Cilia assembly and disassembly are coupled to actin dynamics, ensuring a coherent cellular response during environmental change. How these processes are integrated remains undefined. The histone lysine demethylase KDM3A plays important roles in organismal homeostasis. Loss-of-function mouse models of Kdm3a phenocopy features associated with human ciliopathies, whereas human somatic mutations correlate with poor cancer prognosis. We demonstrate that absence of KDM3A facilitates ciliogenesis, but these resulting cilia have an abnormally wide range of axonemal lengths, delaying disassembly and accumulating intraflagellar transport (IFT) proteins. KDM3A plays a dual role by regulating actin gene expression and binding to the actin cytoskeleton, creating a responsive “actin gate” that involves ARP2/3 activity and IFT. Promoting actin filament formation rescues KDM3A mutant ciliary defects. Conversely, the simultaneous depolymerization of actin networks and IFT overexpression mimics the abnormal ciliary traits of KDM3A mutants. KDM3A is thus a negative regulator of ciliogenesis required for the controlled recruitment of IFT proteins into cilia through the modulation of actin dynamics.

Introduction

Primary cilia are homeostatic sensory organelles driven by intraflagellar transport (IFT) that integrate multiple and diverse cellular inputs critical to normal development and organ homeostasis. Diverse environmental cues affect their assembly and disassembly in a manner that is tightly coupled to the actin cytoskeleton and changes in gene expression. How cells coordinate these processes to effect a coherent cellular response remains unclear.

Actin network architecture is increasingly recognized as a major driver of ciliogenesis in cycling cells (Pitaval et al., 2010). Chemical and genetic perturbations of multiple actin remodeling factors have been found to promote ciliogenesis or affect ciliary length (Kim et al., 2010, 2015; Cao et al., 2012; Hernandez-Hernandez et al., 2013; Kang et al., 2015) and intriguingly rescue ciliogenesis defects in Ift88 (Kim et al., 2010) and IFT121 (Fu et al., 2016) mutant cells. Actin-myosin–mediated contraction participates in ciliary abscission as the neuronal differentiation program is triggered (Das and Storey, 2014). Leptin elongates hypothalamic cilia via transcriptional regulation and actin destabilization (Kang et al., 2015), whereas the microRNA miR-129-3p regulates cilia assembly through the concomitant transcriptional silencing of centrosomal proteins and repression of actin filament formation (Cao et al., 2012). Altogether, these findings highlight the existence of concurrent regulatory mechanisms coordinating ciliogenesis, actin dynamics, and transcription.

Interestingly, mouse mutants for the histone N-lysine demethylase Kdm3a (Jmjd1a, Tsga, or Jhdm2a) are obese, predisposed to diabetes (Inagaki et al., 2009; Tateishi et al., 2009), and have male infertility phenotypes (Okada et al., 2007) overlapping with mouse models of dysfunctional cilia. Structural abnormalities of the acrosome and manchette were observed in mutant Kdm3a mouse spermatids (Kasioulis et al., 2014) similar to those reported for conditional inactivation of Ift88 (Kierszenbaum, 2002; San Agustin et al., 2015). KDM3A localizes to the actin-rich acrosome–acroplaxome region, which is altered in Kdm3a mutant spermatids (Kasioulis et al., 2014). In spite of these phenotypic and functional overlaps, a link between KDM3A and ciliogenesis has not been investigated.

KDM3A is part of the large family of Jumonji C (JmjC) domain–containing proteins (Clissold and Ponting, 2001; Klose et al., 2006; Markolovic et al., 2016). Deregulated expression of KDM3A is associated with colorectal cancer (Uemura et al., 2010), breast cancer (Wade et al., 2015), and hepatocellular carcinoma (Yamada et al., 2012). Hypoxia (Beyer et al., 2008; Pollard et al., 2008; Wellmann et al., 2008) and the HSP90 chaperone machinery (Kasioulis et al., 2014) are reported to modulate KDM3A levels and activity, whereas its interaction with chromatin is modulated by cold-regulated phosphatases (Abe et al., 2015). Overall, these findings indicate that KDM3A plays fundamental roles in organismal homeostasis, integrating multiple environmental cues with physiological states.

KDM3A presents different regulatory modalities and subcellular distributions. Through the catalytic removal of mono- and N-dimethyl groups from lysine 9 of histone H3, KDM3A acts as a transcriptional activator (Yamane et al., 2006) of multiple genes involved in spermatogenesis (Okada et al., 2007), fat storage (Okada et al., 2010), and energy expenditure (Inagaki et al., 2009) as well as of diverse types of cancers (Ohguchi et al., 2016; Zhao et al., 2016). More recently, scaffolding and noncatalytic modes of transcriptional regulation by KDM3A have emerged wherein KDM3A binds to the SWI–SNF (Switch-sucrose nonfermentable) chromatin-remodeling complex (Abe et al., 2015) or Hedgehog-responsive transcription factor GLI1 (Schneider et al., 2015), regulating expression of target genes. Although these functions relate to the role of KDM3A in the nucleus, KDM3A exits the nucleus in response to mechanosensitive stimuli (Kaukonen et al., 2016) and is found at various cytoplasmic sites of somatic cells and during germ cell development (Okada et al., 2007; Yang et al., 2009; Yamada et al., 2012; Kasioulis et al., 2014). Altogether, these findings indicate that KDM3A is a multifunctional protein with highly regulated subcellular distributions and nontranscriptional roles that remain to be explored.

Here, we present evidence on the regulation of actin dynamics by KDM3A and the subsequent impact on IFT and ciliogenesis. KDM3A is required for the stabilization of primary cilia and regulates cilial responsiveness to environmental cues. RNA-sequencing, proteomic, and two-hybrid studies reveal a dual role for KDM3A in this integrated response modulating ciliogenesis. On a global level, it senses extracellular signaling and transcriptionally regulates the free pool of actin available. On a very local level, KDM3A directly binds to the actin cytoskeleton, creating a responsive “actin gate” and regulating IFT within cilia and periciliary compartments. Importantly, deregulated expression of an IFT protein, IFT81, coupled with actin depolymerization phenocopies cilial defects of KDM3A mutants, demonstrating that KDM3A-mediated cytoskeletal changes tightly regulate IFT during ciliary growth.

Results

KDM3A plays a conserved role in the regulation of ciliary dynamics

During the analysis of a deletion mutant mouse model of Kdm3a lacking the catalytic domain JmjC (ΔJC; Kasioulis et al., 2014), we observed that flagella of mutant spermatids were consistently shorter than those of control littermates (Fig. 1, A–C). We then asked whether KDM3A played a broader role in ciliogenesis, examining primary cilia in somatic Kdm3aΔJC cells. Mouse embryonic fibroblasts (MEFs) derived from Kdm3aΔJC mutants show abnormal ranges of ciliary lengths (Fig. 1 D), which can be partially rescued by transfecting a short cytoplasmic isoform of KDM3A (Fig. 1 E, KDM3Ai2-EGFP; Kasioulis et al., 2014). Kdm3aΔJC/ΔJC MEFs also displayed accumulations of IFT88 at the ciliary base and tip compared with the uniform distribution of IFT88 in wild-type cilia (Fig. 1, F and G). These observations suggested that KDM3A could play a role in axonemal dynamics and IFT in primary cilia. Because Kdm3aΔJC mutant cells encode a truncated KDM3A protein, albeit at very low levels (Kasioulis et al., 2014), we generated KDM3Anull cells in the immortalized human retinal epithelium hTERT RPE1 cell line. RPE1 cells have been thoroughly characterized to study ciliary growth, resorption, and axoneme length in response to cell density and serum content of the culture media (Pugacheva et al., 2007; Kim et al., 2015).

Using CRISPR/Cas9 gene editing directed to exon 22 of KDM3A, we generated independent RPE1 cell lines null for KDM3A. Lines 1, 2, and 8 (KDM3Anull) did not express KDM3A protein (Fig. 2 A) and were found to bear compound KDM3A mutations by genotyping (Fig. S1, A and B). As controls, we used the parental RPE1 line (RPE1WT) as well as an RPE1 sister line that has undergone CRISPR/Cas9 editing but was wild type for KDM3A (RPE1CrWT). Cells were plated at equal densities and promoted either to (a) ciliate, by serum withdrawal (−FCS), or (b) resorb cilia, by serum replenishment (+FCS), with or without temperature stress. No differences were observed in the cilial length between controls and KDM3A null cultures after 3 or 24 h without serum (Fig. 2 B, 3 h and 24 h −FCS). In contrast, under resorption conditions, KDM3Anull cilia remained significantly longer than those of controls (Fig. 2 B, resorption +FCS), suggesting that KDM3Anull cells failed to efficiently resorb cilia. Moreover, the proportion of KDM3Anull cells that remained ciliated was unchanged after serum replenishment, in contrast to the efficient resorbtion of ciliated control cells (Fig. 2 C).

