TMEM120A/TACAN: A putative regulator of ion channels, mechanosensation, and lipid metabolism

ABSTRACT

TMEM120A (TACAN) is an enigmatic protein with several seemingly unconnected functions. It was proposed to be an ion channel involved in sensing mechanical stimuli, and knockdown/knockout experiments have implicated that TMEM120A may be necessary for sensing mechanical pain. TMEM120A’s ion channel function has subsequently been challenged, as attempts to replicate electrophysiological experiments have largely been unsuccessful. Several cryo-EM structures revealed TMEM120A is structurally homologous to a lipid modifying enzyme called Elongation of Very Long Chain Fatty Acids 7 (ELOVL7). Although TMEM120A’s channel function is debated, it still seems to affect mechanosensation by inhibiting PIEZO2 channels and by modifying tactile pain responses in animal models. TMEM120A was also shown to inhibit polycystin-2 (PKD2) channels through direct physical interaction. Additionally, TMEM120A has been implicated in adipocyte regulation and in innate immune response against Zika virus. The way TMEM120A is proposed to alter each of these processes ranges from regulating gene expression, acting as a lipid modifying enzyme, and controlling subcellular localization of other proteins through direct binding. Here, we examine TMEM120A’s structure and proposed functions in diverse physiological contexts.

KEYWORDS:

• PIEZO2
• TMEM120A
• TACAN
• ion channel regulation
• mechanosensation

Introduction

Dorsal root ganglion (DRG) neurons are responsible for sensing and encoding somatosensory stimuli into electrical signals through specialized ion channels. Among these channels are those that are mechanically activated and respond to tactile input [1]. Famously, PIEZO2 was the first depolarizing mechanosensitive channel identified in DRG neurons [1]. It is necessary for light touch sensing [2,3], proprioception (sensing limb position in space) [4], and mechanical allodynia (an inflammation-induced event where innocuous touch is perceived as painful) [5,6]. When Piezo2 is knocked down in DRG neurons, the proportion of cells with rapidly inactivating currents is strongly reduced [1]. However, many neurons still respond to mechanical stimuli with currents displaying two other unique inactivating kinetics (intermediate and slow) implying other mechanosensitive channels are present [1,7].

The channel(s) contributing to mechanically induced slowly inactivating currents and those involved in sensing noxious tactile pain remain a challenging mystery to identify [7]. Several candidates have been proposed to be the slowly inactivating mechanosensor including TMEM150C (Tentonin3) [8] and TMEM120A (TACAN) [9]. However, these hypotheses have been controversial.

Mechanically-evoked currents from TMEM150C could not be reproduced in cells lacking expression of endogenous mechanosensitive channels, i.e. HEK293 cells with a Piezo1 knockout (HEK293-P1KO) [10,11] and its role in sensory neurons have also been debated [7,12]. On the other hand, expressing TMEM150C prolonged the inactivation kinetics of three known mechanically-activated channels: PIEZO1, PIEZO2, and TREK-1 [10,11]. Because TMEM150C appears to be a general regulator of these channels [10,11] through direct protein–protein interaction [11], using cell lines with endogenous mechanosensitive channels may have led to confounding results indicating TMEM150C acts as an ion channel responding to mechanical stimuli [8].

TMEM120A’s proposed ion channel function has also been challenged predominantly due to several unsuccessful attempts to reproduce electrophysiological experiments [7,13–18]. However, debating TMEM120A’s function has put the spotlight on a seemingly do-it-all protein linked to lipid metabolism [19–21], antiviral function via regulation of Stimulation of Interferon Genes (STING) [22], inhibition of ion channels (PIEZO2) [13] & Polycystin-2 (PKD2) [17], and regulation of mechanical pain [23–26]. Accompanying all these apparent effects, TMEM120A is highly expressed in diverse cell types e.g. brown and white adipose tissue, the kidneys, the colon [19], trigeminal ganglion neurons [26], and DRG neurons [7,9,13]. Many important research questions across body systems linger due to a lack of mechanistic understanding of TMEM120A’s true function. In this review, we will discuss the electrophysiological analysis of TMEM120A, its structure, and the physiological processes that it has been implicated in.

