MTMR regulates KRAS function by controlling plasma membrane levels of phospholipids

H-, N-, and KRAS proteins are small GTPases that oscillate between inactive GDP-bound and active GTP-bound conformational states and operate in signaling cascades that control cell growth and proliferation (Gorfe and Cho, 2019; Henkels et al., 2021). Consistent with this key regulatory role, activating mutations of RAS are present in ∼20% of human cancers, with the majority occurring in the KRAS isoform (Prior et al., 2020). All RAS isoforms must be localized to the inner leaflet of the plasma membrane (PM) and be spatially organized into nanoclusters for biological activity (Gorfe and Cho, 2019; Henkels et al., 2021). Membrane interactions are mediated by the RAS C-terminal membrane anchor (Zhou et al., 2017). In the case of KRAS4B, the major expressed splice variant of KRAS (Hood et al., 2023) (hereafter KRAS), this anchor comprises a C-terminal farnesylcysteine methyl ester, generated by posttranslational modification, and an adjacent polybasic domain of six continuous lysines (Gutierrez et al., 1989; Hancock et al., 1989, 1990, 1991). Recent work has shown that the KRAS membrane anchor encodes exquisite binding specificity for phosphatidylserine (PtdSer) lipids with one saturated and one desaturated acyl chain (Zhou et al., 2017). This lipid binding specificity is effectively hard-wired into the anchor structure and renders KRAS PM binding, nanoclustering, and hence biological function absolutely dependent on PM PtdSer content (Cho et al., 2012b, 2013; van der Hoeven et al., 2013, 2018).

PtdSer is synthesized in the ER and delivered to the PM against a concentration gradient by the lipid transfer proteins (LTPs), oxysterol-binding protein–related proteins (ORP) 5 and 8, operating at ER-PM membrane-contacting sites (MCSs). One molecule of phosphatidylinositol (PI) 4-phosphate (PI4P), generated by PM-localized PI 4-kinase IIIα (PI4KA) from PI, is exchanged by ORP5/8 for one molecule of ER-localized PtdSer (Chung et al., 2015; Moser von Filseck et al., 2015). PI4P delivered to the ER is converted to PI by the SAC1 phosphatase, which, operating together with PI4KA, maintains the PI4P gradient that in turn drives PtdSer transport to the PM (Chung et al., 2015; Moser von Filseck et al., 2015). Concordantly genetic and pharmacological studies have shown that inhibition or loss of any component of the PtdSer transport machinery, including PI4KA, EFR3A (the PM anchor protein for the kinase), ORP5, ORP8, and SAC1, results in a reduction in PM PtdSer content, displacement of KRAS from the PM, and abrogation of KRAS signaling (Gulyas et al., 2017; Kattan et al., 2019, 2021; Sohn et al., 2018). Other pharmacological approaches to deplete PM PtdSer also dissociate KRAS from the PM (Cho et al., 2012b; Garrido et al., 2020; Miller et al., 2019; Tan et al., 2018, 2019; van der Hoeven et al., 2018). The PM is a dynamic organelle; thus, KRAS on endomembranes is sequestered by a chaperone protein, phosphodiesterase 6δ, and released to perinuclear membranes in the ARL2/3-dependent manner. The negative charge on the recycling endosome membranes then electrostatically traps KRAS, from where vesicular transport returns KRAS to the PM (Chandra et al., 2011; Schmick et al., 2014). KRAS phosphorylation at the Ser181 residue can disrupt this electrostatic interaction, depleting KRAS from the PM (Cho et al., 2016; Kovar et al., 2020).

In this context, we carried out a cell-based high-content screen using a human genomic short interfering RNA (siRNA) library to identify additional potential regulators of KRAS PM localization. Among the hits from the assay were multiple genes that encode proteins with phosphatase activity. Four of these genes encode myotubularin-related proteins (MTMR) 2, 3, 4, and 7, which have 3-phosphatase activity toward PI 3-phosphate (PI3P) and PI(3,5)-bisphosphate (PI(3,5)P₂) (Robinson and Dixon, 2006). In this study, we examined the role of MTMR in the PM localization of KRAS. Our data demonstrate that knockdown (KD) of MTMR 2, 3, 4, or 7 dissociates KRAS, but not HRAS from the PM. We show inter alia that the molecular mechanism involves depletion of the PI4P and PtdSer content of the PM, extending the repertoire of proteins required to support the operation of the ORP5/8 LTPs and hence sustain KRAS PM localization.

To identify novel regulators of KRAS PM interaction, we performed an image-based high-content screen of a human genomic siRNA library. A Caco-2 (human colorectal adenocarcinoma) cell line stably expressing monomeric green fluorescent protein (GFP)–tagged oncogenic mutant KRAS (KRASG12V) was generated and transfected with siRNA pools comprising four different oligonucleotides against each gene. Four days after transfection, cells were imaged using an automated confocal microscope to analyze the extent of KRAS mislocalization from the PM. We then narrowed the list of hits using a counter-screen against HRASG12V to identify KRASG12V-specific mediators. Bioinformatics analysis revealed that 8 of these putative KRASG12V-specific mediators encode proteins with phosphatase activity (Table 1), 4 of which are members of the MTMR family. There are 16 human myotubularins, which exhibit 3-phosphatase activity toward PI3P and PI(3,5)P₂, generating PI and PI5P, respectively (Robinson and Dixon, 2006; Wang et al., 2024). MTMR family has previously been shown to regulate many cellular processes including endocytosis, membrane trafficking, cell proliferation, autophagy, cytokinesis, and cytoskeletal dynamics (Hnia et al., 2012). We selected this MTMR gene set for further analysis.