It was noted during these studies that KDM3Anull cilia displayed a wider range of lengths than RPE1WT under all conditions (Fig. 2 B, analysis of variance F). Also 3 h after serum withdrawal, KDM3Anull cultures had a higher proportion of ciliated cells than controls (Fig. 2 D). Together, these results suggest that mammalian KDM3A negatively regulates ciliogenesis.

Ciliogenesis is usually promoted through serum withdrawal, but cycling RPE1 cells also form cilia in the presence of serum, albeit at lower frequencies (Cao et al., 2012). To confirm the propensity of KDM3Anull cells to ciliate, we examined ciliogenesis in exponentially growing cultures. In the presence of serum, some extremely long cilia with abnormal bulges were observed in KDM3Anull cells (Fig. 2 E, arrow). Importantly, higher proportions of ciliated cells were consistently observed in mutants (Fig. 2 F), in spite of comparable cell density and mitotic index to controls (RPE1WT+CrWT 4.05%, SD = 1.90; KDM3Anull 6%, SD = 1.63; N = 3 biological replicates, n > 300 cells scored/condition). A higher proportion of KDM3Anull cells retained cilia into late G2 (Fig. 2, G and H), whereas most wild-type cilia were already resorbed (Spalluto et al., 2013). However, these changes were neither caused by nor associated with alterations in cell cycle progression (Fig. 2 I). As we did not observe mutant mitotic cells with cilia, we conclude that mutant cilia must be eventually resorbed albeit with altered kinetics.

Both Kdm3aΔJC deletion and KDM3Anull mutant models indicate that KDM3A plays a role in axoneme dynamics, yet the deletion mutants have on average shorter cilia (Fig. 1 D) rather than the wider range of lengths observed in KDM3Anull cells (Fig. 2 B). These contrasting outcomes may be caused by a gain of function of the truncated KDM3A protein encoded by Kdm3aΔJC mutants, as MEFs derived from a second hypomorphic mouse model of Kdm3a (Kdm3aGT; Kasioulis et al., 2014) present, as observed in KDM3Anull RPE1 cells, a wider range of cilial lengths (Fig. S1, C and D). Ciliary resorption in response to serum replenishment and temperature stress was similarly delayed in Kdm3aGT/GT MEFs (Fig. S1 E).

In summary, given the increased proportion of ciliated cells, abnormal range of cilial lengths, and delayed cilial resorption kinetics, our loss-of-function mutants suggest that in the absence of KDM3A, ciliogenesis is facilitated and may be less tightly regulated. Conversely, overexpressing KDM3A reduces cilia formation in the parental RPE1WT line (Fig. S1, F and G).

KDM3A regulates actin expression and binds actin in vivo and in vitro

Given KDM3A’s localization to actin-rich structures and altered biochemical properties of actin in Kdm3a mutant spermatids (Kasioulis et al., 2014), we hypothesized that the effect of KDM3A depletion on ciliogenesis is a consequence of actin dysregulation. As KDM3A had been shown to bind the promoter and regulate the expression of smooth muscle α-actin (Acta2; Lockman et al., 2007), these effects could be caused by altered transcription or protein–protein interactions or both. Indeed, phalloidin staining of KDM3Anull cells (Fig. 3 A) showed consistently decreased abundance of actin filaments compared with RPE1WT (Fig. 3 B), and this was also evident in Kdm3aΔJC/ΔJC MEFs (Fig. S2 A).

The turnover of the actin cytoskeleton is both transcriptionally and posttranscriptionally regulated (Olson and Nordheim, 2010), requiring a pool of actin monomers (G-actin) and multiple actin-remodeling factors that determine F-actin formation (Blanchoin et al., 2014). We therefore investigated both transcriptional changes and protein–protein interactions that could contribute to KDM3A ciliary traits. To address the global transcriptional changes that occur in the absence of KDM3A during ciliary growth and resorption, we used a combined approach of RNA sequencing and quantitative label-free mass spectrometric analysis (Tables S2 [raw datasets] and S3 [thresholds]). Principal-component analysis of RPE1WT and KDM3Anull proteomes first showed that differences between samples primarily reflected serum content of the culture media as opposed to genotypes of the cell lines (Fig. 3 C). This supports previous studies reporting that KDM3A does not cause large global transcriptional changes but is instead a fine regulator of gene expression (Okada et al., 2007). Subsequently, considering that KDM3A is a transcriptional activator (Yamane et al., 2006), we identified genes for which protein or RNA abundance were reduced in KDM3Anull cells compared with RPE1WT controls (Table S3). GO(Gene Ontology) analysis suggests that the transcriptional activity of KDM3A is required to sense the extracellular environment (Fig. 3 D and Table S4). In agreement with previous findings (Lockman et al., 2007), direct interrogation of actin expression in these datasets reveals that smooth α-actin ACTA2 expression (Table S3) as well as protein levels (Fig. 3 E) failed to be up-regulated in response to serum replenishment. A reduced abundance of ACTA2 in mutants could potentially compromise actin polymerization by limiting available actin monomers during ciliary resorption (+FCS). However, as actin levels are normal in KDM3Anull cells in the absence of serum, transcriptional control alone cannot explain perturbations of actin dynamics observed in mutant cells during ciliary growth (−FCS).

Next, we investigated a potentially direct effect of KDM3A on the cytoskeleton through protein interactions using KDM3A immunopurification of total cell extracts followed by mass spectrometry. In parallel, we performed a subsaturation two-hybrid screen using the JMJC domain of KDM3A as bait. The KDM3A antibody significantly enriched 166 proteins from RPE1WT (Fig. 3 F and Table S3), where GO analysis revealed actin binding and actin cytoskeleton as the most enriched terms (Table S4). Strikingly, these immunopurifications showed that endogenous KDM3A copurified with ACTA2 and the actin-capping protein CAPZB, both of which were also found in the two-hybrid screen. A binary interaction between ACTA2 and the JMJC domain of KDM3A was observed by yeast two-hybrid screening (Fig. 3 H), but not with CAPZB (Fig. S2 B). Further confirmation of a direct association between KDM3A and ACTA2 was supported through binding of purified GST-JMJC to pure smooth muscle α-actin (Figs. 3 I and S3 C). The multiple actin-related proteins that copurified with endogenous KDM3A may thus be indirectly enriched through KDM3A binding to actin filaments.

In summary, global analysis of the transcriptional and total proteomic profiles of cells lacking KDM3A show enrichment of GO terms that relate to the extracellular environment that are in line with the increasingly recognized role of KDM3A in environmental adaptation and mechanosensation (Chakraborty et al., 2016; Kaukonen et al., 2016; Pedanou et al., 2016). Immunopurification of endogenous KDM3A, two-hybrid screening, and far–Western blotting revealed that KDM3A interacts with actin in vivo and in vitro. Together with the transcriptional failure of KDM3Anull cells to induce actin expression in response to serum replenishment, these results reveal a novel role for KDM3A in the modulation of actin dynamics that may underpin phenotypic changes of KDM3Anull cells.

KDM3A modulates actin dynamics

To characterize the functional interaction between KDM3A and actin, we further investigated cellular phenotypes known to depend on actin dynamics.

It was recently shown that KDM3A expression is up-regulated when cells are in suspension (Pedanou et al., 2016) and that its cytoplasmic localization depends on mechanosensitive signals (Kaukonen et al., 2016). Indeed, Halo-tagged KDM3A is cytoplasmic and enriched in cortical actin filaments and cellular edges immediately after attachment of suspended cells (Fig. 4 A). We reasoned that the fast wave of actin polymerization (Fig. 4 B) that follows early spreading over solid substrates (Wolfenson et al., 2014) would thus provide functional evidence on the impact of cytoplasmic KDM3A on actin dynamics. Wild-type (RPE1WT and RPE1CrWT) and KDM3Anull (lines 1, 2, and 8) cultures were trypsinized and maintained in suspension in serum-free media for 2 h before plating (Fig. 4 C) or plated directly (Fig. 4 D) onto plastic plates. We found that in both cases, a greater proportion of KDM3Anull cells remained unspread compared with control cultures (Fig. 4, C and D; P0/P1 ratio χ2 = 0.001), indicating that loss of KDM3A interferes with the early stages of cell spreading that require fast actin polymerization.

