3112 Research Article

IntroductionThe mucolipins (TRPML) are a subfamily of channel proteins withinthe transient receptor potential (TRP) superfamily of cation channels(Venkatachalam and Montell, 2007) and the mucolipins 1, 2 and 3(TRPML1, TRPML2 and TRPML3, respectively) each featurepredicted cation channel pores between their transmembranesegments 5 (TM5) and 6 (TM6) (Puertollano and Kiselyov, 2009).Loss-of-function mutations in TRPML1 are implicated inmucolipidosis type IV (MLIV; MIM# 252650) pathogenesis, whereasa spontaneous gain-of-function mutation in TRPML3 is principallyimplicated in the varitint-waddler mouse mutant (Cuajungco andSamie, 2008; Zeevi et al., 2007). The varitint-waddler mouse featureshearing loss, vestibular dysfunction, and abnormal pigmentation (DiPalma et al., 2002). On the cellular level, the A419P gain-of-functionmutation in TRPML3 causes Ca2+ overload, which leads to celldeath in mouse inner ear cells and melanosomes (Grimm et al.,2007; Grimm et al., 2009; Xu et al., 2007). MLIV is an autosomalrecessive neurodegenerative lysosomal storage disorder characterizedby mental and motor retardation, hypotonia, and opthalmologicalabnormalities (Bach, 2005). Loss-of-function mutations in TRPML1lead to heterogeneous lysosomal storage of hydrophobic andhydrophilic macromolecules in cells of MLIV patients and otheranimal models (Bach, 2001; Venkatachalam et al., 2008; Venugopalet al., 2007). At present, TRPML2 is not implicated in anypathological disorder; however, we have recently described lysosomalinclusions in HEK 293 cells featuring significantly reduced nativeTRPML2 expression (Zeevi et al., 2009).

TRPML channels have been implicated in various membrane andprotein sorting mechanisms along endo-lysosomal pathways(Puertollano and Kiselyov, 2009). MLIV pathogenesis is directlylinked to TRPML1 loss-of-function within the late endosomal/lysosomal compartments where the protein is generally found(Manzoni et al., 2004; Pryor et al., 2006; Vergarajauregui andPuertollano, 2006) and, hence, several lysosomal regulatory roleshave been proposed for TRPML1 in these compartments (Dong etal., 2008; Dong et al., 2009; LaPlante et al., 2006; Miedel et al.,2008; Piper and Luzio, 2004; Pryor et al., 2006; Soyombo et al.,2006; Treusch et al., 2004; Venugopal et al., 2009; Vergarajaureguiet al., 2008). TRPML2 features cell surface and recycling endosomallocalization and, as a result, has been implicated in recycling ofglycosylphosphatidylinositol (GPI)-anchored proteins from recyclingendosomes to the cell surface along the Arf6-regulated pathway(Karacsonyi et al., 2007). TRPML3 appears to play roles in theregulation of endocytosis and autophagy that correspond withlocalization of the protein to the cell surface and early endosomesunder basal conditions, and autophagosome localization under stressconditions (Kim et al., 2009; Martina et al., 2009). Taken together,the differing intracellular distributions and proposed protein functionsof each respective TRPML might suggest that paralogous TRPMLsare unlikely to co-regulate particular celluar tasks. However, recentevidence argues otherwise.

Although TRPMLs have shown to be distinct from each otherin their primary intracellular distributions (Karacsonyi et al., 2007;Kim et al., 2009; Martina et al., 2009; Pryor et al., 2006;

Heteromultimeric TRPML channel assemblies play acrucial role in the regulation of cell viability modelsand starvation-induced autophagyDavid A. Zeevi1,*, Shaya Lev2,3,*, Ayala Frumkin1, Baruch Minke2,3 and Gideon Bach1,‡

1Monique and Jacques Roboh Department of Genetic Research, 2Department of Medical Neurobiology and 3The Kühne Minerva Center forStudies of Visual Transduction, Institute of Medical Research Israel-Canada (IMRIC), Faculty of Medicine of the Hebrew University and HadassahHebrew University Hospital, Jerusalem 91120, Israel*These authors contributed equally to this work‡Author for correspondence (bach@hadassah.org.il)

Accepted 5 June 2010Journal of Cell Science 123, 3112-3124 © 2010. Published by The Company of Biologists Ltddoi:10.1242/jcs.067330

SummaryThe mucolipin (TRPML) subfamily of transient receptor potential (TRP) cation channels consists of three members that play variousroles in the regulation of membrane and protein sorting along endo-lysosomal pathways. Loss-of-function mutations in TRPML1 causethe neurodegenerative lysosomal storage disorder, mucolipidosis type IV (MLIV), whereas a gain-of-function mutation in TRPML3 isprincipally implicated in the hearing-impaired and abnormally pigmented varitint-waddler mouse. Currently, TRPML2 is not implicatedin any pathological disorder, but we have recently shown that it is a functional cation channel that physically interacts with TRPML1and TRPML3 to potentially regulate lysosomal integrity. Here, we show that mutant TRPMLs heteromultimerize with other mutantand wild-type TRPMLs to regulate cell viability and starvation-induced autophagy, a process that mediates macromolecular andorganellar turnover under cell starvation conditions. Heteromultimerization of dominant-negative TRPMLs with constitutively activeTRPMLs rescues cells from the cytotoxic effects of TRPML constitutive activity. Moreover, dominant-negative TRPML1 channels,including a mutant channel directly implicated in MLIV pathology, also inhibit starvation-induced autophagy by interacting with andaffecting native TRPML channel function. Collectively, our results indicate that heteromultimerization of TRPML channels plays arole in various TRPML-regulated mechanisms.

Key words: Mucolipin, TRPML, Ion channel heteromultimerization, TRP channels, Starvation-induced autophagy

Jour

nal o

f Cel

l Sci

ence

Vergarajauregui and Puertollano, 2006), all three channels do alsocolocalize with each other, in dual combinations, in a subset ofintracellular vesicles (Zeevi et al., 2009). Furthermore, recombinantas well as native TRPML channels also physically interact witheach other in homo- and heteromultimeric combinations (Curcio-Morelli et al., 2010; Venkatachalam et al., 2006; Zeevi et al.,2009). In addition, TRPML2 and TRPML3 traffic, in part, tolysosomes (Karacsonyi et al., 2007; Kim et al., 2009; Martina etal., 2009; Zeevi et al., 2009) where, like TRPML1, they alsoappear to regulate lysosomal function, in an as-yet-unknownmanner. Indeed, gene-specific knockdown of TRPML2 or TRPML3leads to lysosomal inclusions reminiscent of those found in MLIVpatient TRPML1-null cells (Zeevi et al., 2009), suggesting thatTRPMLs might dynamically interact with each other to sustainlysosomal integrity. Other discoveries in the TRPML field alsopoint towards commonalities in TRPML function. Acidic pHappears to regulate the ion current properties of all three TRPMLs(Cantiello et al., 2005; Dong et al., 2008; Grimm et al., 2007; Kimet al., 2008; Lev et al., 2010; Raychowdhury et al., 2004; Xu et al.,2007), and paralogous missense mutations, within TRPML putativeion channel pores, regulate the activity of each TRPML channel inthe same manner (Dong et al., 2008; Grimm et al., 2007; Lev etal., 2010; Xu et al., 2007). Recently, TRPML1 was also shown toregulate the gene expression of TRPML2 (Samie et al., 2009), andrecombinant TRPML1 was shown to reduce the surface expressionof recombinant TRPML3 (Grimm et al., 2010). Together, thesefindings suggest a more concerted role for TRPMLs in cellregulatory events.

Although, at present, the precise mechanism by which TRPMLsregulate each other’s function to affect cell biological processes isunclear, recent evidence suggests that TRPMLs do so byheteromultimerizing to form distinct cation channel assemblies(Curcio-Morelli et al., 2010). Functional TRPML1–TRPML2 andTRPML1–TRPML3 channel assemblies have been characterizedin a synthetic lipid-bilayer system by single-channel analysis(Curcio-Morelli et al., 2010), but a more comprehensivephysiological interrogation of these and other TRPML channelassemblies has still to be carried out. Here, we assayed TRPMLchannel assemblies, in every possible homo- and heteromericcombination, in the physiological context of native TRPML-expressing tissue-culture cells. Furthermore, in order to demonstratethe direct application of these channel assemblies upon cell biologyand homeostasis, we examined the affect of heterologous–heterologous and heterologous–native TRPML subfamilyinteractions upon cellular function. We show that dominant-negativeloss-of-function TRPML mutants physically interact withconstitutively active TRPMLs to form functional complexes. Thesehomo- and heteromultimeric channel complexes feature reducedactivity, which affects cell viability. Finally, we show thatheteromultimerization of TRPML channels can play a role instarvation-induced autophagy, which suggests that this process isregulated, at least in part, by a concerted effort of more than justone TRPML channel.