To validate the siRNA screen, we generated short-hairpin RNA (shRNA)–mediated stable KD cell lines. T47D (human mammary gland ductal carcinoma) epithelial cells stably expressing GFP-KRASG12V were infected with lentiviruses expressing shRNA targeting MTMR 2, 3, 4, and 7, followed by puromycin selection. For each target gene, we tested 4 different shRNAs and selected the two most effective KD for further experiments. Immunoblotting verified that endogenous expression levels of MTMR 2, 3, and 7 were reduced by the cognate shRNAs (Fig. 1 A). MTMR 4 mRNA levels were also reduced by shRNA expression; we evaluated mRNA because the anti-MTMR 4 antibodies we tested were not able to detect endogenous MTMR 4 (Fig. 1 A). To study KRAS cellular localization, cells were incubated with CellMask, a dye staining cellular membranes (Hu et al., 2013; Maechler et al., 2019; Mönkemöller et al., 2015; Park et al., 2019), and imaged by confocal microscopy. To quantify the extent of KRAS distribution to endomembranes, we calculated Manders’ coefficient, which measures the fraction of CellMask colocalized with GFP-KRASG12V (Cho et al., 2012b; Manders et al., 1993; Rehl et al., 2023). Our data show that while GFP-KRASG12V was predominantly localized to the PM in control cells, it was distributed intracellularly in the KD cell lines (Fig. 1 B and Fig. S1 A). Manders’ coefficients for CellMask are correspondingly higher in the KD cell lines, indicating that KRASG12V is redistributed to endomembranes after silencing MTMR 2/3/4/7. To directly quantify the extent of KRAS dissociation from the PM, intact basal PM sheets from T47D stably co-expressing GFP-KRASG12V and shRNA for MTMR 2/3/4/7 were prepared and labeled with gold-conjugated anti-GFP antibodies and analyzed by electron microscopy (EM). We observed significant reductions in anti-GFP immunogold labeling after the KD, indicating loss of KRASG12V from the inner PM leaflet (Fig. 1 C and Fig. S1 B). Spatial mapping of GFP-KRASG12V on the PM in MTMR 2/3/4/7 KD cells revealed significant decreases in L_max, the peak value of the L(r)-r clustering statistic that measures the extent of KRAS nanoclustering, which is essential for high-fidelity RAS signal transduction (Cho and Hancock, 2013; Cho et al., 2012a; Tian et al., 2007) (Fig. 1 E). Confocal microscopy and EM further demonstrated that while MTMR 2/3/4/7 KD did not disrupt the PM localization and nanoclustering of HRASG12V, it disrupts that of the C-terminal truncated mutant KRAS (CTK), which contains the membrane anchor without the G-domain, from the PM (Fig. 1, B, D, and F; and Fig. S2). Together, our data show that MTMR 2/3/4/7 regulate the PM localization and nanoclustering of KRAS, but not HRAS, independent of the G-domain.

The PM interaction of KRAS, but not HRAS, is dependent on PM PtdSer content (Cho et al., 2012b, 2015; Garrido et al., 2020; Miller et al., 2019; Tan et al., 2018; van der Hoeven et al., 2018; Zhou et al., 2017). To quantify PM PtdSer content, T47D cells stably co-expressing GFP-LactC2, a well-characterized PtdSer probe (Yeung et al., 2008), and shRNA targeting MTMR 2/3/4/7 were incubated with CellMask and imaged by confocal microscopy. GFP-LactC2 was primarily localized to the PM in control cells but was redistributed in MTMR 2/3/4/7 KD cells as reflected by higher Manders’ coefficient values (Fig. 2 A). EM analysis of isolated intact PM sheets prepared from T47D cells expressing GFP-LactC2 and incubated with anti-GFP antibody–conjugated gold shows that gold particle labeling density of the inner basal PM was significantly reduced when cells co-express shRNAs for MTMR 2/3/4/7, indicating depletion of inner PM PtdSer (Fig. 2 B). The loss of inner leaflet PtdSer was not due to flipping to the outer leaflet since annexin V binding to the outer PM leaflet, as measured by flow cytometry, did not increase in MTMR 2/3/4/7 KD cells (Fig. 2 C). Finally, whole-cell lipidomics revealed that MTMR 2/3/4/7 KD resulted in a significant reduction in total cellular PtdSer content (Fig. 2, D and E). Together, these results show that MTMR 2/3/4/7 potently regulate the PM and total levels of PtdSer; consistent with a mechanistic role of MTMR 2/3/4/7 in controlling PtdSer cellular distribution.