In the presence of serum, KDM3Anull cells consistently have rounder cellular edges compared with RPE1WT (Fig. 4 D). Branching or elongation of actin fibers is dynamically driven by the antagonism between ARP2/3 nucleating activity and capping proteins, ultimately determining the length of the actin fiber (Edwards et al., 2014). At the cellular edge, this interplay is manifested as the extent of lamellipodia to filopodia formation (Fig. 4 E), where the underlying actin-remodeling events steer cell migration (Pollard and Borisy, 2003; Koestler et al., 2013). Consistent with these findings, KDM3Anull cells presented reduced cell motility (Fig. 4 F) and prominent leading edges (Fig. S2 D). Conversely, overexpressing KDM3A increases the proportion of cellular membrane ending in spike/filopodia-like protrusions (Fig. 4, H–J; and Fig. S2 E).

Together, the morphological and behavioral features found in the absence or overexpression of KDM3A demonstrate that KDM3A modulates global actin dynamics and suggest that KDM3A binding to actin may compete with or facilitate the binding of different actin-remodeling proteins.

Altered actin dynamics underlies the ciliary traits of KDM3A-null cells

To confirm that a deregulated actin cytoskeleton in the absence of KDM3A underlies the abnormal ciliary phenotypes observed in mutants, we aimed to rescue these phenotypes using chemical modulators of F-actin formation. Depolymerization of F-actin promotes ciliary growth (Kim et al., 2010, 2015). Conversely, actin polymerization is thought to be required for ciliary resorption (Liang et al., 2016).

Interfering with actin dynamics through cytochalasin D (CytoD) treatment promotes ciliogenesis in cycling RPE1 cells without the requirement of serum withdrawal (Cao et al., 2012). Indeed, CytoD treatment alone is sufficient to rescue ciliogenesis in genetic mutants affecting IFT machinery (Kim et al., 2010; Fu et al., 2016). Surprisingly, we instead observed that KDM3Anull cells accumulated abnormal phalloidin-stained foci (Fig. 5 A, 2 µM CytoD) and have decreased proportion of ciliated cells (Fig. 5 B). Live-cell imaging with a ciliary marker (5-HT6-GFP; Berbari et al., 2008) suggests that the decreased proportion of KDM3Anull ciliated cells in 0.5 µM CytoD is likely caused by reduced ciliary stability, as cilia break off the cell body (Fig. 5 C), without the net increase in ciliary length observed in control cultures (Fig. 5 D). CytoD treatment of cycling RPE1WT cultures thus phenocopies the increased proportion of ciliated cells found in KDM3Anull cells in the presence of serum (Fig. 2 E). The low abundance of actin fibers of KDM3A mutants likely facilitates ciliogenesis, but further perturbations in actin polymerization are not tolerated by KDM3Anull cells (Fig. 5 E).

Critically, the converse experiment, where we promoted actin polymerization using jasplakinolide treatment (Kustermans et al., 2008) in the presence of serum, rescued the abnormal resorption of KDM3Anull cilia (Fig. 5, F and G). This suggests that failure of KDM3Anull cells to polymerize actin in response to serum contributes to the abnormal resorption kinetics of mutant cilia (Fig. 5 H).

These pharmacogenetic results support a functionally relevant association between KDM3A and the actin cytoskeleton during ciliogenesis. In the absence of KDM3A, the formation of actin filaments is perturbed and underlies the ciliary defects of KDM3Anull cells.

A requirement for KDM3A in cilial stability and IFT

Live imaging suggested that KDM3Anull cilia were unstable. We investigated whether this instability was caused by the drug treatment or whether it is an inherent property of mutant cilia. Live imaging of cell during ciliary growth (−FCS) revealed that RPE1WT cilia remained fairly stable, whereas KDM3Anull cilia underwent excessive elongation and fragmentation (Videos 1 and 2 and Fig. 6 A). Similar ciliary instability was observed in Kdm3aΔJC mutant MEFs (Video 3). These results suggest that instability of mutant axonemes in the absence of KDM3A reflect an inability to regulate axonemal length. These findings are supported by the unresponsiveness of KDM3Anull cilia to resorb (Video 4) and the wider range of lengths observed in fixed KDM3A mutant cilia (Fig. 2).

The presence of ciliary bulges (Fig. 2 E) and their shedding observed during live imaging of KDM3Anull cells suggest that the instability of KDM3A mutant cilia results from perturbation of IFT. Indeed, IFT staining of MEFs (Fig. 1 F) and RPE1 null for KDM3A (Fig. 6 B) showed abnormal distributions of IFT88 and IFT81. The anomalies were most evident at the ciliary tip (Fig. 6 C) but could also be observed at the ciliary base (Fig. 1 F).

In Chlamydomonas reinhardtii, actin is required for the recruitment of IFT particles to the basal body, where it controls flagellar entry of IFT trains (Avasthi et al., 2014). Accordingly, the abnormal distribution of IFTs observed in KDM3Anull cells could be caused by the permissive and depolymerized state of the actin cytoskeleton in these mutants. If this was true, then perturbing actin dynamics in RPE1WT would phenocopy the IFT anomalies observed in KDM3A mutants. To test this, we treated cells with CytoD or CK-869, a small molecule that specifically inhibits ARP2/3 complex components (Hetrick et al., 2013) copurifying with KDM3A (Fig. 3 F). Perturbations of ARP2/3 activity could contribute to themorphological phenotype of KDM3Anull cells (Fig. 4, E–G). RPE1WT cells treated with either drug in the presence or absence of serum showed large accumulations of IFT81 at the ciliary tip of RPE1WT cells (Fig. 6, C–E) accompanied by ciliary elongation (Fig. 6 F). The effect of CK-869 was particularly interesting, as KDM3Anull cells were extremely responsive to ARP2/3 inhibition, showing ciliary elongation from doses as low as 2 µM (Fig. 6 G), whereas RPE1WT cells only respond from 50 µM (Fig. 6 H). This functional synergism between KDM3A mutants and CK-869 treatment indicates a common underlying mechanism by which KDM3A influences IFT through actin polymerization involving the ARP2/3–actin assembly pathway.

In summary, live imaging studies provide evidence that in the absence of KDM3A, cilia become unstable. Antibody staining showed abnormal distributions of endogenous IFTs in mutant cilia, and this trait can be phenocopied in wild-type cells by perturbing actin dynamics. Using two different drugs interfering with actin nucleation, we demonstrate that in mammalian primary cilia, actin dynamics also modulates IFT. Altogether, these findings provide evidence that in the absence of KDM3A, actin dynamics are perturbed, disrupting IFT and destabilizing cilia.

KDM3Anull cilia reveal an actin-driven constraint for IFT in wild-type cilia

The finding that inhibiting actin polymerization in mammalian control cells phenocopies the abnormal IFT loaded bulges found in KDM3Anull cells suggests that wild-type cilia have an actin-dependent constraint that modulates ciliary IFT. To test this further, we increased IFT protein levels by transfecting tagged IFT81 into RPE1WT and KDM3Anull cells, reasoning that if an actin gate does exist and is dysfunctional in the absence of KDM3A, IFT overexpression would have opposing effects in wild-type and KDM3A mutants.

We generated a functional IFT81-DDK construct capable of rescuing ciliogenesis in an IFT81-null RPE1 line (Fig. S3, A–D). Comparable transfection levels of IFT81-DDK onto RPE1WT and KDM3Anull cells were confirmed by immunoblotting (Fig. 7 A). Live imaging revealed that overexpressing IFT81 enabled formation of more cilia and with faster kinetics in KDM3Anull mutants compared with RPE1WT cells (Fig. 7 B, slopes; and Fig. S3, E and F). Importantly, expression of 5HT6-GFP alone did not recapitulate this increased rate of ciliogenesis in KDM3Anull cells (Fig. 7 C), demonstrating that this difference is not simply caused by an increased propensity to ciliate but is dependent on ectopic IFT81. In the presence of IFT81-DDK, KDM3Anull cilia are also longer than controls (Fig. 7 D). These results did not reflect differences in cell density (Fig. 7 E) or levels of endogenous IFT-B components (Fig. 7 F). This indicates that increasing the abundance of IFT promotes ciliogenesis in KDM3A mutants, but not wild-type controls.

Because CytoD treatment of control cells phenocopied the ciliary traits of KDM3Anull mutants, we simultaneously induced actin depolymerization along with IFT81 overexpression in RPE1WT cultures. Indeed, this dual treatment allows wild-type cells to become sensitive to IFT-B overexpression when actin is depolymerized (Fig. 7 G).