ResultsAll TRPMLs feature similar ion current regulatory elementsthat affect cell viability in a similar mannerWhen comparing the amino acid sequence of the putative channelpore region (ranging from TM5 to TM6) of each respective TRPMLchannel with that of its other subfamily members, a strikingsimilarity is observed (Fig. 1A). The amino acid identity and

3113Multimeric TRPML channel assemblies

similarity of these channels in this region is ~82% and ~89%,respectively. This suggests that all three TRPMLs feature a highlysimilar ion channel pore, at least with regard to structure, implyingthat missense mutations within this region might give rise to similareffects on the channel activity of each TRPML. Indeed, recentstudies have shown that paralogous mutations in this region havesimilar effects on TRPML channel function, regardless of whichparticular TRPML is mutated. The A419P mutation in TRPML3,which is implicated in varitint-waddler pathogenesis (Di Palma etal., 2002), is a gain-of-function mutation that leads to robust,inwardly rectifying currents through Ca2+-permeable andconstitutively active TRPML3 channels (Grimm et al., 2007; Kimet al., 2007; Lev et al., 2010; Nagata et al., 2008; Xu et al., 2007).Likewise, we and others have recently shown that paralogousmutations in TRPML2 and TRPML1 lead to a similar effect on thechannel function of the respective proteins (Dong et al., 2008; Levet al., 2010; Samie et al., 2009; Xu et al., 2007). On the flip-side,the DD458/459KK mutation in TRPML3 is a loss-of-functionmutation that renders the channel inactive (Kim et al., 2009; Xu etal., 2007), whereas the paralogous mutation in TRPML2 alsoinactivates the channel (Lev et al., 2010). To date, theelectrophysiological properties of the paralogous TRPML1 mutanthave not been characterized; however, a cell-based assay with thismutant protein suggests that it lacks channel function as well(Pryor et al., 2006). Thus, all three TRPMLs can be activated andinactivated by the same paralogous mutations, which lends supportto the notion that all three channels feature similar current regulatoryelements.

Importantly, the TRPML activating and inactivating mutationsdescribed above have also been shown to play a role in cellularfunction and homeostasis. The inwardly rectifying, Ca2+-permeable,and constitutively active TRPML3 A419P [TRPML3(P); see Table1] mutant channel causes intracellular Ca2+ overload, whichultimately leads to cell death in TRPML3(P)-expressing cells(Grimm et al., 2007; Kim et al., 2007; Nagata et al., 2008; Xu etal., 2007). This Ca2+-mediated cytotoxic effect of TRPML3(P)mutant channels is reduced, in part, by coexpression with thecalcium extrusion pump, plasma membrane calcium ATPase type2 (PMCA2) (Grimm et al., 2009) (Fig. 1B). Likewise, we haverecently demonstrated a similar effect upon cell viability whenexpressing the paralogous TRPML2 A424P [TRPML2(P); seeTable 1] mutant in the presence and absence of PMCA2 (Lev etal., 2010) (Fig. 1C). Here, we show that the paralogous TRPML1V432P [TRPML1(P); see Table 1] mutant presents with cytotoxiceffects that can also be partially rescued when the channel iscoexpressed with PMCA2 (Fig. 1D). This indicates, for the firsttime, that cell death in TRPML1(P)-expressing cells is also causedby intracellular Ca2+ overload. Interestingly, the TRPML channel-inactivating mutations, described above, have shown to be evenmore effective at neutralizing the cytotoxic effects of TRPML(P)

Table 1. Abbreviated notation of TRPML mutants describedin this study

TRPML mutant Notation

TRPML3 A419P TRPML3(P)TRPML2 A424P TRPML2(P)TRPML1 V432P TRPML1(P)TRPML3 A419P-DD458/459KK TRPML3(P-KK)TRPML2 A424P-DD463/464KK TRPML2(P-KK)TRPML1 V432P-DD471/472KK TRPML1(P-KK)

Jour

nal o

f Cel

l Sci

ence

mutants than coexpression with PMCA2 (Fig. 1B–D). A previousstudy demonstrated that introduction of the TRPML3 channel-inactivating DD458/459KK mutation in cis with the activatingA419P mutation [TRPML3(P-KK); see Table 1] leads to a fullyinactive channel with reduced cytotoxic effects in melan-a2 cells(Xu et al., 2007). Consistent with this study, we show that, underour assay conditions in HeLa cells, the TRPML3(P-KK) mutantpresents with virtually nonexistent cytotoxic effects (Fig. 1B).Moreover, the same dominant-negative effect of the KK mutationover the P mutation is also observed with paralogous TRPML2(P-KK) and TRPML1(P-KK) mutants (Fig. 1C,D; Table 1). Therefore,we conclude that paralogous current-modulating mutations in allthree TRPMLs affect cell viability in a similar manner.

TRPMLs with mutations in their putative ion channel poresstill physically interact with each other in homo- andheteromultimeric combinationsHaving established the cytotoxic effects of TRPML(P) expressionand the non-cytotoxic effects of TRPML(P-KK) expression, weutilized our constructs as tools by which to assay the functionalityof homo- and heteromultimeric TRPML subfamily channelassemblies. Recent studies have demonstrated that the KK

3114 Journal of Cell Science 123 (18)

mutations in TRPML2 and TRPML3 exert a dominant-negativeinhibitory effect on the function of their respective native channels.This effect was attributed to the interactions of TRPML2-KK andTRPML3-KK with and neutralization of the activity of theircorresponding native channel (Karacsonyi et al., 2007; Kim et al.,2009). In a similar manner, we hypothesized that if the TRPML(P)and TRPML(P-KK) mutants were to form channel assemblies witheach other, regardless of channel identity, then the channel-inactiveTRPML(P-KK) mutants should exert a dominant-negative effecton constitutively active TRPML(P) mutants by rendering thecomplex inactive.

Because this hypothesis is predicated upon the ability of allTRPML(P) channels to physically interact with each TRPML(P-KK) channel in functional complexes, we performed co-immunoprecipitation (co-IP) assays to ensure that the channel poremutations (P and KK) do not interfere with previously describedTRPML subfamily protein–protein interactions (Venkatachalam etal., 2006; Zeevi et al., 2009). In supplementary material Fig. S1,we show the physical interactions of TRPML mutants from thesame protein of origin [for example, TRPML1(P) vs. TRPML1(P-KK)] (supplementary material Fig. S1A). Indeed, the introductionof P and KK mutations into all three TRPML channels bears no

Fig. 1. Paralogous current-modulating mutations in all three TRPML channels affect cell viability in a similar manner. (A)Amino acid sequence alignmentof the TM5–TM6 regions of TRPML1, TRPML2, and TRPML3. Residues are shaded by similarity from low (white) to high (black). The locations of the P and KKmutations (see Table 1) are indicated. (B)HeLa cells were co-transfected with tdTomato-TRPML3(P) + GFP/YFP or GFP-TRPML3(P-KK) + tdTomato, with orwithout PMCA2. Alexa-Fluor-647–AnnexinV staining was performed in order to detect early apoptotic cells. The histogram indicates the mean percentage of cellsexhibiting both GFP/YFP and tdTomato fluorescence, under each condition, that also stained positively for Alexa-Fluor-647–AnnexinV. (C)Same as B, withTRPML2(P)-tdTomato or TRPML2(P-KK)-YFP co-transfected cells. (D)Same as B, with tdTomato-TRPML1(P) or GFP-TRPML1(P-KK) co-transfected cells.For B–D, n3 independent experiments of at least 42 cells per assay. Indicated P values are relative to the left-most bar in each histogram: *P<0.05, **P<0.01.

Jour

nal o

f Cel

l Sci

ence

effect upon such homomeric physical interactions. In supplementarymaterial Fig. S2, we show the physical interactions of TRPMLmutants from differing proteins of origin. Indeed, heteromericinteractions between TRPML(P) and TRPML(P-KK) mutants fromdiffering proteins of origin [for example, TRPML3(P) andTRPML2(P-KK)] are also unaffected by the P and KK mutations(supplementary material Fig. S2A). These heteromeric TRPMLchannel interactions appear to be TRPML subfamily-specificbecause none of the TRPML channels co-immunoprecipitated withTRPM8, a related channel from a different TRP channel subfamily(supplementary material Fig. S2B). Thus, we demonstrate thatTRPMLs, selectively, homo- and heteromultimerize with eachother even when possessing mutations in their putative ion channelpores.

Dominant-negative TRPMLs attenuate the cytotoxic effectsof constitutively active TRPML channelsHaving established that TRPML(P) and TRPML(P-KK) channelsphysically interact with each other in homo- and heteromultimericcombinations, we proceeded to assay the effect of theseinteractions upon cellular function. In particular, we sought toclarify the importance of heteromultimeric TRPML subfamilyinteractions. Thus, based on the cytotoxic effects of TRPML(P)expression and the non-cytotoxic effects of TRPML(P-KK)expression, as described above (Fig. 1B–D), we established cellviability as the indicator in our experiments. If the non-toxicTRPML(P-KK) mutants do not exert a dominant-negative effecton the toxic TRPML(P) mutants, then when the proteinsheteromultimerize, we would not expect to observe a significantrescue of TRPML(P) and TRPML(P-KK) coexpressing cells fromTRPML(P)-induced apoptosis. However, if TRPML(P-KK)s doexert a dominant-negative effect over TRPML(P)s, then wewould expect to observe the opposite when the proteinsheteromultimerize.