In mammalian cells, inner PM leaflet PI4P and PI(4,5)P₂ recruit the LTPs, ORP5 and ORP8, to ER-PM MCSs to exchange newly synthesized PtdSer in the ER for PI4P from the PM (Chung et al., 2015; Ghai et al., 2017; Moser von Filseck et al., 2015; Sohn et al., 2018). Depleting PM PI4P or PI(4,5)P₂ concordantly depletes PtdSer and KRAS from the PM (Gulyas et al., 2017; Kattan et al., 2019). Depletion of PI4P at the Golgi complex also mislocalizes PtdSer and KRAS from the PM via an as yet unknown mechanism (Miller et al., 2019). To examine whether silencing MTMR 2/3/4/7 perturbs cellular localization of PI4P and PI(4,5)P₂, MTMR 2/3/4/7 KD T47D cells expressing GFP-P4M-SidM or GFP-PH-PLCδ1, probes for PI4P (Hammond et al., 2014) and PI(4,5)P₂ (Várnai and Balla, 1998), respectively, were imaged by confocal microscopy. GFP-P4M-SidM decorated the PM and Golgi complex in control cells (closed and open arrowheads, respectively, in Fig. 3 A), consistent with previous studies (Hammond et al., 2014; Miller et al., 2019), whereas MTMR 2/3/4/7 KD perturbed GFP-P4M-SidM staining of the PM, but not the Golgi complex (Fig. 3 A). In contrast, the PM localization of GFP-PH-PLCδ1 was unchanged in MTMR 2/3/4/7 KD cells (Fig. 3 A). EM analysis of immunogold-labeled PM sheets also shows a significant reduction in PI4P PM content, but no change in PI(4,5)P₂ PM content after MTMR 2/3/4/7 KD (Fig. 3, B and C). Given the reduction in PM PI4P content, we next examined ORP5 cellular localization. Confocal imaging showed that GFP-ORP5 predominantly localized to the ER-PM MCS in control cells, but substantially redistributed, presumably, to the ER, in MTMR 2/3/4/7 KD cells (Fig. 3 A) (Du et al., 2020; Monteiro-Cardoso et al., 2022). We conclude that MTMR 2/3/4/7 regulate ORP5 localization to the ER-PM MCS by maintaining the PM level of PI4P.

PI3P is synthesized from PI by class II and III PI3Ks, whence it can be converted to PI(3,5)P₂ or PI by PI 5-kinase, or myotubularin family phosphatases, respectively (Jean and Kiger, 2014; Pulido et al., 2013). To examine whether depleting MTMR 2/3/4/7 alters cellular PI3P levels or distribution, MTMR 2/3/4/7 KD T47D cells expressing the well-characterized PI3P probe, GFP-2xFYVE, the tandem FYVE domain of hepatocyte growth factor–regulated tyrosine kinase substrate (Gillooly et al., 2000), were stained with CellMask and imaged by confocal microscopy. Manders’ coefficient values estimating fractions of endomembranes colocalized with GFP-2xFYVE–decorated endosomes were elevated in MTMR 3/4/7, but not MTMR 2, KD cells, suggesting increased PI3P content in 2xFYVE-decorated endosomes (Fig. 4, A and B; and Fig. S3). Also, individual KD of the MTMR 2/3/4/7 gene elevated the total cellular PI3P level (Fig. 4 C), consistent with previous studies (Cao et al., 2008; Lahiri et al., 2015; Ma et al., 2008; Zhao et al., 2019). Our data further show some PM decoration with GFP-2xFYVE in control and MTMR KD cells with no noticeable difference between cell lines, indicative of PI3P at the PM in these cell lines (arrowheads in Fig. 4 A). To formally quantify PM PI3P content, we performed EM analysis on intact basal PM sheets prepared from MTMR 2/3/4/7 KD T47D cells expressing GFP-2xFYVE. The data reveal that silencing MTMR 2/3/4/7 significantly elevated PI3P levels in the PM compared with the control (Fig. 4 D). Intriguingly, confocal microscopy and EM imaging further show GFP-MTMR 2/3/4/7 localization to the PM (Fig. 4, E and F), supported by previous studies reporting the PM localization of other MTMR members, MTMR 5 and 9 (Kim et al., 2003; Zou et al., 2009). We showed that while MTMR 2/3/4/7 KD modestly increases the PM PI3P level by less than twofold, the KD reduces PM PI4P, PtdSer, and KRAS levels to greater extents by two- to threefold (Fig. 1, Fig. 2, and Fig. 3). Together, these data suggest that MTMR 2/3/4/7 localize to the PM, such that genetic depletion disrupts PI3P conversion to PI at the PM. This in turn modestly increases total PI3P content including at the PM and 2xFYVE-decorated endosomes, which leads to substantial inhibition of the PM localization of KRAS, PtdSer, and PI4P.

To examine whether MTMR requires its enzymatic activity for maintaining the PM localization of KRAS, PtdSer, and PI4P, we overexpressed mCherry-tagged MTMR 3 wild type (WT) or C413S, the phosphatase-inactive mutant (Walker et al., 2001), in MTMR 3 KD T47D cells stably expressing GFP-KRASG12V or GFP-LactC2, and analyzed their cellular localization by confocal microscopy and EM. Our data show that while the overexpression of WT MTMR 3 restored the PM binding of GFP-KRASG12V and GFP-LactC2, the C413S mutant did not show any changes (Fig. 5, A–D). We further assessed the PM PI4P and PI3P contents by measuring the PM binding of GFP-P4M-SidM and GFP-2xFYVE, respectively, by EM. Like KRASG12V and PtdSer, MTMR 3 WT overexpression promoted the PM binding of GFP-P4M-SidM and GFP-2xFYVE, but not the C413S mutant (Fig. 5, E and F). We repeated the experiments with MTMR 4 WT and the C407S, the phosphatase-inactive mutant (Lorenzo et al., 2006), and observed a similar trend. The overexpression of WT MTMR 4, but not the inactive mutant, reduced PM PI3P levels, and promoted the PM binding of GFP-KRASG12V, GFP-LactC2, and GFP-P4M-SidM (Fig. S4). Moreover, despite the similar expression level, the ectopic expression of WT MTMR 2/4/7 did not restore the effects of MTMR 3 KD (Fig. 5, A and G). We made a similar observation with rescuing MTMR 4 KD, in which the overexpression of WT MTMR 2/3/7 did not revert the effects of MTMR 4 KD (Fig. S4). Together, these data suggest that MTMR members require their enzymatic activity for maintaining the PM localization of KRAS, PtdSer, and PI4P, and that the enzymatic activity of each MTMR is not functionally redundant.