Together, these findings provide further and independent evidence of deregulated IFT in KDM3A mutant cilia that result from perturbations in actin dynamics. When actin is unperturbed, increasing IFT pools in control cells makes no difference to ciliary growth, indicating that actin allows wild-type cells to maintain tight regulation of IFT entry into the ciliary compartment (Fig. 8 A). In contrast, when actin nucleation is perturbed either genetically (KDM3A mutants) or chemically (CytoD treatment ARP2/3 inhibition), this constraint is lifted.

Discussion

Our results show that KDM3A plays a role in mammalian ciliogenesis through the modulation of actin dynamics. In the absence of KDM3A, the assembly and growth of primary cilia is enhanced. However, cilial stability is compromised such that the length of mutant axonemes and presence of IFT proteins within cilia become unregulated. We find that KDM3Anull cells have reduced abundance of actin filaments. This is because KDM3A plays a transcriptional role regulating actin gene expression and binds directly to actin protein, influencing local actin networks. We demonstrate that the ciliary traits uncovered in KDM3A mutants can be chemically reversed or phenocopied in wild-type cultures by altering actin dynamics. Through independent approaches involving genetics, small-molecule inhibitors, and IFT overexpression, we uncover an actin-mediated constraint on IFT within mammalian cilia. Our findings uncover a functional interaction between the actin cytoskeleton and IFT that is perturbed in the absence of KDM3A.

Cells adapt to new environments, like changes in the serum content of the media, through simultaneous changes in gene expression, cytoskeletal rearrangements, and cilia signaling. This integrated response requires signal transduction relays that communicate the cytoplasmic actin polymerization status with IFT in cilia and histone code in the nucleus. Through the use of KDM3A mutants, we have uncovered a requirement for KDM3A, which acts as a nexus in this intracellular signaling network. RNA sequencing and proteomic profiling reveal a transcriptional role for KDM3A in increasing the pool of actin in response to serum replenishment that is concomitant with actin filament formation and ciliary resorption. Failure of KDM3A mutants to up-regulate actin expression will thus likely contribute to the poor resorption of mutant cilia. During disassembly, formation of new actin filaments at the ciliary base would likely require a pool of free actin to restrict access into the ciliary compartment. Nevertheless, in the absence of serum KDM3Anull mutants contain comparable levels of actin transcripts and protein to those found in control cells, yet it is under these growth conditions that ciliary extension is facilitated in KDM3A mutants. Here, we propose that direct protein interaction of KDM3A with actin locally creates an actin gate at the ciliary base (Fig. 8, B and C; and Video 5) to regulate entry of IFT. In the absence of KDM3A, access to cilia is unrestricted and balance and/or kinetics of IFT are perturbed.

The fact that KDM3A is an enzyme that binds actin through its catalytic domain, together with previous observations that actin proteins could be abnormally methylated in Kdm3a mutant testes (Kasioulis et al., 2014), prompted us to test the demethylase activity of KDM3A in lysine-methylated actin peptides found in this study. However, as yet we have failed to detect changes in methylation by KDM3A in any tested peptide in vitro (unpublished data). Although it remains to be seen whether KDM3A can posttranslationally modify actin in vivo, the binding of KDM3A to actin itself is likely to modulate binding of other actin modifiers, including ARP2/3 and the capping components, to directly tune actin networks.

The functional interplay between KDM3A-ARP2/3 is interesting; KDM3Anull cells respond to lower CK-869 doses than RPEWT cells, suggesting that the ciliary traits of KDM3A mutants have already compromised ARP2/3 function. A role for ARP2/3 at the cilium/centrosome has been suggested. ARP2/3 mediates actin nucleation at centrosomes (Farina et al., 2016). Silencing of ACTR3 or ARP2 in cultured cells promotes ciliogenesis but with unclear kinetics, as cells were phenotyped 2.5–3 d after siRNA depletion was initiated (Kim et al., 2010; Cao et al., 2012). In our study, we used a small-molecule inhibitor, CK-869, which prevents the formation of new actin filaments and unlike CytoD does not destabilize filaments that are already formed (Hetrick et al., 2013). We demonstrate that acutely blocking ARP2/3 activity (3-h treatment) alters the distribution of endogenous IFT-B within cilia. This suggests that continual actin nucleation by ARP2/3 is required for regulated entry of IFT into cilia and cilial length maintenance.

Our identified roles for KDM3A in actin dynamics have further implications. Understanding how cancer cells can subvert actin bundling during metastasis is an emerging and exciting theme in research. Although the brunt of the current focus of overexpression of KDM3A in various types of cancer has been on its transcriptional targets (Krieg et al., 2010; Uemura et al., 2010; Cho et al., 2012; Osawa et al., 2013; Tee et al., 2014; Parrish et al., 2015), our work suggests that cytoplasmic KDM3A may act as an important cytoskeletal rheostat integrating mechanosensation (Kaukonen et al., 2016; Pedanou et al., 2016) with cell adhesion and signaling through cilia. Aberrant length ranges and resorption of KDM3A mutant cilia in response to environmental cues could also contribute to metabolic imbalance in Kdm3a mouse models (Okada et al., 2010), as ciliary length is tightly regulated in response to leptin levels (Han et al., 2014).

Our findings that KDM3A has roles beyond the “direct” regulation of transcription (i.e., acting as an H3K9 demethylase; Yamane et al., 2006) links cytoplasmic sensing of actin polymerization, ciliogenesis, and transcriptional activity in response to serum. KDM3A is a versatile protein that is tightly regulated in response to environmental inputs (Pollard et al., 2008; Krieg et al., 2010; Cheng et al., 2014; Kasioulis et al., 2014; Abe et al., 2015) with both catalytic and structural roles (Abe et al., 2015; Schneider et al., 2015) involving binding to histone (Yamane et al., 2006) and nonhistone proteins (Ramadoss et al., 2017) within nuclear and cytoplasmic compartments (Okada et al., 2007; Yang et al., 2009; Yamada et al., 2012; Kasioulis et al., 2014; Kaukonen et al., 2016). KDM3A thus appears to be capable of integrating epigenetic and cytoskeletal regulation with ciliary IFT in response to extracellular cues.

While our manuscript was in press, actin-mediated scission of the tips of mammalian primary cilia was uncovered as a mechanism to clear activated ciliary receptors (Nager et al., 2017) or to promote ciliary disassembly (Phua et al., 2017), independent of retrograde IFT involvement. The behavior of KDM3Anull cilia is consistent with these findings; through its role modulating actin dynamics, KDM3A would also modulate ciliary ectocytosis at the tip. The relative contribution of unrestricted IFT recruitment to cilia or reduced IFT clearance through ectosytosis in KDM3A mutants is an exciting topic for future research.

Materials and methods

Plasmids

Plasmids encoding the serotonin receptor pEGFPN3-5HT6 (Berbari et al., 2008), IFT81-DDK, and Halo-KDM3A were purchased from Addgene (35624), OriGene (RC216437), and Kazusa DNA Res Institute (FHC05559), respectively. IFT81-mGFP was made by transferring Sgf1-Mlu1 restriction fragment from clone RC216437 into mTagGFP vector PS100048 (OriGene). The KDM3A deletion mutant GFP-ZFNR was cloned using PCR primers containing Bgl2 and Kpn1 sites for cloning into EGFP-C1 vector (Takara Bio Inc.; listed in Table S1) with Halo-KDM3A as template and PfuUltra II DNA polymerase (Agilent Technologies), subsequently confirmed by sequencing. KDM3Ai2-EGFP (Kasioulis et al., 2014) contains all the functional domains of full-length KDM3A except for a nuclear localization signal (Brauchle et al., 2013).

Kdm3a mouse model

The deletion mutant (Kdm3aΔJC) and hypomorphic gene-tarp (Kdm3aGT) mouse models were described previously (Tateishi et al., 2009; Kasioulis et al., 2014). Genotyping was done with PCR primers as in Table S1.