Coexpression of TRPML3(P) with TRPML2(P-KK)significantly reduced cell death, with respect to TRPML3(P) andGFP/YFP coexpressing control cells, from 84.7±3.2 to 26.8±2.5%AnnexinV-positive (early apoptotic) cells (Fig. 2A). Thisobservation suggests that TRPML2(P-KK) channels exert adominant-negative effect over TRPML3(P) channels. In addition,we found that TRPML3(P)-induced apoptosis is further reducedwhen coexpressing the channel with the dominant-negativeTRPML2(P-KK) channel and the calcium extrusion pump,PMCA2 (Fig. 2A), indicating that the dominant-negative effectof TRPML2(P-KK) stems from reducing the cytotoxicity ofTRPML3(P)-induced Ca2+ overload. Moreover, we also observedsimilar cytotoxic rescue in cells coexpressing TRPML3(P) withTRPML1(P-KK) and in cells coexpressing TRPML3(P) withTRPML3(P-KK) (Fig. 2A), which suggests that any TRPML(P-KK) mutant is capable of rescuing cells from TRPML3(P)-induced apoptosis. Importantly, this TRPML(P-KK)-mediatedrescue was not limited to TRPML3(P)-expressing cells only,because similar neutralizing effects of dominant-negativeTRPML(P-KK) channels on the cytotoxic effects of TRPML2(P)(Fig. 2B) and TRPML1(P) (Fig. 2C) expression were alsoobserved in the presence and absence of PMCA2 (Fig. 2B,C).Therefore, these data combined with the findings of the co-IPassays described above (supplementary material Figs S1, S2),suggest that dominant-negative TRPML(P-KK)s attenuate thecytotoxic effects of constitutively active TRPML(P) channelsvis-à-vis homo- and heteromeric physical interactions.

3115Multimeric TRPML channel assemblies

Dominant-negative TRPML2(P-KK) channels significantlyattenuate the currents of TRPML3(P) channels inheteromeric complexesAs suggested by the cell viability experiments, it is likely thatinactive TRPML(P-KK) channel subunits exert their dominant-negative effect over constitutively active TRPML(P) channel

Fig. 2. Dominant-negative TRPML(P-KK)s attenuate the cytotoxic effectsof constitutively active TRPML(P) channels. (A)HeLa cells were co-transfected with tdTomato-TRPML3(P) in combination with PMCA2 (+),empty vector (–), and GFP/YFP-tagged expression constructs, as indicated.Alexa-Fluor-647–AnnexinV staining was performed and the results of stainingare summarized as in Fig. 1B. (B)Same as A, with TRPML2(P)-tdTomato co-transfected cells. (C)Same as A, with tdTomato-TRPML1(P) co-transfectedcells. For A–C, n3 independent experiments of at least 45 cells per assay.Indicated P values are relative to the left-most bar in each histogram: *P<0.05,**P<0.01.

Jour

nal o

f Cel

l Sci

ence

subunits by functionally interacting with and affecting the currentof TRPML(P) channels. To support this notion, we looked atwhether expression of TRPML3(P) together with TRPML2(P-KK)reduces the robust, TRPML3(P)-associated currents that lead tocell Ca2+ overload and death. In HEK cells, where our currentrecordings were performed, we observed a clear difference inoverall cell morphology between cells expressing TRPML3(P)only and those expressing TRPML2(P-KK) + TRPML3(P).Consistent with the results in Fig. 2A, we found that whenTRPML3(P) was expressed alone, most TRPML3(P)-positive cellsfeatured membrane blebbing and loss of morphology, which wasreduced when the channel was coexpressed with TRPML2(P-KK)(Fig. 3A).

In order to clarify the role of the TRPML3(P) current in theexperiments depicted in Fig. 3A, we recorded whole-cell currentsfrom HEK cells expressing TRPML3(P) and TRPML2(P-KK) inthe presence of PMCA2. The coexpression of PMCA2 was usedto help maintain cell integrity, which facilitates whole-cellrecordings. Cells expressing TRPML3(P) in the absence ofTRPML2(P-KK) feature robust, inward rectifying currents (Fig.3B, red), consistent with previous reports describing the constitutivechannel activity of TRPML3(P) (Grimm et al., 2007; Grimm et al.,

3116 Journal of Cell Science 123 (18)

2009; Kim et al., 2007; Lev et al., 2010; Nagata et al., 2008; Xuet al., 2007). By contrast, cells expressing TRPML2(P-KK) in theabsence of TRPML3(P) do not exhibit any significant currents(Fig. 3B, green), implying that TRPML2(P-KK) is an inactivechannel. Importantly, significantly reduced current density wasobserved in cells coexpressing both channels (Fig. 3B, black).Because this effect might be attributed to reduced surface expressionof TRPML3(P) in the presence of TRPML2(P-KK), we assayedthe total and surface expression of both channels when expressedalone and in combination. The total and surface expression ofTRPML2(P-KK) are modestly enhanced by the presence ofTRPML3(P) (Fig. 3C–E) and the total expression of TRPML3(P)is modestly enhanced by the presence of TRPML2(P-KK)(Fig.3F,G). However, the surface expression of TRPML3(P) is greatlyenhanced by the presence of TRPML2(P-KK)(Fig. 3F,H). Thissuggests that TRPML3(P)-mediated current is reduced in thepresence of TRPML2(P-KK), despite enhanced overall surfaceexpression of the constitutively active channel.

These findings (Fig. 3), together with the co-IP (supplementarymaterial Fig. S2A) and cell viability (Fig. 2A) assays describedabove, indicate that TRPML2(P-KK) rescues cells fromTRPML3(P)-induced apoptosis by physically interacting with

Fig. 3. TRPML3(P) current density is significantlyreduced when the channel is coexpressed withdominant-negative TRPML2(P-KK). (A)Wide-fieldconfocal images showing representative cell morphologyof HEK cells transfected with the indicated constructs.tdTomato-TRPML3(P) fluorescence (right panels) isshown next to DIC images (left panels). White arrowsindicate cells with abnormal morphology. Yellow arrowsindicate cells with normal morphology. Scale bars: 20mm.(B)Top: Representative I-V curves of whole-cell currentsmeasured from HEK cells coexpressing the indicatedchannels together with PMCA2. Bottom: Histogram of theaverage current density (at –120 mV) exhibited in HEKcells co-transfected with the indicated channels andPMCA2; n is number of cells analyzed; *P<0.05. (C)HEKcells were co-transfected with PMCA2 and the indicatedexpression constructs (above top panel). Surface proteinswere biotinylated and subsequently isolated fromnormalized total cell lysates. Pictured are immunoblots oftotal lysate (Total) and surface protein (Surface) fractionsthat were probed for GFP+YFP and TRPML2(P-KK)-YFP.The endogenous GAPDH protein was probed in totallysates as a loading control. Transfection of unconjugatedGFP and YFP fluorescent proteins served as a negativecontrol for the assays. (D,E)Mean total and surfaceexpression of TRPML2(P-KK) in the presence ofTRPML3(P) (black), relative to its expression in theabsence of TRPML3(P) (green), as determined from theexperiments depicted in C. Experiments involvingTRPML2(P-KK) alone were assigned a value of 1; n3.(F)Same as C, with tdTomato-tagged proteins beingprobed for total and surface expression. Transfection of theunconjugated tdTomato fluorescent protein served as anegative control for the assays. (G,H)Mean total andsurface expression of TRPML3(P) in the presence ofTRPML2(P-KK) (black), relative to its expression in theabsence of TRPML2(P-KK) (red), as determined from theexperiments depicted in F. Experiments involvingTRPML3(P) alone were assigned a value of 1; n3.

Jour

nal o

f Cel

l Sci

ence

TRPML3(P), in dominant-negative fashion, to form surface-expressed inactive heteromeric channel complexes. Previous studieshave shown that extracellular Ca2+ is the primary driving force incausing intracellular Ca2+ overload and cell death in the presenceof the surface-expressed, constitutively active, inwardly rectifying,Ca2+-permeable TRPML3(P) channel (Grimm et al., 2007; Kim etal., 2007). To confirm that functional physical interactions underliethe effects we have described in Fig. 3, we analyzed the TRPM8current when the channel was coexpressed with TRPML2(P-KK).TRPM8, as we have shown, does not physically interact withTRPML2(P-KK) or any other TRPML(P-KK) dominant-negativemutant (supplementary material Fig. S2B). Consistent with thisfinding, coexpression of TRPML2(P-KK) with TRPM8 did notreduce the outwardly rectifying and menthol-facilitated TRPM8-derived (Varnai et al., 2006) whole-cell currents (supplementarymaterial Fig. S3). This control further supports the notion that thereduction of TRPML3(P) channel activity that occurs in thepresence of TRPML2(P-KK) (Fig. 3B, black) can be explained byspecific physical interactions between the two channels(supplementary material Fig. S2A) that lead to the formation ofheteromultimeric channel assemblies with reduced channel activity.Conversely, where there is no interaction of TRPML2(P-KK) with

3117Multimeric TRPML channel assemblies

another TRP channel (such as TRPM8), there is no dominant-negative inhibitory effect on the current density of that channel.

