear fractions under each condition (Supplementary Figure S9).Taken together, these results suggest that PIP5K1A blocks the binding of prelet-7a-1 with XPO5 and then affects the C / N ratio of pre-miRNAs.
Ne xt, we inv estigated whether the kinase acti vity of PIP5K1A is r equir ed for its function in miRN A bio genesis.First, we created a kinase-dead PIP5K1A mutant protein ( PIP5K1A -D309N), which is already known to prevent the production of PIP2 ( 43 ).We confirmed this by immunofluorescence assay with an anti-PIP2 antibody (Supplementary Figure S10).We found that PIP5K1A -D309N OE cells showed decreased mature let-7a and unchanged prelet-7a-1 le v els relati v e to wild type cells by northern blotting and qRT-PCR assays (Figure 3 H-K), decreased mature let-7b and let-7c le v els relati v e to wild type cells by qRT-PCR (Figure 3 K) and lower prelet-7a-1 / prilet-7a-1 le v els in the cytoplasmic fraction (Figure 5 F) as also seen in PIP5K1A -WT OE cells.Howe v er, the prilet-7a-1 le v els were not changed (Supplementary Figure S11).These re- sults suggest that the kinase activity of PIP5K1A may be important for prilet-7 le v els at an early step of miRNA biogenesis, howe v er it is not r equir ed for regulating the export of prelet-7 from the nucleus and mature let-7 le v els at the later step of miRNA biogenesis.

DISCUSSION
Taken together, our study points to a previously unrecognized function of PIP5K1A in miRNA biogenesis.In summary, we have found that PIP5K1A regulates the ability of the pre-miRNA / XPO5 complex to participate in miRN A bio genesis / export.We propose a model based on our findings in RKO cell lines (Figure 6 ).In wild type cells, PIP5K1A interferes with the interaction between pre-miRNAs and XPO5 to maintain normal mature miRNA le v els.Our model explains how PIP5K1A's interference, which is reduced in PIP5K1A KD cells, allows more pre-miRNAs to bind to XPO5 and be exported to cytoplasm leading to more mature miRNAs to r epr ess downstr eam gene expression.
Our study has demonstrated phosphatidylinositol-4phospha te 5-kinase PPK-1 / PIP5K1A regula tes miRNA le v els from C. elegans to humans.We have found that ppk-1 mRN A directl y binds to LIN-28 RN A pr otein fr om our LIN-28 CLIP-seq analysis ( 30 ) and that ppk-1 mRNA levels decrease in a lin-28 ( n719 ) mutant compared to wild type animals by a qRT-PCR assay.Interestingly, we also found that in ppk-1 RNAi animals, lin-28 mRNA le v els decreased compared to control RNAi animals.These results indicate that LIN-28 and PPK-1 function on each other in a feed forward loop.How this occurs will r equir e further analysis.Consistent with lin-28 and ppk-1 acting positi v ely on each other, we found that ppk-1(lf) leads to similar heter ochr onic phenotypes as lin-28(lf).lin-28 mutant animals exhibit a decreased seam cell number as a result of skipping a symmetric cell division step in the L2 stage and exhibit pr ecocious expr ession of adult cell fates.Similar to this, we also found a decreased seam cell number in ppk-1 RNAi animals and a precocious adult cell fate phenotype.Loss of both genes also results in increased let-7 le v els.
To answer why ppk-1 RNAi animals showed decreased lin-28 mRNA le v els, we specula te tha t it is possibly because the increase of mature let-7 le v els allows for increased negati v e regulation of its target lin-28 mRNA.It has been reported that the human PIP5K1A mRNA is also bound by LIN28 ( 30 , 44 ) and PIP5K1A positi v ely regulates LIN28 mRNA le v els from our r esults (Figur e 3 M), suggesting that it has a conserved mechanism from C. elegans to humans.
We found that PIP5K1A associates with the importin ␤family factor XPO5 in human cells.While C. elegans PPK-1 is an ortholog of human PIP5K1A, unfortunately C. elegans lacks an XPO5 orthologue ( 45 ).Howe v er C. elegans has XPO-1, which is also an importin ␤-family and seems to function as the major nuclear export receptor ( 45 ).Our preliminary data indicate that PPK-1 associates with XPO-1 by a co-immunoprecipitation assay (unpublished data), howe v er this possible interaction needs to be fully investigated in the future.

Figure 1 .
Figure 1.ppk-1 is a new heter ochr onic gene.( A ) qRT-PCR analysis of ppk-1 mRNA le v els in the wild type and lin-28 ( n719 ) mutant animals at the L2 stage.pmp-3 mRNA was used as an endogenous control.( B ) qRT-PCR analysis of lin-28 mRNA and ppk-1 mRNA le v els fr om contr ol RNAi and ppk-1 (RNAi) animals at the peak time point of lin-42 mRNA le v els.act-1 mRNA was used as an endogenous control.( C-E ) ajm-1::gfp (Apical junction marker) expression in control RNAi and ppk-1 (RNAi) animals at the mid-L3 (m-L3), late-L3 (l-L3) and early-L4 (e-L4) stages, respecti v ely.The percentages of animals exhibiting any seam cell fusion and the number of animals examined are sho wn belo w the images.Arro ws indicate the fusion of seam cells.V: Vulva.n ≥ 20, Scale bar: 10 m. ( F ) col-19::gfp (Adult specific mar ker) e xpression in control RN Ai, ppk-1 (RN Ai), let-7(n2853) ; ppk-1 (RN Ai) and lin-29(n546) ; ppk-1 (RNAi) animals at the L4 stage.V: Vulva.n ≥ 20, Scale bar: 10 m.The percentages of animals with col-19::gfp expression and number of animals examined are shown to the right of the images.( G, H ) Precocious alae formation in control RNAi and ppk-1 (RNAi) animals at early-L4 (e-L4) and mild-L4 (m-L4) stages, respecti v ely.V: Vulva.n ≥ 20, Scale bar: 10 m.The percentages of animals with alae formation and number of animals examined are shown below the images.( I ) Seam cell number quantifications of control RNAi and ppk-1 (RNAi) animals from the L1 stage to the L4 stage.n ≥ 20 (actual number sho wn belo w the bar).( J ) Seam cell number quantifications from the indicated animals at the L4 stage.n ≥ 20 (actual number shown below the bar).All experiments were performed with at least three biological replicates.All data are represented as mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001 and NS: not significant.