Cell lines and CRISPR/Cas9 mutants

Primary MEF cultures were established by mincing embryonic day 11.5–13.5 embryos in DMEM supplemented with antibiotics, FCS, and β-mercaptoethanol and grown in 3% O2. Telomerase immortalized human retinal pigment epithelial cells (hTERT-RPE1, referred to here as RPE1WT) were from Takara Bio Inc. and maintained in DMEM/F-12 (1:1; Thermo Fisher Scientific). RPE1WT cells were engineered with a double nicking strategy using CRISPR/Cas9 nickases following reported guidelines (Ran et al., 2013). Oligonucleotides to produce guide RNAs (listed in Table S1) were designed with CRISPR Design Tool (http://tools.genome-engineering.org). In brief, guides cloned onto pX461 or pX462 (Addgene) mammalian expression plasmids were cotransfected onto RPE1WT cells followed by 24 h of 3 µg/ml puromycin selection and individual GFP-positive cells sorted by FACS onto 96-well plates to establish the RPE1 lines KDM3Anull and IFT81null. RPE1CrWT derives from noneventful CRISPR targeting of IFT81, subsequent sequencing confirmed wild-type alleles for both IFT81 and KDM3A. Mutations were confirmed by immunoblotting with KDM3A and IFT81 antibodies and sequencing of genomic PCR (primers are listed in Table S1) cloned into pGEM-T Easy (Promega) using T7 and SP6 primers.

Transfections

Parental and engineered RPE1 lines were transfected with Lipofectamine2000 (Thermo Fisher Scientific) following the manufacturer’s instructions, grown without antibiotic the day before, and transfected in the absence of serum for 5 h. All cotransfections were done mixing the DNA of the relevant cotransfected plasmids before the addition of transfection reagents.

Antibodies

Antibodies were from the following sources: N-KDM3A (12835; Proteintech), C-KDM3A (NB100-77282; Novus Biologicals), anti–acetylated α-tubulin (T7451; Sigma-Aldrich), γ-tubulin (GTU-88; Sigma-Aldrich), IFT88 (13967-1-AP; Proteintech), IFT81 (11744-1-AP; Proteintech), GAPDH (5019A-2; ImGENEX), ARL13B (17711-1-AP; Proteintech), DDK tag (TA50011; OriGene), and Halo (G9281; Promega). Secondary antibodies for immunofluorescence and immunoblotting are Fab′2 IgG Alexa Fluor from Molecular Probes and HRP conjugated from EMD Millipore and Sigma-Aldrich.

Drug treatments

CytoD (25023; EMD Millipore), CK-869 (C9124; Sigma-Aldrich), and jasplakinolide (1689-05; BioVision) were dissolved as recommended by the manufacturers and used at the indicated concentrations. Controls for each drug treatment contained the relevant solvent only.

Ciliary phases

For the study of cilia dynamics, cells were plated 24 h in advance. In all cases, stocks of all cell lines were maintained at comparable cell densities. For the studies of ciliogenesis in exponentially growing cultures, cells were not allowed to reach confluency during two consecutive passages before assaying ensuring a consistent number of cycling cells. For cilia resorption studies, cells were plated between 40% and 50% confluency followed by a further 24, 48, or 72 h culture without serum (−FCS, as indicated in each panel); for ciliary growth assays (3–24 h), cells were plated at 60–70% confluency. Resorption was induced by replacing serum-free media with media containing 10% FCS (+FCS) for the indicated times. Temperature-induced resorption was done by adding 10% FCS and floating the tissue culture plates on a water bath maintained at 42°C for the indicated times.

Cell spreading assays

Actively growing cultures were trypsinized and resuspended in complete media, replated immediately, or spun down and placed in suspension without serum for 90 min in a cell incubator. Replated cells were imaged 90–120 min after replating on uncoated plastic plates.

Yeast two-hybrid studies

The initial yeast two-hybrid screen was performed using an embryonic eye mouse cDNA library (Kasioulis et al., 2014) with the JMJC domain of KDM3A as bait cloned through partial EcoRI digestion into the DNA-binding domain encoding plasmid pGBKT7. Bait and prey plasmids from blue colonies growing in quadruple selection (Leu/Trp/Ade/His) were rescued by cell lysis and bacterial transformation in ampicillin (pGADT7) or kanamycin (pGBKT7) plates and sequenced. Plasmids encoding CAPZB and ACTA2 were cotransformed back into the yeast two-hybrid Gold strain (Takara Bio Inc.) along with JMJC-pGKBT7, as described in the Matchmaker two-hybrid system protocol. Independent colonies grown in double selection (Leu/Trp) where resuspended, and a 1/100 dilution was replated in 0.2 µg/ml aureobasidin A double selection plates in the presence of X-α-Gal (Takara Bio Inc.).

Immunofluorescence

Cells were grown on glass coverslips and fixed (10 min, 4% methanol-free formaldehyde), except for those shown in Fig. 6 (C and D), which were fixed with ice-cold methanol for 5 min. Fixed cells were permeabilized (15 min in 0.25% Triton X-100) and blocked in 2% BSA before incubation with antibodies and DAPI and mounted in Prolong Gold (Thermo Fisher Scientific). Rhodamine–phalloidin (R415; Thermo Fisher Scientific) added to the secondary antibody incubation was used to visualize F-actin of fixed cells.

Imaging

Live-cell microscopy was done using glass-bottom plates (662892; Greiner Bio-one) in a heated chamber with a source of nitrogen using a 40× plan Fluor 0.75 NA dry lens in an A1R confocal microscope (Nikon) in the relevant culture media for each growth condition. Movies shown were acquired as to minimize photobleaching. Postacquisition analyses were done with NIS-Elements (Nikon) on maximum-intensity projections using Fiji (ImageJ) and IMARIS 8.4 (Bitplane). Ciliary length studies on fixed cells were done with wide-field microscopy or confocal imaging as indicated in each panel using a 40× dry lens and FIJI software for measurements.

Cell lysis and immunopurifications

Total extracts for immunoblots were obtained by lysing cells in 150 mM Tris-HCl, pH 7.5, 50 mM NaCl, 0.5% IGEPAL, 1% Triton X-100, 1 mM EDTA, 0.5% deoxycholate, Benzoase DNase (EMD Millipore), and Halt protease inhibitor (1860932; Thermo Fisher Scientific). Total cell extracts for total cell proteome were done in PBS and 1% SDS in the presence of protease inhibitor, sonicated, and cleared by 10-min centrifugation at 14,000 rpm. For KDM3A interactome, RPE1WT (in duplicate), RPE1WTcr, and KDM3Anull (line 2, in triplicate) cells were plated the day before and grown in the absence of serum for 3 h, lysed in 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10% glycerol, 5 mM EDTA, 0.5% IGEPAL, and protease inhibitors and incubated with N-KDM3A antibody (12835; Proteintech) for 2 h at 4°C followed by 15-min incubation with PureProteome magnetic beads (EMD Millipore) to concentrate the immunoglobulin complexes. Three successive washes were performed to reduce the IGEPAL concentration.

Far–Western blot

Serial dilutions (50–200 ng/µl) of pure actin derived from rabbit skeletal muscle (AKL99; Cytoskeleton, Inc.) resolved in 4–12% acrylamide were transferred onto nitrocellulose and stained with Ponceau red to visualize loadings. Bacterially expressed GST or GST-JMJC immobilized in glutathione magnetic beads (8821; Proteintech) were washed and eluted in glutathione buffer. Membrane blocked with 5% BSA was incubated with eluted GST proteins in 10 mM Tris, pH 7.6, 100 mM KCl, 1 mM MgCl2, 0.1 mM EDTA, 1 mM DTT, 0.1% (vol/vol) Tween-20, and 2% BSA for 1 h. Subsequent steps followed standard immunoblotting protocol using GST antibody to detect GST proteins retained on membrane.

Mass spectrometry

On-bead purified complexes were digested with trypsin, reduced, alkylated, and desalted, and two runs of each sample were analyzed by liquid chromatography tandem mass spectrometry as described previously (Turriziani et al., 2014). Datasets can be found in ProteomeXchange identifier PXD004334. Data were analyzed using the MaxQuant software suite by searching against the UniProt HUMAN database with protein N terminus (acetylated), M (oxidation). Label-free quantification values were determined by MaxQuant as described previously (Cox et al., 2014). In the search for posttranslational modifications, databases were also searched for peptides with mass modifications as follows: mono-, di-, and trimethylation (delta mass of 14.0157, 28.0313, and 42.04695 D, respectively).

RNA sequencing

RNA was isolated from RPE1WT and KDM3Anull (line 2) cells in triplicate cultures under ciliary growing or resorption conditions (Fig. 3 A) using RNeasy Plus (QIAGEN) according to the manufacturer’s instructions. RNA integrity score was calculated with RNA 6000 Nano reagent (Agilent Technologies) in a 2100 Bioanalyzer (Agilent Technologies). All samples used for the preparation of the RNA-sequencing library had a RIN (RNA Integrity Value) score above 9. Library construction used poly(A) RNA selection, and all 100-bp reads were obtained using Illumina HiSeq2000. Cufflinks, CummeRbund, and Picard were used for alignments to reference genome, transcript assembly, quantification, and statistical analysis, respectively, done at Oxford Gene Technology (The Molecular Genetics Company). The mean read depth for each sample was of 13 × 106.