Dominant-negative TRPML3(P-KK) channels significantlyattenuate the currents of TRPML1(P) channels inheteromeric complexesLike recombinant TRPML3(P), recombinant TRPML1(P) is alsoa surface-expressed, constitutively active, inwardly rectifying,Ca2+-permeable channel (Dong et al., 2008; Dong et al., 2009;Samie et al., 2009; Xu et al., 2007) that is implicated in causingcell death (Fig. 1D). In HEK cells, TRPML1(P) causes celldegeneration that is attenuated when the channel is coexpressedwith the inactive TRPML3(P-KK) channel (Fig. 4A). Consistentwith this finding, TRPML1(P)-derived whole-cell currents (Fig.4B, red) are significantly reduced in the presence of TRPML3(P-KK) (Fig. 4B, black). However, given that native TRPML1 isstrictly intracellular (Kim et al., 2009; Zeevi et al., 2009), it ispossible that coexpression with TRPML3(P-KK) reducesTRPML1(P)-derived whole-cell currents by also reducing thesurface expression of TRPML1(P) in favor of the naturalTRPML1 intracellular localization. Fig. 4C–E shows that theoverall surface expression of TRPML3(P-KK) is enhanced in the

Fig. 4. TRPML1(P) current density is significantly reducedwhen the channel is coexpressed with dominant-negativeTRPML3(P-KK). (A)Same experimental procedures as in Fig.3A with the indicated constructs. tdTomato-TRPML1(P)fluorescence (right panels) is shown next to DIC images (leftpanels); Scale bars: 20mm. (B)Same experimental procedures asin Fig. 3B; *P<0.05. (C)HEK cells were co-transfected withPMCA2, GFP-TRPML3(P-KK), and either tdTomato ortdTomato-TRPML1(P). Total and surface expression of GFP-TRPML3(P-KK) was analyzed as described in Fig. 3C.(D,E)Summary of experiments in C as in Fig. 3D,E; n3.(F)HEK cells were co-transfected with PMCA2, tdTomato-TRPML1(P), and either GFP or GFP-TRPML3(P-KK). Totaland surface expression of tdTomato-TRPML1(P) was analyzedas described in Fig. 3F. (G,H)Summary of experiments in F asin Fig. 3G,H; n3.

Jour

nal o

f Cel

l Sci

ence

presence of TRPML1(P), and Fig. 4F–H shows that the overallsurface expression of TRPML1(P) is enhanced even more so inthe presence of TRPML3(P-KK). Thus, TRPML1(P)-mediatedcurrent is reduced in the presence of TRPML3(P-KK) despite itsenhanced overall surface expression. Together, these resultsindicate that TRPML3(P-KK) reduces TRPML1(P)-inducedapoptosis (Fig. 2C) by physically interacting with TRPML1(P),in dominant-negative fashion, to form surface-expressed inactiveheteromeric channel complexes.

3118 Journal of Cell Science 123 (18)

Dominant-negative TRPML1(P-KK) channels significantlyattenuate the currents of TRPML2(P) channels inheteromeric complexesLike recombinant TRPML3(P) and TRPML1(P), recombinantTRPML2(P) is also a surface-expressed, constitutively active,inwardly rectifying, Ca2+-permeable channel (Dong et al., 2008;Lev et al., 2010; Samie et al., 2009) that is implicated in causingcell death (Fig. 1C). In HEK cells, TRPML2(P) causes celldegeneration that is attenuated when the channel is coexpressed

Fig. 5. TRPML2(P) current density is significantlyreduced when the channel is coexpressed withdominant-negative TRPML1(P-KK). (A)Sameexperimental procedures as in Fig. 3A with the indicatedconstructs. TRPML2(P)-tdTomato fluorescence (rightpanels) is shown next to DIC images (left panels). Scalebars: 20mm. (B)Same experimental procedures as inFig. 3B; *P<0.05. (C)HeLa cells were transfected withGFP-TRPML1(P-KK) and immunostained for CD63.Cells were then imaged by confocal microscopy. Scalebar: 10mm. (D)Images similar to those in C, from atleast 15 cells in three independent experiments, wereused to determine the mean percentage of GFP-TRPML1(P-KK) that overlapped with CD63. (E–J) Same experimental procedures as in Fig. 4C–Hwith the indicated constructs; n3 independentexperiments per assay.

Jour

nal o

f Cel

l Sci

ence

with the inactive TRPML1(P-KK) channel (Fig. 5A). Consistentwith this finding, TRPML2(P)-derived whole-cell currents (Fig.5B, red) are significantly reduced in the presence of TRPML1(P-KK) (Fig. 5B, black).

We have found that the majority of recombinant TRPML1(P-KK) is expressed intracellularly and colocalizes, for the most part,with the late endosomal/lysosomal marker CD63 (Fig. 5C,D), likewild-type TRPML1 (Vergarajauregui and Puertollano, 2006). Still,a small fraction of TRPML1(P-KK) also reaches the cell surface(Fig. 5E). Interestingly, a recent report has shown that wild-type(WT)-TRPML3 surface expression and agonist-induced whole-cell currents are reduced when the channel is coexpressed with theintracellular WT-TRPML1 (Grimm et al., 2010). However, thesame report also suggested that coexpression of WT-TRPML1with TRPML3(P) did not bear any effect upon TRPML3(P) whole-cell currents. Thus, we asked whether TRPML1(P-KK) exerts itsdominant-negative effect over TRPML2(P) by interacting with thechannel intracellularly and preventing it from reaching the cellsurface, or by interacting with the channel on the cell surface andattenuating its current.

Fig. 5E–G shows that a small fraction of overexpressedTRPML1(P-KK) reaches the cell surface. Interestingly, this fractionis enhanced when the channel is coexpressed with TRPML2(P)(Fig. 5E–G). Fig. 5H–J shows that TRPML2(P) total and surfaceexpression are modestly enhanced in the presence of TRPML1(P-KK). This suggests that TRPML2(P) remains surface-expressedeven in the presence of the primarily intracellular, but partiallysurface-expressed TRPML1(P-KK). Thus, our data indicate thatTRPML1(P-KK) exerts its dominant-negative effect onTRPML2(P) currents in surface-expressed heteromeric inactivechannel assemblies that prevent TRPML2(P)-mediated apoptosis(Fig. 2B).

Together, our findings from co-IP, cell viability assays, whole-cell current recordings, and cell surface expression assays indicatethat TRPML mutant regulation of cell viability is probably mediatedby cell-surface-localized heteromultimeric protein–proteininteractions that constitute functional or non-functionalheteromultimeric TRPML channel assemblies.

Dominant-negative TRPML1 channels inhibit starvation-induced autophagyHaving established that gain-of-function and dominant-negative,loss-of-function TRPML mutants form heteromultimeric TRPMLchannel assemblies that affect cell viability, we next asked whetherheteromultimeric TRPML channel assemblies composed ofdominant-negative and WT-TRPMLs affect starvation-inducedautophagy (SIA). The process of SIA is not directly regulated byTRPML1 (Vergarajauregui et al., 2008), but is directly regulatedby TRPML3 (Kim et al., 2009; Martina et al., 2009) and possiblyTRPML2 (whose role in SIA has not yet been determined).

Autophagy is a process whereby macromolecules and organellesare processed for degradation in lysosomes via specializedintracellular vesicles in response to various cell stressors including,but not limited to, starvation conditions (Cuervo, 2004b; Yorimitsuand Klionsky, 2005). The autophagic process begins with theformation of double-membrane autophagosomes and can bedetected by examining the distribution of autophagosomal lightchain three (LC3) in cells because this protein is recruited intoautophagosomes upon induction of autophagy (Klionsky et al.,2007). LC3 displays a soluble cytosolic distribution under basalconditions, which transforms into a punctate vesicular distribution

3119Multimeric TRPML channel assemblies

as LC3 becomes lipidated during autophagy. Thus, autophagyinduction, under starvation conditions, can be quantified bycounting LC3-positive puncta within RFP-LC3-transfected cells.

The potential role of TRPML2 in the regulation of starvation-induced autophagy has not been elucidated. Thus, we assayedRFP-LC3 distribution in cells stably transfected with microRNA-adapted short hairpin RNAs (shRNAmirs) designed to knockdownnative TRPML2 or TRPML3 expression (Zeevi et al., 2009). Asseen in Fig. 6A and the summary in Fig. 6B, a primarily diffusecytosolic RFP-LC3 distribution in fed control cells (transfectedwith non-silencing shRNAmir) is transformed into a punctatedistribution in starved control cells, where the autophagic machineryis expected to be upregulated. In addition, consistent with a previousreport describing the active role of TRPML3 in the regulation ofSIA (Kim et al., 2009), we confirm that knockdown of nativeTRPML3 expression reduces autophagosome formation in starvedTRPML3 knockdown cells, with respect to starved control cells(Fig. 6A,B). Importantly, in starved TRPML2 knockdown cells,we did not detect any discrepancy in autophagic levels with respectto control (Fig. 6A,B). Moreover, whereas overexpression ofrecombinant WT-TRPML3 markedly enhances autophagosomeformation in starved transfected cells (Kim et al., 2009; Martina etal., 2009) with respect to starved control cells; overexpression ofWT-TRPML2 does not affect typical autophagic levels under thesame conditions (Fig. 6C,D). Together, these results indicate forthe first time that, unlike TRPML3, TRPML2 does not appear toplay an active role in SIA.