While there are nine catalytically active MTMR members (Wang et al., 2024), single KD of MTMR 2/3/4/7 is sufficient to elevate PI3P contents and dissociates KRAS, PtdSer, and PI4P from the PM (Figs. 1, 2, 3, and 4). One plausible explanation is that depleting single MTMR member may perturb the enzymatic activities of other MTMR members, resulting in significant disruption of PI3P metabolism. To test this notion, we first examined whether the loss of multiple MTMR members can induce additive effects in these phenotypes. MTMR 3 KD T47D cells were infected with lentiviruses expressing shRNA targeting MTMR 2, 4, or 7, but the double KD resulted in cytotoxicity (not shown). We then measured endogenous protein levels of MTMR 2/3/7 after KD of a single MTMR member. Our immunoblot data show that the protein expression of MTMR 2/3/7 was concomitantly reduced upon MTMR 2, 3, 4, or 7 KD (Fig. 6 A), suggesting the protein expression of MTMR 2/3/7 is interdependent. Moreover, while the ectopic expression of WT MTMR 3 in MTMR 3 KD cells restored the protein levels of MTMR 2 and 7, the overexpression of WT MTMR 2, 4, or 7 in MTMR 3 KD cells did not restore the endogenous MTMR 3 protein level (Fig. 6, B and C). These data suggest that the protein expression, and thereby the enzymatic activity, of MTMR 2/3/4/7 is interdependent.

Since MTMR 2/3/4/7 KD dissociates KRASG12V from the PM, we used two biological systems to evaluate effects on KRAS signal transmission. First, we analyzed signaling by the KRAS ortholog, let-60, in the invertebrate model system Caenorhabditis elegans, where constitutively active mutation of let-60 induces a quantifiable multivulva phenotype (Reiner et al., 2008). We used RNAi designed to target C. elegans MTMR orthologs and examined whether MTMR KD could suppress the multivulva phenotype associated with the expression of LET-60 G13D (n1046). Our data show that silencing mtm-3 (ortholog of MTMR 3) or mtm-6 (ortholog of MTMR 7) abrogated multivulva formation, and indeed was equipotent with two previously reported potent suppressors of LET-60 G13D signaling (heo-1 and riok-1) (Smith and Levitan, 2004; Weinberg et al., 2014) (Fig. 7 A). Secondly, we examined it in human cancer model cell lines. The proliferation of MTMR 2/3/4/7 KD T47D cells stably expressing GFP-KRASG12V, but not GFP-HRASG12V, was significantly inhibited compared with parental cells replete for each MTMR gene (Fig. 7, B and C), suggesting the growth inhibitory effects of MTMR 2/3/4/7 KD are KRAS-specific. We also examined the proliferation and RAS signaling in human pancreatic ductal adenocarcinoma (PDAC) cell lines, MIA PaCa-2 and PANC-1, which harbor KRAS with activating point mutation, G12C and G12D, respectively, and their growth depends on the oncogenic KRAS signaling (Hayes et al., 2016; Kovar et al., 2020; Rehl et al., 2023). For a control PDAC cell line, we included BxPC-3, which harbors WT KRAS, and its growth is independent of the KRAS signaling. We showed that MTMR 2/3/4/7 KD inhibited the proliferation of MIA PaCa-2 and PANC-1, but not BxPC-3 cells (Fig. 7 D). For assessing RAS signal output, we measured active phosphorylation sites of ERK1/2 and Akt for the RAS/MAPK and RAS/PI3P/Akt pathways, respectively. Our immunoblotting data show a trend of decreased phospho-ERK after MTMR KD in KRAS-dependent PDAC, except for MTMR 4 and 7 in PANC-1 cells (Fig. 7 E). To further examine the ERK activity, we measured the expression level of the transcriptional factor, c-MYC, since active ERK enhances its stability (Sears et al., 1999). Our data show that the c-MYC protein level was significantly decreased upon MTMR 2/3/4/7 KD in KRAS-dependent, but not KRAS-independent, PDAC cell lines (Fig. 7 E). Akt phosphorylation was elevated in MTMR 4 and 7 KD cells, while a tendency of reduction was observed in MTMR 2 and 3 KD cells, suggesting that MTMR 2/3/4/7 differentially regulate the PI3K/Akt signaling in an oncogenic KRAS-independent manner. We further measured the endogenous KRAS expression level since KRAS protein stability is dysregulated after the PM dissociation (Cho et al., 2012b; Garrido et al., 2020; Kovar et al., 2020). Our immunoblot data show that endogenous KRAS expression was significantly reduced upon MTMR 2/3/4/7 KD in KRAS-dependent, but not KRAS-independent, PDAC cell lines (Fig. 7 E). We observed that while the reduced c-MYC level and growth inhibition in MTMR 4 and 7 KD PANC-1 cells are consistent, ppERK levels in these cells were elevated. A similar observation has been reported, where an ERK inhibitor paradoxically increases ERK phosphorylation while simultaneously decreasing its activity in PDAC, suggesting that kinome reprogramming compensates for the inhibition of MEK-induced ERK phosphorylation, but does not affect ppERK activity (Hayes et al., 2016; Rehl et al., 2023). It is plausible that blocking PI-3 phosphatase activity of MTMR 4 or 7 in PANC-1 cells may differentially regulate ERK phosphorylation and its activity compared with other PDAC cell lines. Together, our study provides strong evidence that the MTMR phosphatases regulate KRAS signal transmission by maintaining PtdSer and KRAS PM localization and their genetic loss dissociates the PM localization of PtdSer and KRAS, but not HRAS, resulting in the inhibition of KRAS/MAPK/c-MYC signal output and instability of KRAS protein expression in KRAS-dependent PDAC.