Electrophysiological analysis of TMEM120A

In 2020, TMEM120A was dubbed “TACAN” and proposed to be an ion channel involved in sensing high-threshold mechanical stimuli with slowly inactivating kinetics [9]. This proposal was based on the following experimental data: Overexpression of TMEM120A in heterologous systems (CHO, HEK293, and HEK293-P1KO cells) increased stretch-induced currents above baseline in the range of a single- or sub-picoampere at high-thresholds. TMEM120A’s closely related homolog (TMEM120B) which shares 69% sequence identity and has a “nearly identical” structure to TMEM120A [18] did not increase stretch-activated currents in the same systems [9]. Substrate deflection using micropillar arrays in cells overexpressing TMEM120A elicited macroscopic currents. However, membrane indentation using a blunt glass probe, a standard technique studying PIEZO channels, was insufficient to generate currents above baseline, which was attributed to cell rupture prior to reaching a hypothesized high-threshold. Knockdown of Tmem120A in mouse DRG neurons reduced the proportion of those displaying slowly inactivating mechanically-activated currents evoked by membrane indentation. And finally, insertion of recombinant TMEM120A into planar lipid bilayers generated ionic currents [9]. In this section, we discuss these experiments and subsequent studies which largely challenged TMEM120A’s ion channel function.

Beginning with membrane stretch as a stimulus, applied as negative pressure through the patch pipette in cell attached patch clamp measurements, six studies indicated that they were unable to detect currents above baseline in cells expressing TMEM120A [13–18] see also Figure 1. Fully quantified data for all stimuli was reported for four of these studies [13,16–18] using PIEZO1 as a positive control. In two cases, no data quantification was provided and only representative traces were presented [14,15]. Ke. et al. reported no significant difference in the average current amplitudes between HEK293T cells transfected with TMEM120A and vector-transfected cells, but did observe some small currents in 4 out of 10 TMEM120A expressing cells [18]. The authors discuss these currents and indicate they are likely leak that was induced by high-threshold negative pressure. These currents did not proportionally respond to increased stimuli, and they persisted after stimulation was relieved [18]. This leak is consistent with other studies where it was observed in both vector-only, and TMEM120A expressing cells [13,16].

Figure 1. TMEM120A inhibits PIEZO2 but not PIEZO1 channels [13]. (a) N2A-Piezo1-Knockout cells were transfected with PIEZO2-GFP ± TMEM120A-tdTomato, and patched in a whole-cell configuration (−60 mV). Peak PIEZO2-GFP currents in the absence (black) or presence (red) of TMEM120A over increasing membrane indentation depths. (b) Peak current amplitudes at 6.4 µm (Mann-Whitney, Mean ± SEM). (c) Representative traces. (d) N2A-Piezo1-Knockout cells transfected with PIEZO1-GFP ± TMEM120A-tdTomato, TMEM120A-tdTomato, or tdTomato patched in a cell-attached configuration (−80 mV). Peak currents with increasing negative pressure. (e) Peak current amplitudes at −55 mmHg (Mann-Whitney, Mean ± SEM). (f) Representative traces. Data from Del Rosario, J & Gabrielle M, et. al., JGP 2022 [13].

Another study using COS-7 cells did observe stretch-activated currents from a TMEM120A mutant (M207A) [27]. Currents from this mutant were sixfold greater than wild-type & mock transfected cells, which had similar mean steady-state currents in the single-digit picoampere range [27]. However, no statistical analysis was performed comparing GFP-only, TMEM120A, or TMEM120A-M207A groups to show if these differences were statistically significant [27]. How this mutant affects TMEM120A’s structure will be discussed in greater detail in the next section.

An alternative mechanical stimulation in the whole-cell configuration uses a blunt glass probe to indent the cell’s membrane which is routinely used to activate PIEZO1 and PIEZO2 channels [1]. However, TMEM120A currents could not be elicited via membrane indentation in heterologous systems [9,16,18]. This was proposed to be due to cell rupture prior to reaching the hypothesized necessary high-threshold [9].