Overexpression of a dominant-negative TRPML3-KK constructinhibits starvation-induced autophagy by, presumably, interactingwith native TRPML3 and reducing its channel activity (Kim et al.,2009). Here, we show that the TRPML3(P-KK) dominant-negativemutant also inhibits autophagosome formation in starved cells(Fig. 6C,D). Strikingly, we also found that overexpression of theTRPML1(P-KK) dominant-negative mutant unexpectedly inhibitedSIA (Fig. 6C,D; see below).

The TRPML1(P-KK) mutant physically interacts with WT-TRPML1 and WT-TRPML2 (supplementary material Fig. S4),suggesting that endogenous TRPML1 and TRPML2 function canbe compromised when these native channels complex with theTRPML1(P-KK) dominant-negative channel. However, previouslydescribed experiments indicate that SIA is not regulated byendogenous TRPML1 function because neither the absence(Vergarajauregui et al., 2008) nor overexpressed presence of theTRPML1 channel (confirmed in Fig. 6C,D) leads to any discernibleeffect upon SIA (Vergarajauregui et al., 2008). Our data alsodemonstrate a similar argument regarding endogenous TRPML2function in SIA regulation (Fig. 6A–D). Thus, the inhibition of SIAby TRPML1(P-KK) cannot be attributed to physical interactionswith, and channel inactivation of, native TRPML1 or nativeTRPML2. Rather, a more likely explanation for this effectimplicates heteromultimeric interactions between TRPML1(P-KK)and native TRPML3, a channel that plays a definitive SIA-stimulatory role (Kim et al., 2009; Martina et al., 2009). Indeed,supplementary material Fig. S4 demonstrates that, in addition tointeracting with WT-TRPML1 and WT-TRPML2, TRPML1(P-KK) also physically interacts with WT-TRPML3, which suggeststhat endogenous TRPML3 function can also be compromised whenthe native channel complexes with the TRPML1(P-KK) dominant-negative channel. Therefore, we propose that inhibition of SIA,resulting from TRPML1(P-KK) overexpression, is caused byformation of inactive heteromultimeric channel assemblies

Jour

nal o

f Cel

l Sci

ence

comprised of TRPML1(P-KK) and native TRPML3. This is astriking demonstration of how a single TRPML1 mutant can affecta physiological function mediated by another TRPML channel (i.e.TRPML3).

In view of our findings in Figs 1, 2 and 5, it is likely thatTRPML1(P-KK) is an inactive channel. We also propose that thedominant-negative nature of TRPML1(P-KK) correlates with itsability to physically interact with wild-type or constitutively activeTRPML channels to form inactive channel assemblies(supplementary material Figs S1, S2, S4). Accordingly, we testedSIA in cells expressing the known channel-impairedTRPML1(F465L) mutant (Dong et al., 2008; Kiselyov et al., 2005).The F465L missense mutation in TRPML1 was first described incompound heterozygosity with a TRPML1-null allele in an MLIVpatient (Bargal et al., 2001; Kiselyov et al., 2005). In addition, thismutation does not interfere with the ability of TRPML1 tophysically interact with both WT-TRPMLs (supplementary materialFig. S5A) and constitutively active TRPML(P) mutants (data notshown). Therefore, we reasoned that if impaired channel activityof one component of a TRPML channel assembly leads to adominant-negative effect, then heteromultimeric channel assemblies

3120 Journal of Cell Science 123 (18)

composed of native TRPML3 and the channel-impairedTRPML1(F465L) mutant should result in functionally reducedactivity of these assemblies. This in turn might affect SIA, if SIAis regulated by TRPML heteromultimerization. As seen in Fig.6C,D, this is precisely what occurs. Inhibition of SIA byTRPML1(F465L) parallels the inhibition found in TRPML3(P-KK)- and TRPML1(P-KK)-expressing cells. This suggests thatsignificantly reduced channel activity of one, naturally occurring,mutated TRPML channel results in dominant-negative inhibitionof native TRPML3 protein function at the cellular level by renderingthe heteromultimeric complex less active.

Dominant-negative properties of TRPML1(F465L)The TRPML1(F465L) mutation displays autosomal recessiveinheritance in humans (Bargal et al., 2001), meaning that thedominant-negative behavior, demonstrated in our assay of SIA,should not necessarily be expected even when the channel isoverexpressed in cells. As a result, we sought to further characterizethe dominant-negative properties of TRPML1(F465L) in our cellviability assays (described in Fig. 2). As a negative control for theassay, we compared the cell viability of TRPML(P) +

Fig. 6. Dominant-negative TRPML1 mutantsinhibit starvation-induced autophagy.(A,B)Knockdown of native TRPML2 expressiondoes not affect autophagic response in starvedcells. (A)HEK cells, stably transfected with non-silencing (control), TRPML3-depleting (TRPML3KD), or TRPML2-depleting (TRPML2 KD)shRNAmirs, were transfected with RFP-LC3followed by incubation in feeding (Fed) orstarvation (Starved) media for 2 hours.Representative confocal images of RFP-LC3 arepresented for each experiment. Scale bars: 10mm.(B)Histogram of the average number of RFP-LC3puncta in A, as determined from at least 13 cells inthree independent experiments. TRPML3-specificand TRPML2-specific knockdown was verified inthe appropriate cell lines, as previously described(Zeevi et al., 2009), in parallel with eachexperiment. (C,D)TRPML1(P-KK) andTRPML1(F465L) mutants inhibit starvation-induced autophagy. (C)HeLa cells were co-transfected with RFP-LC3 and the indicatedGFP/YFP-tagged expression constructs, followedby treatment as in A. Representative confocalimages of RFP-LC3 are presented. Scale bars:10mm. (D)Histogram of the average number ofRFP-LC3 puncta in C, as determined from at least10 cells in three independent experiments.

Jour

nal o

f Cel

l Sci

ence

TRPML1(F465L) co-transfected cells with that of TRPML(P) +WT-TRPML1 co-transfected cells in order to preclude thepossibility that WT-TRPML1 channels [which physically interactwith TRPML(P) mutants (data not shown)] might also rescueTRPML(P) expressing cells. Our results indicate that neither WT-TRPML1 nor TRPML1(F465L) channels cause apoptosis(supplementary material Fig. S5B). However, only coexpression of

3121Multimeric TRPML channel assemblies

TRPML(P) mutants with TRPML1(F465L), as opposed to WT-TRPML1, significantly rescued cells from apoptosis(supplementary material Fig. S5B). This finding is corroborated inHEK cells where TRPML1(F465L) rescues TRPML3(P)-expressing cells from cell degeneration (Fig. 7A).

TRPML1(F465L) displays impaired channel activity (Fig. 7B,green), but its inhibition of the TRPML3(P) whole-cell current

Fig. 7. TRPML3(P) current density is significantlyreduced when the channel is coexpressed withdominant-negative TRPML1(F465L). (A)Sameexperimental procedures as in Fig. 3A with theindicated constructs. tdTomato-TRPML3(P)fluorescence (right panels) is shown next to DICimages (left panels). Scale bars: 20mm. (B)Sameexperimental procedures as in Fig. 3B; **P<0.01.(C,D)Same experimental procedure as in Fig. 5C,Dwith the TRPML1(F465L)-YFP construct. n3independent experiments of at least 15 cells. (E–J)Same experimental procedures as in Fig. 4C–H withthe indicated constructs. n3 independentexperiments per assay.

Jour

nal o

f Cel

l Sci

ence

density is noteworthy (Fig. 7B, black) because most of theoverexpressed channel is expressed intracellularly and colocalizes,for the most part, with CD63 (Fig. 7C,D). Nevertheless, Fig. 7E–G demonstrates that despite its predominant intracellularlocalization, a small fraction of overexpressed TRPML1(F465L)reaches the cell surface. Moreover, this surface fraction is tripledwhen the channel is coexpressed with TRPML3(P) (Fig. 7E–G). Inaddition, Fig. 7H–I shows that although total TRPML3(P)expression levels are unaffected by TRPML1(F465L), its surfaceexpression levels are modestly enhanced (Fig. 7J). This suggeststhat TRPML3(P) remains surface-expressed even in the presenceof the primarily intracellular, but partially surface-expressedTRPML1(F465L). Thus, our results indicate that TRPML1(F465L)can function as a dominant-negative channel by complexing withsurface-expressed TRPML3(P) to form heteromeric inactivechannel assemblies that prevent TRPML3(P)-mediated apoptosis(supplementary material Fig. S5B).

Together, our data suggest that TRPML channels with reducedactivity, such as the naturally occurring TRPML1(F465L) mutant,are capable of affecting the function of homo- andheteromultimeric TRPML channel assemblies in dominant-negative fashion.