We have identified new roles of the phosphatases, MTMR 2/3/4/7, in maintaining the PM localization of PI4P, PtdSer, and KRAS. We show that KD of MTMR 2/3/4/7 dissociates PtdSer and KRAS from the PM and also reduces the total cellular PtdSer level. Concomitantly, MTMR 2/3/4/7 KD elevates PM PI3P content, reduces PM PI4P content, and dissociates ORP5 from the ER-PM MCS. Collecting these observations together, we propose that the PM-localized MTMR phosphatases are required to help maintain PM levels of PI for generating PM PI4P by PI4KA, which in turn maintains total PtdSer levels and drives the ER-to-PM transport of PtdSer. PI is synthesized from phosphatidic acid at the cytosolic face of ER, where a fraction of PI is used for the synthesis of glycosylphosphatidylinositol-anchored proteins (Blunsom and Cockcroft, 2020), with the remaining PI being widely distributed across intracellular cytosolic membranes (Ashlin et al., 2021; Pemberton et al., 2020; Zewe et al., 2020). PI delivered from the ER directly to the PM by the LTPs, Nir2/3, and TMEM24 (Posor et al., 2022), is a crucial substrate for PM polyphosphoinositides (PPIn). Simple overexpression of Nir2/3 and TMEM24 elevates the synthesis and total amount of PM PPIn (Chang et al., 2013; Kim et al., 2015; Lees et al., 2017), suggesting that PM PI is constantly converted to PPIn and the PI supply to the PM is rate-limiting. In this context, the pathway to PI generation from PI3P by MTMR 2/3/4/7 at the PM emerges as a critical alternative supplier of PI to allow sufficient PM PI4P generation and hence PtdSer transport by the ORP5/8 machinery (Fig. 8). In addition to the PM, MTMR 2/3/4/7 localize to endosomes and promote their trafficking by regulating endosomal PI3P and PI(3,5)P₂ levels (Bonneick et al., 2005; Franklin et al., 2011; Mochizuki and Majerus, 2003; Naughtin et al., 2010; Previtali et al., 2007; Walker et al., 2001). Thus, perturbing endosomal trafficking by MTMR 2/3/4/7 KD will block the delivery of endosomal PI to the PM, further contributing to the reduced PM PI content. Therefore, when MTMR 2/3/4/7 are depleted, the PM PtdSer level falls and KRAS dissociates from the PM (Fig. 8). We did not investigate the mechanism whereby MTMR 2/3/4/7 KD reduced total cellular PtdSer content, but previous work has shown that PM PI4P depletion indirectly blocks PtdSer synthase 1 and 2 activities (Sohn et al., 2016). Since PM PI4P levels were reduced in MTMR 2/3/4/7 KD cells, the same mechanism is likely responsible. Furthermore, inactivating mutations in MTMR 2 are found in Charcot–Marie–Tooth disease type 4B1, a severe autosomal recessive neuropathy with demyelination and myelin outfoldings of the nerve. Studies have proposed that disrupted endosomal trafficking by elevating endosomal PI3P and PI(3,5)P₂ due to inactive MTMR 2 likely contributes to the demyelination of nerve cells (Bonneick et al., 2005; Previtali et al., 2007). Intriguingly, PtdSer is the main component of myelin sheath (Ma et al., 2022), and inhibition of PI4KA in Schwann cells reduces PM and total levels of PtdSer and PI4P, and induces aberrant myelination (Alvarez-Prats et al., 2018). Thus, in addition to disrupted endosomal trafficking by MTMR 2 inactivation, it is plausible that the reduced PM levels of PtdSer and PI4P further contribute to the demyelination observed in Charcot–Marie–Tooth disease type 4B1.

We have demonstrated that the protein expression of MTMR 2/3/4/7 is interdependent, whence single KD of MTMR 2, 3, 4, or 7 concomitantly reduces the protein levels of MTMR 2/3/4/7 (Fig. 6 A), which likely accounts for the observation that a loss of MTMR 2, 3, 4, or 7 is sufficient to elevate cellular PI3P level despite the presence of nine catalytically active MTMR members (Cao et al., 2008; Lahiri et al., 2015; Ma et al., 2008; Zhao et al., 2019). Also, while the overexpression of WT MTMR 3 in MTMR 3 KD cells restores endogenous protein levels of MTMR 2/4/7, the overexpression of WT MTMR 2/4/7 does not restore the endogenous MTMR 3 protein level (Fig. 6, B and C). These data are consistent with our rescue experiments, in which the overexpression of WT MTMR 3, but not the other MTMR members, in MTMR 3 KD cells restores the KD effects of MTMR 3. For the interdependence of MTMR 2/3/4/7 protein expression, one plausible model is that the catalytically active and inactive MTMR members form one large, interconnected machinery, which may regulate the subcellular localization of MTMR members for maintaining cellular PI3P metabolism. The nine enzymatically active MTMR members need to form homo- or heterodimers with the seven inactive members via its coiled-coil domain for their enzymatic activity and subcellular localization (Hnia et al., 2012; Wang et al., 2024). MTMR 3 and 4 form homodimers and heterodimers with each other (Lorenzo et al., 2006). MTMR 2 forms homodimers and heterodimers with inactive members, MTMR 5, 12, and 13 (Kim et al., 2003; Nandurkar et al., 2003; Robinson and Dixon, 2005), while MTMR 7 heterodimerizes with the inactive MTMR 9 (Mochizuki and Majerus, 2003). Also, MTMR 2 promotes MTMR 13 membrane localization and they depend on each other for maintaining their protein expression (Ng et al., 2013; Robinson and Dixon, 2005). Thus, the loss of a single MTMR member could perturb the subcellular localization of its partner MTMR members, resulting in their protein degradation. This in turn disrupts the interconnected network of MTMR members, resulting in the subcellular mislocalization and degradation of multiple MTMR members, and thereby perturbed PI3P metabolism. A systematic analysis of subcellular localization and protein expression of catalytically active and inactive MTMR members after genetic depletion of a single MTMR member will provide deeper insights into the interdependence of MTMR members.