Beaulieu, et al. performed siRNA knockdown of Tmem120A in DRG neurons from mice expressing tdTOMATO in Trpv1 positive cells [9]. Trpv1 is a marker for small diameter C-fibers and these cells are considered nociceptors [28]. Whole-cell experiments with blunt glass probe indentations were performed using mouse DRG neurons due to their capacity to withstand greater stimuli. The proportion of these neurons displaying slowly inactivating currents decreased in the Tmem120A knockdown group to 1 out of 31 neurons from 6 out of 21 neurons in cells transfected with non-targeting siRNA [9]. Two follow-up studies using Tmem120A knockdowns in DRG neurons from wild-type mice, however, were unable to detect a change in proportion of slowly inactivating current [7,13]. Furthermore, knocking down Tmem120A did not change the proportion of slowly inactivating kinetics in IB4 positive DRG neurons [7] (non-peptidergic C-fibers) [28].

Beaulieu, et al. performed additional experiments using these Trpv1 positive DRG neurons for cell-attached patch clamp recordings. They reported that mechanical currents in the cell-attached mode in response to negative pressure decreased in Tmem120A knockdown neurons compared to wild type, which had similar sub-picoampere amplitudes to previous overexpression experiments [9]. There has been no independent effort to replicate this experiment yet.

Micropillar array deflection was employed as another mechanical stimulation modality in an attempt to record whole-cell TMEM120A currents in HEK293-P1KO cells [9]. In this technique, cells are cultured atop micropillars, and a blunt glass probe is used to strike these pillars resulting in a mechanical stimulus via substrate deflection [29,30]. TMEM120A-expressing cells (n = 6) displayed large, but variable macroscopic currents using this technique while mock transfected cells did not (n = 9) [9]. However, the mechanically activated currents did not return to baseline (did not deactivate) after cessation of the mechanical stimuli [9]. It is unclear why these currents do not deactivate, but TRPV4 currents with larger stimuli have been reported also not to deactivate (although they appear to do so at smaller deflections) [30]. Additionally, slowly inactivating currents from DRG neurons measured using micropillar array deflection appear to also deactivate [29]. This experiment unlike those using membrane stretch as a stimulus has not yet been independently performed and published.

To differentiate between TMEM120A being an ion channels, versus an accessory subunit, or ion channel modulator, Beaulieu et al. incorporated the purified TMEM120A protein into planar lipid bilayers and observed unitary single-channel currents that were inhibited by two blockers of mechanically activated channels, GdCl₃ and GsMTx4 [9]. While these data suggest that the TMEM120A protein itself is the pore forming subunit, the unitary conductance of the currents in lipid bilayers were ~ 25-fold higher than those attributed to TMEM120A in cell-attached patches. This is concerning, as the conductance of ion channels in planar lipid bilayers is usually comparable in lipid bilayers and cells [31,32]. It has to be noted, however, that a ~ threefold larger channel conductance was also reported for PIEZO1 currents in lipid bilayers compared to that in cells [33]. Channel activity was observed in the absence of any mechanical stimulus in planar lipid bilayers, which was interpreted by high resting tension in the lipid bilayers, due to their large curvature diameter [9].

Subsequent publications attempted to reconstitute the purified TMEM120A into either lipid bilayers, or lipid vesicles to assess its potential ion channel function [14–16]. Niu, et al. performed similar experiments observing transient currents only when TMEM120A was inserted at a high protein-to-lipid ratio [14]. The authors indicate that this is more consistent with disruptions to the lipid bilayer than channel function [14]. These perturbations associated with high protein-to-lipid ratios are also observed in GUVs and proteoliposomes, which is discussed later in this section.

Artificial lipid bilayers can present several challenges in performing and interpreting experiments [34]. Synthetic membranes can become permeable to ions and water when near phase transition resulting in a phenomenon dubbed “lipid ion channels” [34]. Experimental manipulations to bilayers can adjust their melting point affecting their proximity to phase transition. Perturbating factors affecting thermodynamic properties of synthetic lipid bilayers are voltage changes, temperature, drugs targeted to the membrane, protein insertion, etc [34–37]. For example, a synthetic membrane consisting of DMPC:DLPC (10:1) at 20.6º C and +200 mV with no protein inserted can generate single-channel recordings nearly indistinguishable from those of TRPM2 and TRPM8 [37]. Insertion of non-channel proteins can also shift the melting point of synthetic membranes in a dose-dependent manner e.g. cytochrome b₅ and band 3 protein [34]. In our opinion, additional experiments are necessary to assess whether TMEM120A is mediating ion flux in these synthetic membranes or if insertion of recombinant protein at a high ratio compared to lipids generates nonspecific channel-like currents.