GO term analyses

Genes or proteins differentially enriched in proteome or interactome datasets were analyzed using DAVID (the Database for Annotation Visualization and Integrated Discovery).

Statistics

Statistics were done using GraphPad Prism 5 software. One-way analysis of variance was used for the initial assessment of multiple comparisons. In those cases where a significant difference was observed (P < 0.01), a pairwise two-tailed Students t test was applied to compare individual conditions. χ2 t test was applied to compare percentages as indicated in each panel. Asterisks indicate p-values (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001) calculated through the statistical analysis indicated in each case. N represents the number of experimental or biological replicates, and n represents the number of events measured in all experimental or biological replicates as indicate.

Online supplemental material

Table S1 shows sequences of primers used to clone KDM3A expression plasmids, CRISPR/Cas9 editing, and all genotyping. Table S2 contains raw values of RNA sequencing reads and proteomic studies. Table S3 shows thresholding criteria and the resulting lists of genes that were subsequently used for GO term analysis. Table S4 shows enriched GO terms. Fig. S1 shows genotypes of KDM3Anull cells, ciliary anomalies of Kdm3aGT/GT hypomorph MEFs, and effect of overexpressing KDM3A in the percentage of ciliated cells. Fig. S2 shows phalloidin staining of Kdm3aΔJC/ΔJC MEFs, assessment of KDM3A interaction with actin by two-hybrid screening and far–Western blotting, and detailed measurement criteria used to quantify the extent of filopodia in transfected cells. Fig. S3 shows genotyping of IFT81null cells, functionality of IFT81-DDK construct, and identical transfection efficiencies of 5HT6-EGFP+IFT81-DDK onto RPE1WT and KDM3Anull cells. Videos 1 and 2 show instability of KDM3Anull cilia. Video 3 shows instability of Kdm3aΔJC/ΔJC cilia. Video 4 illustrates the unresponsive behavior of KDM3Anull cilia to serum-induced resorption. Video 5 illustrates that KDM3Ai2-EGFP can be found in cilia.

Acknowledgments

We thank the Structural Genomics Consortium (Oxford) for assistance in protein purification for catalytic studies and the Institute of Genetics and Molecular Medicine technical staff for support with animal husbandry, imaging, and sequencing. We also thank Georgios Kanellos and Brad J. Nolen for expert advice.

C.J. Schofield thanks Cancer Research UK and the Wellcome Trust for funding. This work was supported by core funding from the Medical Research Council (MC_UU_12018/26; P.L. Yeyati, G. Mali, I. Kasioulis, and P. Mill).

The authors declare no competing financial interests.

Author contributions: P.L. Yeyati contributed to the design, execution, and interpretation of the overall research plan. J. Willis, A. von Kriegsheim, and A. Finch performed mass spectrometric analyses, which were interpreted by A. von Kriegsheim and G. Mali. R. Schiller, A. Kawamura, and C.J. Schofield tested KDM3A catalytic activity. I. Kasioulis contributed to KDM3A cloning and subcellular localization experiments. I.R. Adams and C. Playfoot explored KDM3A chromatin binding. N. Gilbert, V. van Heyningen, and I.J. Jackson contributed to experimental design and provided expert advice. J. Sakai shared unpublished data and expert advice. P. Mill oversaw experimental design and interpretation. P.L. Yeyati, C.J. Schofield, and P. Mill wrote the manuscript with contributions from all authors.