We have observed that MTMR 2/3/4/7 KD reduces oncogenic KRAS protein expression (Fig. 7 E). Previous studies have reported that MTMR KD stimulates autophagy via elevating PI3P content in the phagophore, a double-membrane structure that matures into an autophagosome, which recruits autophagy-related proteins for autophagosome biogenesis (Nishimura and Tooze, 2020; Wang et al., 2024). Also, KRAS degradation occurs via the autophagy/lysosome-mediated mechanism, in which ADAM9 loss or inhibition enhances the interaction between KRAS and plasminogen activator inhibitor 1 (PAI-1), which acts as a specific autophagy receptor binding to light chain 3 (LC3) for stimulating lysosomal degradation of KRAS (Huang et al., 2024). Together, our data suggest that the PM-dissociated KRAS undergoes its degradation by enhanced autophagy induced by MTMR KD. However, this may only partially account for our observation. Since MTMR 2/3/4/7 KD induces PM dissociation of KRAS without the G-domain (Fig. S2), it is likely that the KD induces the PM dissociation of WT KRAS expressed in BxPC-3, which was not degraded upon MTMR 2/3/4/7 KD (Fig. 7 E). Thus, the KRAS degradation upon MTMR 2/3/4/7 KD is unlikely induced by KRAS PM dissociation. Also, while oncogenic KRAS/MAPK signaling promotes autophagy in PDAC (Bryant et al., 2019), we observed MTMR 2/3/4/7 KD reduces the MAPK/c-MYC signaling in PDAC harboring oncogenic KRAS. Thus, the inhibition of MAPK/c-MYC signaling by MTMR KD would block autophagy, but promote it. Deeper insights for KRAS degradation in terms of oncogenic KRAS signaling and cellular PI3P content need to be elucidated.

In sum, our study proposes a new role of the MTMR phosphatases in maintaining PM PI4P and PtdSer contents by maintaining an adequate supply of PM PI from PI3P. Depleting MTMR 2/3/4/7 in turn reduces PM and total cellular PtdSer levels, blocking KRAS PM localization and KRAS signaling. Thus, the mechanisms that maintain PM PI content may contain useful targets to abrogate oncogenic KRAS signaling. Our study further proposes an interdependent network of MTMR members for maintaining their protein expression and their phosphatase activity.

shRNA glycerol sets for MTMR 2 (RHS4533-EG8898), MTMR 3 (RHS4533-EG8897), MTMR 4 (RHS4533-EG9110), MTMR 7 (RHS4533-EG9108), and human cDNA of MTMR 2 (MHS6278-202760229), MTMR 3 (MHS6278-202800229), MTMR 4 (MHS6278-202801286), and MTMR 7 (MHS6278-211688199) were purchased from GE Healthcare Dharmacon. Antibodies were purchased for detecting MTMR 2 (PA5-22748; 1:1,000) and c-MYC (13-2500, 1:1,000) from Invitrogen, MTMR 3 (sc-393779; 1:1,000) from SCBT, and MTMR 7 (AB150458; 1:1,000) from Abcam. Antibodies for detecting c-MYC (10828-1-AP; 1:3,000), β-actin (60008-1-Ig; 1:5,000), MTMR 3 (21336-1-AP; 1:1,000), and GFP (66002-1-lg; 1:4,000) were purchased from Proteintech. Anti-ppERK (4370; 1:3,000), anti-pAkt (S473) (4060; 1:1,000), anti-total ERK (4696; 1:1,000), and anti-total Akt (2920; 1:1,000) antibodies for immunoblotting were from Cell Signaling Technology. Anti-RFP antibody (sc-390909; 1:1,000) was from Santa Cruz. Goat anti-mouse IgG (G21040; 1:2,000), anti-rabbit IgG secondary antibodies (G21234; 1:5,000), and CellMask Deep Red PM stain (C10046; 1:5,000) were purchased from Invitrogen. Staurosporine (BIA-S1086) was purchased from BioAustralis.

A whole-genome siRNA screen was performed in Caco-2 cells stably expressing GFP-KRASG12V using a pre-aliquoted Silencer siRNA library (Ambion) at 30 nM final concentration. A total of 21,687 genes were represented in the library and were arrayed into glass-bottomed sixty-nine 384-well plates in triplicate. Gene silencing was induced for 4 days via reverse transfection using DharmaFECT 1 (GE Healthcare). For controls, each plate included nontargeting control siRNA and siRNA targeting FNTA (farnesyltransferase type 1 subunit alpha) as negative and positive controls, respectively. Also, each plate included the Kif11 siRNA that targets a gene essential for cell survival served as a control for transfection and KD efficiency. After gene silencing, cells were fixed with 4% paraformaldehyde (PFA) and stored in 1x phosphate-buffered saline (PBS) at 4°C until imaged. Plates were imaged using a Nikon A1R confocal microscope. Images were acquired as a 2-by-1 montage with a 20× objective lens using the GFP-confocal mode with NIS-Elements automated imaging software. Images were scored for KRAS PM dissociation/intracellular accumulation of GFP-KRASG12V and cytotoxicity. 253 genes were selected from the primary screen based on KRAS PM dissociation, and after a counter-screen with GFP-HRASG12V to identify KRAS-specific mediators, we refined the candidate list to 157. The 157 genes were further validated with an independent secondary screen involving silencing each gene with Dharmacon SMARTpool ON-TARGET plus siRNA in Caco-2 cells expressing GFP-KRASG12V and scoring for KRAS PM dissociation. A further bioinformatics analysis revealed that eight of these genes encode proteins with phosphatase activity (Table 1).