Additionally, Beaulieu et al. observed a robust decrease in current from synthetic bilayers containing TMEM120A when GsMTx4 was applied [9]. GsMTx4 is an amphipathic peptide from spider toxin which acts as an inhibitor of mechanosensitive ion channels [38]. Interestingly, the peptide was reported to have no effect [38] or modestly potentiated [39] potassium selective TREK-1 channels. GsMTx4 does not act as a pore blocker but rather enters the membrane in a tension-dependent fashion, and effectively makes force transfer from mechanical stimuli on the membrane to the channels less efficient [39]. Once again, applying macromolecules to synthetic bilayers can be difficult to interpret as amphipathic peptide insertion can alter their thermodynamic properties [35]. It has been reported that anesthetics reducing the melting point of membranes can dose-dependently decrease the current from “lipid ion channel” events in synthetic bilayers where no channel proteins are inserted e.g. appearing to inhibit a channel when none are present [36]. Without additional experimentation including calorimetry of the synthetic bilayers containing TMEM120A and GsMTx4, it is difficult to draw conclusions regarding channel activity, or inhibition if artifacts are present due to protein insertion. Another potentially useful avenue to make these lipid bilayer experiments more informative could be adjusting solvents to an imbalance inducing osmotic tension events [33] in combination with GsMTx4 application.

While direct mechanical force cannot be applied to planar lipid bilayers, other artificial membranes were employed to test TMEM120A’s mechanosensitivity. Purified recombinant TMEM120A was inserted into giant unilamellar vesicles (GUVs) and proteoliposomes [14–16]. At high protein-to-lipid concentrations, transient currents were again detected but when this ratio was lowered there was no current [14,15]. The authors again point to this as leak associated with high concentrations of protein being inserted into these membranes. When negative pressure was applied to GUVs and proteoliposomes with TMEM120A, no change in currents were detected [15,16]. For example, Rong, et al. observed only 2 GUV patches out of 217 displaying stretch-induced mini-conductance channel activities [16]. The authors report that only one of these patches responded at both −80 and +80 mV and the other only at −80 mV [16].

Overall, the majority of the electrophysiological data published after the original publication [9] do not support that TMEM120A is an ion channel involved in sensing mechanical stimuli [7,13–18]. One publication, however, showed increased currents in cells transfected with a mutant of TMEM120A, and suggested the presence of a potential ion permeation pathway based on molecular dynamics simulation [27].

Can we say anything definitive about the ion channel function of TMEM120A? In the authors’ opinion, based on all the current data, it is unlikely that TMEM120A is an ion channel, but it cannot be fully excluded. Science progresses by convergence of evidence from independent observations. After the cloning of Piezo channels in 2010, such converging evidence emerged very quickly, and those channels became well-established mechanically activated ion channels playing clear roles in physiological processes such as light touch and proprioception without much controversy [40]. For the mechanosensory channels responsible for hearing, the process unfolded much slower. TMC1 was identified in 2003 as a transmembrane protein, mutations in which cause deafness in humans, and the idea that it may be the mechanosensory channel responsible for hearing, was immediately proposed [41]. It took almost two decades of pain-staking research and many arguments and counterarguments for TMC1 and TMC2 to eventually become accepted as the mechanotransduction channel in the inner ear, mainly because of the highly complicated machinery requiring six additional proteins for the channel complex to function [42]. In our opinion, demonstrating that TMEM120A is an ion channel would require substantial future work, that includes finding an experimental system in which robust and reproducible currents can be evoked and/or identifying potential additional subunits required to confer mechanosensitivity.