References

References
Abe
,
Y.
,
R.
Rozqie
,
Y.
Matsumura
,
T.
Kawamura
,
R.
Nakaki
,
Y.
Tsurutani
,
K.
Tanimura-Inagaki
,
A.
Shiono
,
K.
Magoori
,
K.
Nakamura
, et al
2015
.
JMJD1A is a signal-sensing scaffold that regulates acute chromatin dynamics via SWI/SNF association for thermogenesis
.
Nat. Commun.
6
:
7052
.
Avasthi
,
P.
,
M.
Onishi
,
J.
Karpiak
,
R.
Yamamoto
,
L.
Mackinder
,
M.C.
Jonikas
,
W.S.
Sale
,
B.
Shoichet
,
J.R.
Pringle
, and
W.F.
Marshall
.
2014
.
Actin is required for IFT regulation in Chlamydomonas reinhardtii.
Curr. Biol.
24
:
2025
2032
.
Berbari
,
N.F.
,
A.D.
Johnson
,
J.S.
Lewis
,
C.C.
Askwith
, and
K.
Mykytyn
.
2008
.
Identification of ciliary localization sequences within the third intracellular loop of G protein-coupled receptors
.
Mol. Biol. Cell.
19
:
1540
1547
.
Beyer
,
S.
,
M.M.
Kristensen
,
K.S.
Jensen
,
J.V.
Johansen
, and
P.
Staller
.
2008
.
The histone demethylases JMJD1A and JMJD2B are transcriptional targets of hypoxia-inducible factor HIF
.
J. Biol. Chem.
283
:
36542
36552
.
Blanchoin
,
L.
,
R.
Boujemaa-Paterski
,
C.
Sykes
, and
J.
Plastino
.
2014
.
Actin dynamics, architecture, and mechanics in cell motility
.
Physiol. Rev.
94
:
235
263
.
Brauchle
,
M.
,
Z.
Yao
,
R.
Arora
,
S.
Thigale
,
I.
Clay
,
B.
Inverardi
,
J.
Fletcher
,
P.
Taslimi
,
M.G.
Acker
,
B.
Gerrits
, et al
2013
.
Protein complex interactor analysis and differential activity of KDM3 subfamily members towards H3K9 methylation
.
PLoS One.
8
:
e60549
.
Cao
,
J.
,
Y.
Shen
,
L.
Zhu
,
Y.
Xu
,
Y.
Zhou
,
Z.
Wu
,
Y.
Li
,
X.
Yan
, and
X.
Zhu
.
2012
.
miR-129-3p controls cilia assembly by regulating CP110 and actin dynamics
.
Nat. Cell Biol.
14
:
697
706
.
Chakraborty
,
D.
,
W.
Cui
,
G.X.
Rosario
,
R.L.
Scott
,
P.
Dhakal
,
S.J.
Renaud
,
M.
Tachibana
,
M.A.K.
Rumi
,
C.W.
Mason
,
A.J.
Krieg
, and
M.J.
Soares
.
2016
.
HIF-KDM3A-MMP12 regulatory circuit ensures trophoblast plasticity and placental adaptations to hypoxia
.
Proc. Natl. Acad. Sci. USA.
113
:
E7212
E7221
.
Cheng
,
M.-B.
,
Y.
Zhang
,
C.-Y.
Cao
,
W.-L.
Zhang
,
Y.
Zhang
, and
Y.F.
Shen
.
2014
.
Specific phosphorylation of histone demethylase KDM3A determines target gene expression in response to heat shock
.
PLoS Biol.
12
:
e1002026
.
Cho
,
H.-S.
,
G.
Toyokawa
,
Y.
Daigo
,
S.
Hayami
,
K.
Masuda
,
N.
Ikawa
,
Y.
Yamane
,
K.
Maejima
,
T.
Tsunoda
,
H.I.
Field
, et al
2012
.
The JmjC domain-containing histone demethylase KDM3A is a positive regulator of the G1/S transition in cancer cells via transcriptional regulation of the HOXA1 gene
.
Int. J. Cancer.
131
:
E179
E189
.
Clissold
,
P.M.
, and
C.P.
Ponting
.
2001
.
JmjC: cupin metalloenzyme-like domains in jumonji, hairless and phospholipase A2β
.
Trends Biochem. Sci.
26
:
7
9
.
Cox
,
J.
,
M.Y.
Hein
,
C.A.
Luber
,
I.
Paron
,
N.
Nagaraj
, and
M.
Mann
.
2014
.
Accurate proteome-wide label-free quantification by delayed normalization and maximal peptide ratio extraction, termed MaxLFQ
.
Mol. Cell. Proteomics.
13
:
2513
2526
.
Das
,
R.M.
, and
K.G.
Storey
.
2014
.
Apical abscission alters cell polarity and dismantles the primary cilium during neurogenesis
.
Science.
343
:
200
204
.
Edwards
,
M.
,
A.
Zwolak
,
D.A.
Schafer
,
D.
Sept
,
R.
Dominguez
, and
J.A.
Cooper
.
2014
.
Capping protein regulators fine-tune actin assembly dynamics
.
Nat. Rev. Mol. Cell Biol.
15
:
677
689
.
Farina
,
F.
,
J.
Gaillard
,
C.
Guérin
,
Y.
Couté
,
J.
Sillibourne
,
L.
Blanchoin
, and
M.
Théry
.
2016
.
The centrosome is an actin-organizing centre
.
Nat. Cell Biol.
18
:
65
75
.
Fu
,
W.
,
L.
Wang
,
S.
Kim
,
J.
Li
, and
B.D.
Dynlacht
.
2016
.
Role for the IFT-A complex in selective transport to the primary cilium
.
Cell Reports.
17
:
1505
1517
.
Han
,
Y.M.
,
G.M.
Kang
,
K.
Byun
,
H.W.
Ko
,
J.
Kim
,
M.S.
Shin
,
H.K.
Kim
,
S.Y.
Gil
,
J.H.
Yu
,
B.
Lee
, and
M.S.
Kim
.
2014
.
Leptin-promoted cilia assembly is critical for normal energy balance
.
J. Clin. Invest.
124
:
2193
2197
.
Hernandez-Hernandez
,
V.
,
P.
Pravincumar
,
A.
Diaz-Font
,
H.
May-Simera
,
D.
Jenkins
,
M.
Knight
, and
P.L.
Beales
.
2013
.
Bardet-Biedl syndrome proteins control the cilia length through regulation of actin polymerization
.
Hum. Mol. Genet.
22
:
3858
3868
.
Hetrick
,
B.
,
M.S.
Han
,
L.A.
Helgeson
, and
B.J.
Nolen
.
2013
.
Small molecules CK-666 and CK-869 inhibit actin-related protein 2/3 complex by blocking an activating conformational change
.
Chem. Biol.
20
:
701
712
.
Inagaki
,
T.
,
M.
Tachibana
,
K.
Magoori
,
H.
Kudo
,
T.
Tanaka
,
M.
Okamura
,
M.
Naito
,
T.
Kodama
,
Y.
Shinkai
, and
J.
Sakai
.
2009
.
Obesity and metabolic syndrome in histone demethylase JHDM2a-deficient mice
.
Genes Cells.
14
:
991
1001
.
Kang
,
G.M.
,
Y.M.
Han
,
H.W.
Ko
,
J.
Kim
,
B.C.
Oh
,
I.
Kwon
, and
M.-S.
Kim
.
2015
.
Leptin elongates hypothalamic neuronal cilia via transcriptional regulation and actin destabilization
.
J. Biol. Chem.
290
:
18146
18155
.
Kasioulis
,
I.
,
H.M.
Syred
,
P.
Tate
,
A.
Finch
,
J.
Shaw
,
A.
Seawright
,
M.
Fuszard
,
C.H.
Botting
,
S.
Shirran
,
I.R.
Adams
, et al
2014
.
Kdm3a lysine demethylase is an Hsp90 client required for cytoskeletal rearrangements during spermatogenesis
.
Mol. Biol. Cell.
25
:
1216
1233
.
Kaukonen
,
R.
,
A.
Mai
,
M.
Georgiadou
,
M.
Saari
,
N.
De Franceschi
,
T.
Betz
,
H.
Sihto
,
S.
Ventelä
,
L.
Elo
,
E.
Jokitalo
, et al
2016
.
Normal stroma suppresses cancer cell proliferation via mechanosensitive regulation of JMJD1a-mediated transcription
.
Nat. Commun.
7
:
12237
.
Kierszenbaum
,
A.L.
2002
.
Intramanchette transport (IMT): managing the making of the spermatid head, centrosome, and tail
.
Mol. Reprod. Dev.
63
:
1
4
.
Kim
,
J.
,
J.E.
Lee
,
S.
Heynen-Genel
,
E.
Suyama
,
K.
Ono
,
K.
Lee
,
T.
Ideker
,
P.
Aza-Blanc
, and
J.G.
Gleeson
.
2010
.
Functional genomic screen for modulators of ciliogenesis and cilium length
.
Nature.
464
:
1048
1051
.
Kim
,
J.
,
H.
Jo
,
H.
Hong
,
M.H.
Kim
,
J.M.
Kim
,
J.-K.
Lee
,
W.D.
Heo
, and
J.
Kim
.
2015
.
Actin remodelling factors control ciliogenesis by regulating YAP/TAZ activity and vesicle trafficking
.
Nat. Commun.
6
:
6781
.
Klose
,
R.J.
,
E.M.
Kallin
, and
Y.
Zhang
.
2006
.
JmjC-domain-containing proteins and histone demethylation
.
Nat. Rev. Genet.
7
:
715
727
.
Koestler
,
S.A.
,
A.
Steffen
,
M.
Nemethova
,
M.
Winterhoff
,
N.
Luo
,
J.M.
Holleboom
,
J.
Krupp
,
S.
Jacob
,
M.
Vinzenz
,
F.
Schur
, et al
2013
.
Arp2/3 complex is essential for actin network treadmilling as well as for targeting of capping protein and cofilin
.
Mol. Biol. Cell.
24
:
2861
2875
.
Krieg
,
A.J.
,
E.B.
Rankin
,
D.
Chan
,
O.
Razorenova
,
S.
Fernandez
, and
A.J.
Giaccia
.
2010
.
Regulation of the histone demethylase JMJD1A by hypoxia-inducible factor 1 alpha enhances hypoxic gene expression and tumor growth
.
Mol. Cell. Biol.
30
:
344
353
.
Kustermans
,
G.
,
J.
Piette
, and
S.
Legrand-Poels
.
2008
.
Actin-targeting natural compounds as tools to study the role of actin cytoskeleton in signal transduction
.
Biochem. Pharmacol.
76
:
1310
1322
.
Liang
,
Y.
,
D.
Meng
,
B.
Zhu
, and
J.
Pan
.
2016
.
Mechanism of ciliary disassembly
.
Cell. Mol. Life Sci.
73
:
1787
1802
.
Lockman
,
K.
,
J.M.
Taylor
, and
C.P.
Mack
.
2007
.
The histone demethylase, Jmjd1a, interacts with the myocardin factors to regulate SMC differentiation marker gene expression
.
Circ. Res.
101
:
e115
e123
.