T47D (HTB-13; ATCC) and BxPC3 (CRL-1687; ATCC) cells were maintained in RPMI-1640 medium (30-2001; ATCC) supplemented with 2 mM L-glutamine (CA009-010; GenDEPOT) and 10% fetal bovine serum (FBS) (16000044; Gibco). For T47D cells, 10 μg/ml human recombinant insulin (12585-014; Gibco) was also supplemented. MIA PaCa-2 (CRL-1420; ATCC), PANC-1 (CRL-1469; ATCC), and Caco-2 (HTB-37; ATCC) cells were maintained in DMEM with high glucose (CM002; Gibco) supplemented with 2 mM L-glutamine and 10% FBS. All cells were grown at 37°C with 5% CO₂ and frequently tested for Mycoplasma using MycoAlert Mycoplasma Detection Kit (LT07-318; Lonza).

T47D cells were transfected with GFP-KRASG12V, GFP-HRASG12V, or GFP-LactC2 in a pEF6 vector (Invitrogen) using TransIT-BrCa Transfection Reagent (MIR 5500; MiRus), and selected with 10 μg/ml blasticidin for 3 wk. After the selection, cells were maintained in 5 μg/ml blasticidin. For generating MTMR KD cell lines, WT T47D cells were infected with lentiviruses expressing shRNA targeting MTMR 2/3/4/7. Cells were selected by puromycin (1 μg/ml) for 1 wk and maintained in 0.5 μg/ml puromycin.

5 × 10⁵ T47D cells were seeded on a coverslip on day 1 in RPMI-1640 medium containing 10% FBS, 2 mM L-glutamine, and 10 μg/ml insulin (complete growth medium). On day 2, fresh complete growth medium was supplemented. On day 3, cells were fixed with 4% PFA for 30 min, quenched with 50 mM NH₄Cl for 10 min, and mounted with Vectashield (H-1000; Vector Laboratories). For calculating Manders’ coefficient, cells were incubated with CellMask stain (1:5,000) for 1 h in complete growth medium, fixed with 4% PFA, and imaged by confocal microscopy. Using ImageJ software 1.53k, images were converted to 8-bit images, and a threshold to a control pixel of each image was set. The fraction of red fluorophore–conjugated CellMask colocalizing with GFP-tagged proteins was calculated using Manders’ coefficient plugin downloaded from Wright Cell Image Facility.

T47D cells stably expressing GFP-KRASG12V with MTMR 2/3/4/7 KD were cultured in complete growth medium containing 0.5 µg/ml puromycin. Triplicate samples of 1 × 10⁶ cells were prepared in 333 μl Dulbecco’s PBS (without Ca²⁺ and Mg²⁺, 14190144; Invitrogen). Lipid extraction and analysis using electron spray ionization and MS/MS were performed at Lipotype GmbH, as described previously (Gerl et al., 2012; Sampaio et al., 2011). Automated processing of acquired mass spectra, identification, and quantification of detected lipid species were done by LipidXplorer software. Only lipid identifications with a signal-to-noise ratio >5, an absolute abundance of at least 1 pmol, and a signal intensity fivefold higher than in corresponding blank samples were considered for further data analysis. The abundance of lipids is represented as a heat map with a log₂ scale relative to control (scrambled shRNA) cells.

The total amount of PI(3)P was examined in a quantitative and competitive ELISA format assay according to the manufacturer’s instructions (K-3300; Echelon Biosciences). Briefly, MTMR 2/3/4/KD T47D cells grown on T75 flasks were collected in ice-cold 0.5 M trichloroacetic acid (TCA) and centrifuged at 1,000 × g for 10 min at 4°C. The pellet was washed with 3 ml 5% TCA/1 mM EDTA, and neutral lipids were removed using MeOH:CHCl₃ (2:1), followed by acidic lipid extraction using MeOH:CHCl₃:12N HCl (80:40:1). The acidic lipid extracts were further separated into organic and aqueous phases, and the organic phase was dried in a vacuum dryer. The dried lipid was rehydrated in PBS-T 3% protein stabilizer and sonicated in a water bath. The sample was then used to measure the PI3P level using the mass ELISA kit. Absorbance was read at 450 nm on Synergy H1 Plate Reader (BioTek).

Preparation of cell lysates and immunoblotting were performed as described previously (Cho et al., 2011). Briefly, cells were washed twice with ice-cold 1x PBS. Cells were harvested in lysis buffer B containing 50 mM Tris-Cl (pH 7.5), 75 mM NaCl, 25 mM NaF, 5 mM MgCl₂, 5 mM EGTA, 1 mM DTT, 100 μM NaVO₄, 1% NP-40, plus protease and phosphatase inhibitors. SDS-PAGE and immunoblotting were generally performed using 20–40 μg of lysate from each sample group. Signals were detected by enhanced chemiluminescence (34578 and 34075; Thermo Fisher Scientific) and imaged using Amersham Imager 600 (GE Healthcare). ImageJ software (v1.53k) was used to quantify band intensity.