Figure 2 .
Figure 2. PPK-1 regulates miRNA expression and functions in lin-28 / let-7 heter ochr onic pathway.( A, E ) Northern blot analyses of mature let-7 expression from the indicated animals (at the peak time point of lin-42 le v els).5.8S rRNA was used as a loading control.All experiments were performed with three biolo gical replicates.( B , F ) Quantitation of let-7 / 5.8S le v els from northern b lot anal yses using ImageJ.( C, G ) qRT-PCR anal ysis of mature let-7, lin-41 and hbl-1 le v els from the indicated animals (at the peak time point of lin-42 le v els).miRNA le v els were normalized to U18 snoRNA.lin-41 and hbl-1 mRNA le v els were normalized to act-1 mRNA.( D ) qRT-PCR analysis of ppk-1 mRNA le v els in the wild type and PPK-1 OE animals.act-1 mRNA was used as an endo genous control.(H ) qRT-PCR anal yses of mature lin-4 , miR-75, miR-77 and miR-237 le v els in control RNAi and ppk-1 (RNAi) animals.U18 snoRNA was used as an endogenous control.All experiments were performed with at least three biological replicates.All data are represented as mean ± SEM. * P < 0.05, ** P < 0.01, **** P < 0.0001 and NS: not significant.

Figure
FigurePIP5K1Aregulates miRNA expression .( A, G ) qRT-PCR analysis of PIP5K1A mRNA le v els from indicated RKO cell lines.GAPDH mRNA was used as an endogenous control.( B, H ) Northern blot analyses of mature let-7a and prelet-7a-1 expression from the indicated cells.5.8S rRNA was used as a loading control.All experiments were performed with at least thr ee biological r eplica tes.(C , D, I, J) Quantita tion of let-7a / 5.8S ( C , I ) or prelet-7a-1 / 5.8S ( D, J ) le v els using Image J from the indicated cells.( E, K ) qRT-PCR analysis of mature let-7a, let-7b and let-7c le v els from indicated RKO cell lines.U6 snRNA was used as an endogenous control.( F, L ) qRT-PCR analysis of prilet-7a-1 and prilet-7b le v els from indicated RKO cell lines.pri-miRNA le v els were normalized to GAPDH mRNA.( M ) qRT-PCR analysis of LIN-28A , MYC and HMGA2 mRNA le v els in control and PIP5K1A KD RKO cell lines.GAPDH mRNA was used as an endogenous control.( N ) qRT-PCR analysis of mature miR-122 , miR-125a , miR-125b and miR-886 le v els in control and PIP5K1A KD RKO cells.U6 snRNA was used as an endogenous control.All experiments wer e performed with at least thr ee biological r eplica tes.All da ta ar e r epr esented as mean ± SEM. * P < 0.05, ** P < 0.01, *** P < 0.001 and NS: not significant.

Figur e 4 .
Figur e 4. PIP5K1A physicall y interacts with and localizes with XPO5.(A, B) Physical interaction between PIP5K1A and XPO5.( A ) Endogenous PIP5K1A of HEK293 cells was immunoprecipitated and then the XPO5 protein was analyzed by immunoblotting.Normal IgG was used as a negati v e control.( B ) HEK293 cells were transfected with HA::XPO5 and XPO5 was immunoprecipitated with anti-HA antibody.Normal IgG was used as a negati v e. ( C ) Representati v e image of confocal imaging of co-localization between PIP5K1A and XPO5 in RKO cells.DAPI was used to stain the nuclear DNA.n = 30 cells, Scale bar: 7 m.( D ) Cross-sections of a 3D volume reconstruction.The presence of the colocalization signal was examined by generation of a 3D cross-section reconstructed from a confocal z-stack.The colocalization (yello w, arro w) of PIP5K1A (green) and XPO5 (red) in the nucleus in three dimensions (3D), as shown in the x,y, x,z, and y,z planes.( E ) qRT-PCR analysis of XPO5 mRNA le v els in control and PIP5K1A KD RKO GAPDH mRNA was used as an endogenous control.The data are represented as mean ± SEM.Individual experiments were performed in triplicate.NS: Not significant.( F ) Western blot of XPO5 protein levels in control and PIP5K1A KD RKO cells.GAPDH mRNA was used as a loading control.

Figure 6 .
Figure 6.A schematic model of how PIP5K1A / XPO5 regulates pre-miRNA export from the nucleus to the cytoplasm.