DiscussionIn this report, we demonstrated the ability of all three TRPMLs tointeract with each other, in every possible dual combination, toform homo- and heteromultimeric channel assemblies. Moreover,we also found that TRPML channel assemblies form, even whenvarious missense mutations are located within the putative channelpore region of each subunit. This property of TRPML channelsbecame useful when we assayed the effects of heteromultimericTRPML channel assemblies, comprised of mutant and wild-typechannel subunits, upon cellular function. Indeed, our observationthat TRPML1(P-KK) inhibits starvation-induced autophagysuggests a nascent function for TRPML1 in the regulation of SIA.This arises from the ability of TRPML1 to complex with, andaffect the functionality of, TRPML3 channels even when possessingdominant-negative mutations in its channel pore. Furthermore,our findings that the MLIV-associated channel-impairedTRPML1(F465L) mutant inhibits SIA, and also behaves indominant-negative fashion in multiple assays, implies that channelfunctionality is key to establishing some TRPML-based dominant-negative phenotypes. Indeed, this report is the first to demonstratedominant-negative properties in an MLIV-associated missensemutant protein.

Although a thorough clinical and phenotypic assessment has notbeen performed with carriers of the F465L mutation in TRPML1,the mutation originally presented with apparent autosomal recessiveinheritance in the family where the mutation was identified (Bargalet al., 2001). On the other hand, we have seen that this mutationaccrues dominant-negative properties to the TRPML1 channel,and it should also be noted that heterozygote carriers of MLIV-associated mutations, in general, exhibit slightly more lysosomalinclusions than normal in electron micrographs (our unpublishedobservations). Hence, if the F465L mutation is truly dominant-negative, why are the carriers of this mutation presumablyasymptomatic for MLIV? We offer three possible explanations.First, it could be that the F465L mutant transcript is degraded inF465L/wild-type heterozygotes, thereby eliminating the negativeeffects of TRPML1(F465L) expression. Second, assuming thatF465L expression levels might parallel wild-type transcript

3122 Journal of Cell Science 123 (18)

expression levels, it is possible that functional TRPML1 channelcomplexes (comprised of wild-type-only channels), and non-functional F465L-containing TRPML1 channel complexes, coexistwithin cells. If proper lysosomal function requires only a minimalamount of functional TRPML1 channels, then this might alsoexplain why F465L carriers are asymptomatic for MLIV. Third, itis possible that F465L channels are only selectively dominant,which means that they are dominant in regulating certain cellularfunctions but not necessarily in other more crucial cellularfunctions. Due to inaccessibility of cultured cells from the familyof the F465L-bearing MLIV patient and those of the patient himself,we cannot explore the possible explanations for the recessiveinheritance of the F465L mutation in TRPML1. Nonetheless, thenotion that an MLIV-associated mutation exhibits a dominant-negative effect on physiological mechanisms offers newpossibilities in developing therapeutically relevant strategies fortreating TRPML-related pathologies.

Recombinant TRPML1(F465L) traffics primarily to lateendosomal/lysosomal intracellular compartments, as does nativeTRPML1 and a fraction of native TRPML3 (Kim et al., 2009;Zeevi et al., 2009). Autophagy is an intracellular process (Cuervo,2004a) that is also regulated by TRPML3 (Kim et al., 2009). Thissuggests that TRPML1(F465L) is likely to inhibit SIA bycomplexing with native TRPML3 in intracellular compartments.However, because some recombinant TRPML1(F465L) reachesthe cell surface (Kiselyov et al., 2005), we cannot preclude thepossibility that the channel also complexes with surface-localizedTRPML3 to inhibit SIA. However, this latter possibility is unlikelybecause autophagy is an intracellular process (Cuervo, 2004a).Nevertheless, the partial surface expression of TRPML1(F465L)and other recombinant TRPML1 mutants did become useful in thisstudy in describing functional surface-expressed heteromericchannel complexes. Future studies should investigate thesephysiological interactions in a native system in order to verify theimplications of heteromeric channel assemblies.

The notion that heteromeric interactions among TRPMLs canpotentially affect cellular function and homeostasis has significantramifications when investigating TRPML function. Our studydemonstrates the importance of investigating the effect of a singlemissense TRPML mutant on the functions and related mechanismsof not one but all three native TRPML channels. Indeed,recombinant dominant-negative missense TRPML mutants affordinvestigators a valuable tool for potentially downregulating allthree native TRPMLs at once. This might facilitate comparisonsbetween three-TRPML-coding human and murine TRPML modelsand Caenorhabditis elegans and Drosophila models that code fora single TRPML ortholog. Such comparisons could ultimatelyserve to translate discoveries from TRPML-based animal modelsinto therapies for TRPML-associated disorders such as MLIV.

In addition, future studies should also take into account thepotential residual effects of the absence or abundance of a singleTRPML upon the functions of remaining or other TRPML paralogs.For instance, it could be that MLIV pathogenesis results, at leastin part, from the downregulation of native TRPML2 and TRPML3that results from the absence of TRPML1 in MLIV cells. Forexample, high levels of basal (as opposed to starvation-induced)autophagy have been detected in MLIV patient TRPML1-null cells(Vergarajauregui et al., 2008), whereas overexpression of TRPML3leads to upregulation of basal autophagic levels (Kim et al., 2009;Martina et al., 2009). This, taken together with our observation thatnative TRPML1 dynamically interacts with native TRPML3 (Zeevi

Jour

nal o

f Cel

l Sci

ence

3123Multimeric TRPML channel assemblies

et al., 2009), suggests that there might be interplay betweenTRPML1 and TRPML3 in the regulation of basal autophagy. WhenTRPML1 is present, the ability of TRPML3 to stimulate theformation of autophagosomes is attenuated accordingly; but whenTRPML1 is absent, as is mostly the case in MLIV cells (Bargal etal., 2001; Sun et al., 2000), stimulation of autophagy by TRPML3isleft unregulated, allowing for the increased autophagosomeformation that is found in MLIV patient cells. Future studies shouldinvestigate this potential TRPML1–TRPML3 interplay in light ofour findings here.

Materials and MethodsExpression constructsThe TRPML1-YFP expression construct, used in this study, was described previously(Zeevi et al., 2007). The EGFP ORF, minus a stop codon, and the complete EYFPORF were derived from the pEGFP-C2 and pEYFP-N1 vectors (Clontech),respectively, and cloned into a slightly modified version of the pcDNA3 vector(Invitrogen). tdTomato nucleotide sequences, with and without a stop codon, werederived from the pRSET-B tdTomato bacterial expression construct, kindly providedby Roger Tsien (UCSD, San Diego, CA), and cloned into the same pcDNA3 vector.Subsequently, the human TRPML1 and TRPML3 ORFs were cloned downstream toand in-frame with the GFP and tdTomato sequences on the pcDNA3-GFP andpcDNA3-tdTomato vectors, respectively. Similarly, the human TRPML2 ORF wascloned upstream to and in-frame with EYFP and tdTomato sequences on theappropriate pcDNA3-EYFP and pcDNA3-tdTomato vectors, respectively. Theresulting TRPML fusion-protein constructs were then used to derive TRPML mutantsby site-directed mutagenesis according to a previously described method (Ho et al.,1989). Insert orientation and polymerase fidelity were verified by restriction enzymemapping and sequencing for each construct. The TRPM8, pmRFP-LC3, pEGFP-TRPML3, and PMCA2z/a constructs, used in this study, were kindly provided bySharona E. Gordon (University of Washington, Seattle, WA), Tamotsu Yoshimori(Osaka University, Osaka, Japan), Shmuel Muallem (University of TexasSouthwestern Medical Center, Dallas, TX), and Peter G. Gillespie (Oregon HearingResearch Center, Portland, OR), respectively.

Cell cultureHeLa and HEK cells were grown at 37°C in Dulbecco’s modified Eagle’s medium(DMEM) supplemented with 10% fetal bovine serum and 1% Pen-Strep (BiologicalIndustries). Transfections were performed with the Fugene 6 (Roche) or TransIt(Mirus) transfection reagents with equal amounts of cDNA. Wherever not specified,transfections were supplemented with GFP, YFP, tdTomato, or empty vector cDNAin order to equalize DNA concentrations in all transfections. For immunoprecipitationand electrophysiology experiments, PMCA2 was added to all transfections unlessthe experiment included WT-TRPML constructs. HEK cells, stably transfected withnon-silencing, TRPML3-depleting, or TRPML2-depleting shRNAmirs were describedpreviously (Zeevi et al., 2009).

AnnexinV stainingHela cells were grown on coverslips and transfected with fluorescent proteinconstructs and PMCA2 or empty vector. At 18 hours post-transfection, cells werestained with Alexa-Fluor-647–AnnexinV (BioLegend) and analyzed on a confocalmicroscope (Olympus Fluoview 300 IX70) as previously described (Lev et al.,2010).