Markolovic
,
S.
,
T.M.
Leissing
,
R.
Chowdhury
,
S.E.
Wilkins
,
X.
Lu
, and
C.J.
Schofield
.
2016
.
Structure-function relationships of human JmjC oxygenases-demethylases versus hydroxylases
.
Curr. Opin. Struct. Biol.
41
:
62
72
.
Nager
,
A.R.
,
J.S.
Goldstein
,
V.
Herranz-Pérez
,
D.
Portran
,
F.
Ye
,
J.M.
Garcia-Verdugo
, and
M.V.
Nachury
.
2017
.
An Actin network dispatches ciliary GPCRs into extracellular vesicles to modulate signaling
.
Cell.
168
:
252
263.e14
.
Ohguchi
,
H.
,
T.
Hideshima
,
M.K.
Bhasin
,
G.T.
Gorgun
,
L.
Santo
,
M.
Cea
,
M.K.
Samur
,
N.
Mimura
,
R.
Suzuki
,
Y.-T.
Tai
, et al
2016
.
The KDM3A-KLF2-IRF4 axis maintains myeloma cell survival
.
Nat. Commun.
7
:
10258
.
Okada
,
Y.
,
G.
Scott
,
M.K.
Ray
,
Y.
Mishina
, and
Y.
Zhang
.
2007
.
Histone demethylase JHDM2A is critical for Tnp1 and Prm1 transcription and spermatogenesis
.
Nature.
450
:
119
123
.
Okada
,
Y.
,
K.
Tateishi
, and
Y.
Zhang
.
2010
.
Histone demethylase JHDM2A is involved in male infertility and obesity
.
J. Androl.
31
:
75
78
.
Olson
,
E.N.
, and
A.
Nordheim
.
2010
.
Linking actin dynamics and gene transcription to drive cellular motile functions
.
Nat. Rev. Mol. Cell Biol.
11
:
353
365
.
Osawa
,
T.
,
R.
Tsuchida
,
M.
Muramatsu
,
T.
Shimamura
,
F.
Wang
,
J.
Suehiro
,
Y.
Kanki
,
Y.
Wada
,
Y.
Yuasa
,
H.
Aburatani
, et al
2013
.
Inhibition of histone demethylase JMJD1A improves anti-angiogenic therapy and reduces tumor-associated macrophages
.
Cancer Res.
73
:
3019
3028
.
Parrish
,
J.K.
,
M.
Sechler
,
R.A.
Winn
, and
P.
Jedlicka
.
2015
.
The histone demethylase KDM3A is a microRNA-22-regulated tumor promoter in Ewing Sarcoma
.
Oncogene.
34
:
257
262
.
Pedanou
,
V.E.
,
S.
Gobeil
,
S.
Tabariès
,
T.M.
Simone
,
L.J.
Zhu
,
P.M.
Siegel
, and
M.R.
Green
.
2016
.
The histone H3K9 demethylase KDM3A promotes anoikis by transcriptionally activating pro-apoptotic genes BNIP3 and BNIP3L
.
eLife.
5
:
1
16
.
Phua
,
S.C.
,
S.
Chiba
,
M.
Suzuki
,
E.
Su
,
E.C.
Roberson
,
G.V.
Pusapati
,
M.
Setou
,
R.
Rohatgi
,
J.F.
Reiter
,
K.
Ikegami
, and
T.
Inoue
.
2017
.
Dynamic remodeling of membrane composition drives cell cycle through primary cilia excision
.
Cell.
168
:
264
279.e15
.
Pitaval
,
A.
,
Q.
Tseng
,
M.
Bornens
, and
M.
Théry
.
2010
.
Cell shape and contractility regulate ciliogenesis in cell cycle-arrested cells
.
J. Cell Biol.
191
:
303
312
.
Pollard
,
T.D.
, and
G.G.
Borisy
.
2003
.
Cellular motility driven by assembly and disassembly of actin filaments
.
Cell.
112
:
453
465
.
Pollard
,
P.J.
,
C.
Loenarz
,
D.R.
Mole
,
M.A.
McDonough
,
J.M.
Gleadle
,
C.J.
Schofield
, and
P.J.
Ratcliffe
.
2008
.
Regulation of Jumonji-domain-containing histone demethylases by hypoxia-inducible factor (HIF)-1alpha
.
Biochem. J.
416
:
387
394
.
Pugacheva
,
E.N.
,
S.A.
Jablonski
,
T.R.
Hartman
,
E.P.
Henske
, and
E.A.
Golemis
.
2007
.
HEF1-dependent Aurora A activation induces disassembly of the primary cilium
.
Cell.
129
:
1351
1363
.
Ramadoss
,
S.
,
G.
Guo
, and
C.-Y.
Wang
.
2017
.
Lysine demethylase KDM3A regulates breast cancer cell invasion and apoptosis by targeting histone and the non-histone protein p53
.
Oncogene.
36
:
47
59
.
Ran
,
F.A.
,
P.D.
Hsu
,
J.
Wright
,
V.
Agarwala
,
D.A.
Scott
, and
F.
Zhang
.
2013
.
Genome engineering using the CRISPR-Cas9 system
.
Nat. Protoc.
8
:
2281
2308
.
San Agustin
,
J.T.
,
G.J.
Pazour
, and
G.B.
Witman
.
2015
.
Intraflagellar transport is essential for mammalian spermiogenesis but is absent in mature sperm
.
Mol. Biol. Cell.
26
:
4358
4372
.
Schneider
,
P.
,
J.M.
Bayo-Fina
,
R.
Singh
,
P.
Kumar Dhanyamraju
,
P.
Holz
,
A.
Baier
,
V.
Fendrich
,
A.
Ramaswamy
,
S.
Baumeister
,
E.D.
Martinez
, and
M.
Lauth
.
2015
.
Identification of a novel actin-dependent signal transducing module allows for the targeted degradation of GLI1
.
Nat. Commun.
6
:
8023
.
Spalluto
,
C.
,
D.I.
Wilson
, and
T.
Hearn
.
2013
.
Evidence for reciliation of RPE1 cells in late G1 phase, and ciliary localisation of cyclin B1
.
FEBS Open Bio.
3
:
334
340
.
Tateishi
,
K.
,
Y.
Okada
,
E.M.
Kallin
, and
Y.
Zhang
.
2009
.
Role of Jhdm2a in regulating metabolic gene expression and obesity resistance
.
Nature.
458
:
757
761
.
Tee
,
A.E.
,
D.
Ling
,
C.
Nelson
,
B.
Atmadibrata
,
M.E.
Dinger
,
N.
Xu
,
T.
Mizukami
,
P.Y.
Liu
,
B.
Liu
,
B.
Cheung
, et al
2014
.
The histone demethylase JMJD1A induces cell migration and invasion by up-regulating the expression of the long noncoding RNA MALAT1
.
Oncotarget.
5
:
1793
1804
.
Turriziani
,
B.
,
A.
Garcia-Munoz
,
R.
Pilkington
,
C.
Raso
,
W.
Kolch
, and
A.
von Kriegsheim
.
2014
.
On-beads digestion in conjunction with data-dependent mass spectrometry: a shortcut to quantitative and dynamic interaction proteomics
.
Biology (Basel).
3
:
320
332
https://doi.org/10.3390/biology3020320.
Uemura
,
M.
,
H.
Yamamoto
,
I.
Takemasa
,
K.
Mimori
,
H.
Hemmi
,
T.
Mizushima
,
M.
Ikeda
,
M.
Sekimoto
,
N.
Matsuura
,
Y.
Doki
, and
M.
Mori
.
2010
.
Jumonji domain containing 1A is a novel prognostic marker for colorectal cancer: in vivo identification from hypoxic tumor cells
.
Clin. Cancer Res.
16
:
4636
4646
.
Wade
,
M.A.
,
D.
Jones
,
L.
Wilson
,
J.
Stockley
,
K.
Coffey
,
C.N.
Robson
, and
L.
Gaughan
.
2015
.
The histone demethylase enzyme KDM3A is a key estrogen receptor regulator in breast cancer
.
Nucleic Acids Res.
43
:
196
207
.
Wellmann
,
S.
,
M.
Bettkober
,
A.
Zelmer
,
K.
Seeger
,
M.
Faigle
,
H.K.
Eltzschig
, and
C.
Bührer
.
2008
.
Hypoxia upregulates the histone demethylase JMJD1A via HIF-1
.
Biochem. Biophys. Res. Commun.
372
:
892
897
.
Wolfenson
,
H.
,
T.
Iskratsch
, and
M.P.
Sheetz
.
2014
.
Early events in cell spreading as a model for quantitative analysis of biomechanical events
.
Biophys. J.
107
:
2508
2514
.
Yamada
,
D.
,
S.
Kobayashi
,
H.
Yamamoto
,
Y.
Tomimaru
,
T.
Noda
,
M.
Uemura
,
H.
Wada
,
S.
Marubashi
,
H.
Eguchi
,
M.
Tanemura
, et al
2012
.
Role of the hypoxia-related gene, JMJD1A, in hepatocellular carcinoma: clinical impact on recurrence after hepatic resection
.
Ann. Surg. Oncol.
19
(
S3
,
Suppl 3
):
S355
S364
.
Yamane
,
K.
,
C.
Toumazou
,
Y.
Tsukada
,
H.
Erdjument-Bromage
,
P.
Tempst
,
J.
Wong
, and
Y.
Zhang
.
2006
.
JHDM2A, a JmjC-containing H3K9 demethylase, facilitates transcription activation by androgen receptor
.
Cell.
125
:
483
495
.
Yang
,
J.
,
I.
Ledaki
,
H.
Turley
,
K.C.
Gatter
,
J.-C.M.
Montero
,
J.-L.
Li
, and
A.L.
Harris
.
2009
.
Role of hypoxia-inducible factors in epigenetic regulation via histone demethylases
.
Ann. N. Y. Acad. Sci.
1177
:
185
197
.
Zhao
,
Q.-Y.
,
P.-J.
Lei
,
X.
Zhang
,
J.-Y.
Zheng
,
H.-Y.
Wang
,
J.
Zhao
,
Y.-M.
Li
,
M.
Ye
,
L.
Li
,
G.
Wei
, and
M.
Wu
.
2016
.
Global histone modification profiling reveals the epigenomic dynamics during malignant transformation in a four-stage breast cancer model
.
Clin. Epigenetics.
8
:
34
.

    Abbreviations used:
     
  • CytoD

    cytochalasin D

  •  
  • IFT

    intraflagellar transport

  •  
  • JmjC

    Jumonji C

  •  
  • MEF

    mouse embryonic fibroblast

Author notes

V. van Heyningen's present address is Institute of Ophthalmology, University College London, EC1V 9EL London, England, UK.

This article is available under a Creative Commons License (Attribution 4.0 International, as described at https://creativecommons.org/licenses/by/4.0/).

Supplementary data