Messenger RNA (mRNA) was extracted from T47D cells infected with lentiviruses expressing shRNA targeting MTMR 4 using the RNeasy kit (74104; Qiagen) according to the manufacturer’s instructions. 2 μg of total RNA was converted to cDNA using Superscript II reverse transcriptase (18064-014; Thermo Fisher Scientific). To verify KD, forward and reverse primers specific for human MTMR 4 exons 1 and 5, and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) exons 1 and 3 were designed: for MTMR 4, 5′-CCA​AGC​CAA​GGA​TCT​GTT​CCC-3′ and 5′-TGT​GTG​AGA​CTC​TCC​AGA​CGT-3′, and for GAPDH, 5′-GGA​GCG​AGA​TCC​CTC​CAA​AAT-3′ and 5′-GGC​TGT​TGT​CAT​ACT​TCT​CAT​GG-3′, respectively. cDNA was amplified by PCR (P2311; GenDEPOT), and the products were resolved by 1.5% agarose gel electrophoresis and visualized by ethidium bromide staining.

let-60 (n1046) worms were kindly provided by Swathi Arur (MD Anderson Cancer Center, Houston, TX, USA). RNAi was induced by feeding let-60 worms through the adult stage with E. coli HT115, producing double-stranded RNA to target genes. The presence of the multivulva phenotype was scored using a DIC/Nomarski microscope. All RNAi clones were from the C. elegans RNAi (Ahringer) collection (Source BioScience) and were sequenced.

MTMR 2/3/4/7 KD T47D cells stably expressing GFP-KRASG12V or GFP-HRASG12V (1 × 10⁴ per well), or human PDAC cell lines, BxPC3, MIA PaCa-2, and PANC-1 (3 × 10⁴ per well), were seeded in triplicate onto a 96-well plate in complete growth medium containing 1.0 μg/ml puromycin. Cells were cultured for 4 days, and fresh complete growth medium containing 1.0 μg/ml puromycin was supplemented every 24 h. On day 5, cell proliferation was assayed by counting cells using a Countess II cell counter (Invitrogen).

MTMR 2/3/4/7 KD T47D cells stably expressing GFP-KRASG12V were incubated with Cy5-conjugated annexin V (559934; BD Pharmingen) according to the manufacturer’s instructions. Annexin V–positive cells were counted using BD Accuri C6 Analyzer with a 670-nm LP filter. GFP-KRASG12V T47D cells expressing scrambled shRNA were treated with 1 μM staurosporine for 6 h to induce apoptosis.

PM sheets were prepared from Caco-2 or T47D cells expressing GFP-tagged proteins of interest and fixed as previously described (Hancock and Prior, 2005; Prior et al., 2003a, 2003b). For univariate analysis, PM sheets were labeled with anti-GFP antibody conjugated to 4.5-nm gold particles. Digital images of the immunogold-labeled PM sheets were taken in a transmission electron microscope. Intact 1-μm² areas of the PM sheet were identified using ImageJ software (v1.53k), and the (x, y) coordinates of the gold particles were determined (Hancock and Prior, 2005; Prior et al., 2003a, 2003b). K-functions (Ripley, 1977) were calculated and standardized on the 99% confidence interval (CI) for univariate functions (Diggle et al., 2000; Hancock and Prior, 2005; Prior et al., 2003a, 2003b). In the case of univariate functions, a value of L(r) – r greater than the CI indicates significant clustering, and the maximum value of the function (L_max) estimates the extent of clustering. Differences between replicated point patterns were analyzed by constructing bootstrap tests as described previously (Diggle et al., 2000; Plowman and Hancock, 2005), and the statistical significance against the results obtained with 1,000 bootstrap samples was evaluated.

The following settings were used for confocal microscope imaging acquisition: (1) Make and model of microscope: Olympus Fluoview F1000; (2) type, magnification, and numerical aperture of the objective lenses: Olympus UPlanSApo 60×/1.35 oil, ∞/0.17/FN26.5 UIS2 BFP1; (3) temperature: room temperature; (4) imaging medium: Vectashield (H-1000; Vector Laboratories); (5) fluorochromes: GFP and mCherry; (6) Camera make and model: N/A; (7) acquisition software: FV10-ASW ver.4.2c; (8) any software used for imaging processing subsequent to data acquisition: none.

Fig. S1 shows MTMR 2/3/4/7 regulate the PM localization of KRASG12V. Fig. S2 shows MTMR 2/3/4/7 KD disrupts the PM localization and nanoclustering of CTK. Fig S3 shows MTMR 2/3/4/7 regulate cellular PI3P contents. Fig. S4 shows the ectopic expression of WT MTMR 4 restores the effects of MTMR 4 loss.

We would like to thank Karen Henkels for her contribution to this work.

This work was supported by the Wright State University Seed Grant and the National Cancer Institute (R00-CA188593 to K.-j. Cho), National Institute of General Medical Sciences (2R35GM119412 to G.R.V. Hammond), and Cancer Prevention and Research Institute of Texas (CPRIT RP200047 to J.F. Hancock).

Author contributions: T.E. Lange: conceptualization, formal analysis, investigation, and writing—review and editing. A. Naji: investigation. R. van der Hoeven: investigation, methodology, resources, validation, and writing—review and editing. H. Liang: investigation. Y. Zhou: data curation, formal analysis, investigation, and validation. G.R.V. Hammond: methodology and writing—review and editing. J.F. Hancock: conceptualization, funding acquisition, project administration, and writing—review and editing. K.-j. Cho: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, supervision, validation, visualization, and writing—original draft, review, and editing.