AntibodiesMonoclonal anti-GFP (which recognizes both GFP and YFP) and polyclonal anti-DsRed (which recognizes tdTomato) antibodies were purchased from Clontech.Polyclonal anti-GFP (598; which recognizes both GFP and YFP), polyclonal anti-TRPM8 (including a TRPM8-blocking peptide), monoclonal anti-GAPDH (6C5),and monoclonal anti-CD63 were from MBL (Woburn, MA), Alomone Labs(Jerusalem, Israel), Santa Cruz Biotechnology and BD Biosciences, respectively. Allsecondary antibodies, except for Alexa Fluor 647 goat anti-mouse IgG (Invitrogen),were purchased from Jackson ImmunoResearch.

Immunoprecipitation and immunoblottingHEK cells were solubilized on ice in lysis buffer containing 1% Triton X-100, 25mM Tris–HCl (pH 7.5), 150 mM NaCl, 5 mM EDTA, and complete TM proteaseinhibitor cocktail (Santa Cruz Biotechnology). After centrifugation to clear debris,cell lysates were then divided into separate tubes for immunoprecipitation with twodifferent antibodies in the presence of Protein A/G PLUS-Agarose (Santa CruzBiotechnology) at 4°C overnight, with shaking. Unless indicated otherwise,immunoprecipitation was performed with the GFP/YFP-recognizing polyclonal anti-GFP antibody, or with the tdTomato-recognizing polyclonal anti-DsRed antibody.Following overnight incubation, all immunoprecipitates were washed five timeswith lysis buffer followed by a final wash in 50 mM Tris–HCl (pH 8). Precipitated

proteins were eluted with 2× LDS sample buffer containing 2× NuPAGE ReducingAgent (Invitrogen) and heated to 60°C for 10 minutes prior to separation on 4–12%NuPAGE Novex Bis-Tris gels (Invitrogen), and then immunoblotted according tostandard procedures. Immunoprecipitates were blotted for GFP/YFP with themonoclonal anti-GFP antibody; for tdTomato, with the polyclonal anti-DsRedantibody; or for TRPM8, with the polyclonal anti-TRPM8 antibody. Secondarydetection was performed with HRP-conjugated goat anti-mouse or mouse anti-rabbitIgG light-chain-specific antibodies, where appropriate.

Surface biotinylation assayTransfected HEK cell surface proteins were biotinylated and isolated from normalizedcell lysates, as previously described (Lev et al., 2010). Total lysate and cell surfacetdTomato-tagged and GFP/YFP-tagged proteins were probed by immunoblot usingeither the polyclonal anti-DsRed or the polyclonal anti-GFP antibodies, respectively.The mouse monoclonal anti-GAPDH antibody was used as loading control.Immunoblots were quantitated by densitometry analysis using Photoshop software(Adobe Systems).

ElectrophysiologyHEK cells were seeded onto polylysine-coated plates at 25% confluence. At 20–24hours before the experiment, cells were transfected to induce expression of channels.Whole-cell currents were recorded at room temperature using borosilicate patchpipettes of 3–5 M and an Axopatch 200B (Molecular Devices) voltage-clampamplifier. Voltage-clamp pulses in the whole-cell configuration, using voltage rampsfrom –140 to 140 mV in 1 second, were generated and the data captured using aDigidata 1440A interfaced to a computer running the pClamp 10 software (MolecularDevices). Currents were filtered using the 8-pole low pass Bessel filter of the patch-clamp amplifier at 5 kHz and sampled at 50 kHz. To measure I–V curves withminimal distortions, only cells with low (<10 MΩ) series resistance were used andthe series resistance was compensated by 80%.

SolutionsThe extracellular Tyrode solution was prepared as a Ca2+ nominal based solution towhich either 0.5 mM EGTA was added for TRPM8 experiments, or 1.5 mM CaCl2for all other experiments. The solution was titrated to pH 7.4 and otherwise contained145 mM NaCl, 5 mM KCl, 1 mM MgCl2, 15 mM HEPES, and 10 mM glucose. Theintracellular solution was titrated to pH 7.2 and contained 130 mM CsCl, 1 mMMgCl2, 10 mM EGTA, 10 mM HEPES, 2 mM Na2-ATP and 4.1 mM CaCl2. Cellswere perfused via a BPS-4 valve control system (Scientific Instruments) at a rate of~30 chamber volumes per minute. Menthol was added to the extracellular solutionat a final concentration of 200 mM.

Confocal microscopyFor immunofluorescence, HeLa cells were grown on coverslips and transfected witheither GFP-TRPML1(P-KK) or TRPML1(F465L)-YFP. Cells were then fixed andimmunostained according to a previously described method (Zeevi et al., 2009).Immunostaining for CD63 was performed with monoclonal anti-CD63 and AlexaFluor 647 goat anti-mouse antibodies. Confocal images were acquired on an OlympusFluoview 300 IX70 microscope, with a 60 oil immersion objective, and processedoff-line with ImageJ software.

For assaying autophagy, HeLa cells were co-transfected with pmRFP-LC3 andanother GFP/YFP-tagged protein construct. At 18 hours post-transfection, the growthmedium was replaced with either fresh growth or starvation medium (Earle’s balancedsalt solution; Biological Industries) for 2 hours at 37°C. Following treatment, cellswere fixed in 4% formaldehyde for 30 minutes at room temperature, washed threetimes with 200 mM NH4Cl in phosphate-buffered saline, and subsequently mountedonto slides with Antifade Solution (Vysis). Confocal images of cells exhibiting bothRFP and GFP/YFP emission were then acquired and processed as described above.Confocal images of transfected HEK cells were acquired as previously described(Lev et al., 2010).

Data analysisData were analyzed and plotted using pClamp 10 (Molecular Devices) and Sigma Plot8.02 (Systat software) software. Confocal images were imported as tiff single imagesinto Photoshop (Adobe Systems), where they were subsequently cropped and resized.

Statistical analysisStudent’s t-test was used for statistical analysis. All error bars show s.e.m.

This work was supported by the Rita Altura foundation (to G.B.),the Moscona and Minerva Foundations (to B.M.), and by Grants fromthe National Institute of Health (RO1 EY 03529 to B.M.), the IsraelScience Foundation (ISF, to G.B. and B.M.), the US-Israel BinationalScience Foundation (BSF to B.M.), and the German-Israel Foundation(GIF to B.M.). Deposited in PMC for release after 12 months.

Supplementary material available online athttp://jcs.biologists.org/cgi/content/full/123/18/3112/DC1

Jour

nal o

f Cel

l Sci

ence

3124 Journal of Cell Science 123 (18)

ReferencesBach, G. (2001). Mucolipidosis type IV. Mol. Genet. Metab. 73, 197-203.Bach, G. (2005). Mucolipin 1, endocytosis and cation channel-a review. Pflugers Arch.

451, 313-317.Bargal, R., Avidan, N., Olender, T., Ben Asher, E., Zeigler, M., Raas-Rothschild, A.,

Frumkin, A., Ben-Yoseph, O., Friedlender, Y., Lancet, D. et al. (2001). Mucolipidosistype IV: Novel MCOLN1 mutations in Jewish and non-Jewish patients and the frequencyof the disease in the Ashkenazi Jewish population. Hum. Mutat. 17, 397-402.

Cantiello, H. F., Montalbetti, N., Goldmann, W. H., Raychowdhury, M. K., Gonzalez-Perrett, S., Timpanaro, G. A. and Chasan, B. (2005). Cation channel activity ofmucolipin-1: the effect of calcium. Pflugers Arch. 451, 304-312.

Cuajungco, M. P. and Samie, M. A. (2008). The varitint-waddler mouse phenotypes andthe TRPML3 ion channel mutation: cause and consequence. Pflugers Arch. Eur. J.Physiol. 457, 463-473.

Cuervo, A. M. (2004a). Autophagy: in sickness and in health. Trends Cell Biol. 14, 70-77.

Cuervo, A. M. (2004b). Autophagy: many paths to the same end. Mol. Cell. Biochem. 263,55-72.

Curcio-Morelli, C., Zhang, P., Venugopal, B., Charles, F. A., Browning, M. F., Cantiello,H. F. and Slaugenhaupt, S. A. (2010). Functional multimerization of mucolipin channelproteins. J. Cell Physiol. 222, 328-335.

Di Palma, F., Belyantseva, I. A., Kim, H. J., Vogt, T. F., Kachar, B. and Noben-Trauth,K. (2002). Mutations in Mcoln3 associated with deafness and pigmentation defects invaritint-waddler (Va) mice. Proc. Natl. Acad. Sci. USA 99, 14994-14999.

Dong, X. P., Cheng, X., Mills, E., Delling, M., Wang, F., Kurz, T. and Xu, H. (2008).The type IV mucolipidosis-associated protein TRPML1 is an endolysosomal iron releasechannel. Nature 455, 992-996.

Dong, X. P., Wang, X., Shen, D., Chen, S., Liu, M., Wang, Y., Mills, E., Cheng, X.,Delling, M. and Xu, H. (2009). Activating mutations of the TRPML1 channel revealedby proline-scanning mutagenesis. J. Biol. Chem. 284, 32040-32052.

Grimm, C., Cuajungco, M. P., van Aken, A. F., Schnee, M., Jors, S., Kros, C. J., Ricci,A. J. and Heller, S. (2007). A helix-breaking mutation in TRPML3 leads to constitutiveactivity underlying deafness in the varitint-waddler mouse. Proc. Natl. Acad. Sci. USA104, 19583-19588.

Grimm, C., Jors, S. and Heller, S. (2009). Life and death of sensory hair cells expressingconstitutively active TRPML3. J. Biol. Chem. 284, 13823-13831.

Grimm, C., Jors, S., Saldanha, S. A., Obukhov, A. G., Pan, B., Oshima, K., Cuajungco,M. P., Chase, P., Hodder, P. and Heller, S. (2010). Small molecule activators ofTRPML3. Chem. Biol. 17, 135-148.

Ho, S. N., Hunt, H. D., Horton, R. M., Pullen, J. K. and Pease, L. R. (1989). Site-directed mutagenesis by overlap extension using the polymerase chain reaction. Gene77, 51-59.

Karacsonyi, C., Miguel, A. S. and Puertollano, R. (2007). Mucolipin-2 localizes to theArf6-associated pathway and regulates recycling of GPI-APs. Traffic 8, 1404-1414.

Kim, H. J., Li, Q., Tjon-Kon-Sang, S., So, I., Kiselyov, K. and Muallem, S. (2007).Gain-of-function mutation in TRPML3 causes the mouse Varitint-Waddler phenotype.J. Biol. Chem. 282, 36138-36142.

Kim, H. J., Li, Q., Tjon-Kon-Sang, S., So, I., Kiselyov, K., Soyombo, A. A. andMuallem, S. (2008). A novel mode of TRPML3 regulation by extracytosolic pH absentin the varitint-waddler phenotype. EMBO J. 27, 1197-1205.

Kim, H. J., Soyombo, A. A., Tjon-Kon-Sang, S., So, I. and Muallem, S. (2009). TheCa(2+) channel TRPML3 regulates membrane trafficking and autophagy. Traffic 10,1157-1167.

Kiselyov, K., Chen, J., Rbaibi, Y., Oberdick, D., Tjon-Kon-Sang, S., Shcheynikov, N.,Muallem, S. and Soyombo, A. (2005). TRP-ML1 is a lysosomal monovalent cationchannel that undergoes proteolytic cleavage. J. Biol. Chem. 280, 43218-43223.

Klionsky, D. J., Cuervo, A. M. and Seglen, P. O. (2007). Methods for monitoringautophagy from yeast to human. Autophagy 3, 181-206.

LaPlante, J. M., Sun, M., Falardeau, J., Dai, D., Brown, E. M., Slaugenhaupt, S. A.and Vassilev, P. M. (2006). Lysosomal exocytosis is impaired in mucolipidosis type IV.Mol. Genet. Metab. 89, 339-348.

Lev, S., Zeevi, D. A., Frumkin, A., Offen-Glasner, V., Bach, G. and Minke, B. (2010).Constitutive activity of the human TRPML2 channel induces cell degeneration. J. Biol.Chem. 285, 2771-2782.

Manzoni, M., Monti, E., Bresciani, R., Bozzato, A., Barlati, S., Bassi, M. T. andBorsani, G. (2004). Overexpression of wild-type and mutant mucolipin proteins inmammalian cells: effects on the late endocytic compartment organization. FEBS Lett.567, 219-224.

Martina, J. A., Lelouvier, B. and Puertollano, R. (2009). The calcium channel mucolipin-3 is a novel regulator of trafficking along the endosomal pathway. Traffic 10, 1143-1156.

Miedel, M. T., Rbaibi, Y., Guerriero, C. J., Colletti, G., Weixel, K. M., Weisz, O. A.and Kiselyov, K. (2008). Membrane traffic and turnover in TRP-ML1-deficient cells:a revised model for mucolipidosis type IV pathogenesis. J. Exp. Med. 205, 1477-1490.

Nagata, K., Zheng, L., Madathany, T., Castiglioni, A. J., Bartles, J. R. and Garcia-Anoveros, J. (2008). The varitint-waddler (Va) deafness mutation in TRPML3 generatesconstitutive, inward rectifying currents and causes cell degeneration. Proc. Natl. Acad.Sci. USA 105, 353-358.

Piper, R. C. and Luzio, J. P. (2004). CUPpling calcium to lysosomal biogenesis. TrendsCell Biol. 14, 471-473.

Pryor, P. R., Reimann, F., Gribble, F. M. and Luzio, J. P. (2006). Mucolipin-1 is alysosomal membrane protein required for intracellular lactosylceramide traffic. Traffic7, 1388-1398.

Puertollano, R. and Kiselyov, K. (2009). TRPMLs: in sickness and in health. Am. J.Physiol. Renal. Physiol. 296, F1245-F1254.

Raychowdhury, M. K., Gonzalez-Perrett, S., Montalbetti, N., Timpanaro, G. A.,Chasan, B., Goldmann, W. H., Stahl, S., Cooney, A., Goldin, E. and Cantiello, H.F. (2004). Molecular pathophysiology of mucolipidosis type IV: pH dysregulation ofthe mucolipin-1 cation channel. Hum. Mol. Genet. 13, 617-627.

Samie, M. A., Grimm, C., Evans, J. A., Curcio-Morelli, C., Heller, S., Slaugenhaupt, S.A. and Cuajungco, M. P. (2009). The tissue-specific expression of TRPML2 (MCOLN-2) gene is influenced by the presence of TRPML1. Pflugers Arch. 459, 79-91.

Soyombo, A. A., Tjon-Kon-Sang, S., Rbaibi, Y., Bashllari, E., Bisceglia, J., Muallem,S. and Kiselyov, K. (2006). TRP-ML1 regulates lysosomal pH and acidic lysosomallipid hydrolytic activity. J. Biol. Chem. 281, 7294-7301.

Sun, M., Goldin, E., Stahl, S., Falardeau, J. L., Kennedy, J. C., Acierno, J. S., Jr,Bove, C., Kaneski, C. R., Nagle, J., Bromley, M. C. et al. (2000). Mucolipidosis typeIV is caused by mutations in a gene encoding a novel transient receptor potentialchannel. Hum. Mol. Genet. 9, 2471-2478.

Treusch, S., Knuth, S., Slaugenhaupt, S. A., Goldin, E., Grant, B. D. and Fares, H.(2004). Caenorhabditis elegans functional orthologue of human protein h-mucolipin-1is required for lysosome biogenesis. Proc. Natl. Acad. Sci. USA 101, 4483-4488.

Varnai, P., Thyagarajan, B., Rohacs, T. and Balla, T. (2006). Rapidly inducible changesin phosphatidylinositol 4,5-bisphosphate levels influence multiple regulatory functionsof the lipid in intact living cells. J. Cell Biol. 175, 377-382.

Venkatachalam, K. and Montell, C. (2007). TRP channels. Annu. Rev. Biochem. 76, 387-417.

Venkatachalam, K., Hofmann, T. and Montell, C. (2006). Lysosomal localization ofTRPML3 depends on TRPML2 and the mucolipidosis-associated protein TRPML1. J.Biol. Chem. 281, 17517-17527.

Venkatachalam, K., Long, A. A., Elsaesser, R., Nikolaeva, D., Broadie, K. and Montell,C. (2008). Motor deficit in a drosophila model of mucolipidosis type IV due to defectiveclearance of apoptotic cells. Cell 135, 838-851.

Venugopal, B., Browning, M. F., Curcio-Morelli, C., Varro, A., Michaud, N.,Nanthakumar, N., Walkley, S. U., Pickel, J. and Slaugenhaupt, S. A. (2007).Neurologic, gastric, and opthalmologic pathologies in a murine model of Mucolipidosistype IV. Am. J. Hum. Genet. 81, 1070-1083.

Venugopal, B., Mesires, N. T., Kennedy, J. C., Curcio-Morelli, C., LaPlante, J. M.,Dice, J. F. and Slaugenhaupt, S. A. (2009). Chaperone-mediated autophagy is defectivein mucolipidosis Type IV. J. Cell. Physiol. 219, 344-353.

Vergarajauregui, S. and Puertollano, R. (2006). Two di-leucine motifs regulate traffickingof mucolipin-1 to lysosomes. Traffic 7, 337-353.

Vergarajauregui, S., Connelly, P. S., Daniels, M. P. and Puertollano, R. (2008). Autophagicdysfunction in mucolipidosis type IV patients. Hum. Mol. Genet. 17, 2723-2737.

Xu, H., Delling, M., Li, L., Dong, X. and Clapham, D. E. (2007). Activating mutationin a mucolipin transient receptor potential channel leads to melanocyte loss in varitint-waddler mice. Proc. Natl. Acad. Sci. USA 104, 18321-18326.

Yorimitsu, T. and Klionsky, D. J. (2005). Autophagy: molecular machinery for self-eating. Cell Death Differ. 12, 1542-1552.

Zeevi, D. A., Frumkin, A. and Bach, G. (2007). TRPML and lysosomal function.Biochim. Biophys. Acta 1772, 851-858.

Zeevi, D. A., Frumkin, A., Offen-Glasner, V., Kogot-Levin, A. and Bach, G. (2009). Apotentially dynamic lysosomal role for the endogenous TRPML proteins. J. Pathol.219, 153-162.

Jour

nal o

f Cel

l Sci

ence