TMEM120A’s structure & similarities to ELOVL7

Five cryogenic electron-microscopy (cryo-EM) structures of TMEM120A have recently been published [14–16,18,27]. Each structure indicates TMEM120A forms symmetrical dimers comprised of six transmembrane domains per protomer (Figure 2A). Furthermore, each subunit contains two distinct regions: a coiled-coil domain in the N-termini, and the transmembrane domains (TMD) in the C-termini. The N-terminal coiled-coil domain has been predicted to be involved in dimerization which the cryo-EM structures [27] and truncation experiments support [17]. Interestingly, the C-terminal TMD is comprised of an α-helical barrel with a deep pocket/tunnel and four groups found that it is only open to one end [14–16,18]. Most structures identified coenzyme-A (CoA) binding in this region [14–16,18] (Figure 2B) while one indicated this density was cholesterol [27]. TMEM120A’s structure does not bear similarity to any known ion channel structure, but it has structural homology with a lipid modifying enzyme called Elongation of Very Long Chain Fatty Acids 7 (ELOVL7) [14,15,18] (Figure 2C-E). Of note, proteins lacking resemblance to known ion channels does not rule out the possibility they are one. The structure of the Cystic Fibrosis Transmembrane Conductance Regulator (CFTR) channel for example, is not similar to other ion channels, as it resembles a family of transporters (ATP-Bound Cassette Transporters) [44].

Figure 2. The structure of human TMEM120A containing one CoA molecule per protomer [16]. (a) Front view. Each protomer labeled either orange or green, and CoA in cyan. (b) Expanded view of box from (A) showing side chains of conserved residues that are catalytically important to ELOVL7: W193 (purple), H196 (red), H197 (yellow). (c) Front view of human TMEM120A protomer. (d) Front view of human ELOVL7. (e) Merged view of TMEM120A and ELOVL7. (f) Top view of TMEM120A protomer with putative constriction sites: M207 (green), W210 (blue), F217 (red), C310 (purple). (g) Bottom view of TMEM120A protomer with CoA (cyan). Images were generated using PyMOL from publicly available pdb files. TMEM120A structure from Rong, Y. et. al., eLife. 2021 [16], PDB ID: 7F3T. ELOLV7 structure from Nie, L. et. al., Nature Structure and Molecular Biology. 2021 [43], PDB ID: 6Y7F.

TMEM120A’s TMD α-helical barrels leave a few lingering questions. As previously stated, without manipulation, these spaces containing a CoA (or cholesterol) are open on one side and sealed on the other (presumed intracellular and extracellular facing respectively). But what may happen when TMEM120A is influenced by lipid depletion or stretch? A TMEM120A structure free of CoA revealed these α-helical barrels are largely empty with the possibility to allow for water molecules, or ions to enter from intracellular space but not exit due to a constriction of this space by four amino acids (M207, W210, F219 & C310) at the extracellular side [16] (Figure 2F). This study also notes that this space is blocked on the intracellular side by CoA when it is bound to the protein [16] (Figure 2G).

Molecular Dynamics (MD) simulations with membrane tension applied have been used to identify ion conducting pathways of novel mechanosensitive ion channels, e.g., the OSCAs [45]. MD simulations modeling TMEM120A under 35 mN/m tension predicts that water molecules can pass through the α-helical barrels from an extracellular to intracellular space [27]. This hypothetical ion conduction pathway is also reported to be restricted by three amino acids (N165, M207 & F223) undergoing conformational changes in MD simulations when tension is applied [27]. Although M207 seems to be a common restrictor of this space, it is not highly conserved across species [16]. Mutating M207 to a smaller alanine increased the pore radius in MD simulations under 35 mN/m tension [27]. The M207A-TMEM120A mutant also appeared to generate stretch-induced currents in cell-attached electrophysiology experiments [27]. Of note, the structure used for MD simulations was the only one identified with cholesterol bound instead of CoA [27]. It is unclear whether the putatively identified ion conducting pathway is present in MD simulations when TMEM120A is bound to CoA or cholesterol since CoA appears to block water molecules from entering intracellulary [16]. Whether TMEM120A could potentially act as a channel operating through a novel mechanism involving lipid enrichment and depletion is unknown. This would require significant structural and electrophysiological investigation.

Another hypothesized role for TMEM120A’s TMD α-helical barrels may be a space for enzymatic activity given its structural homology to ELOVL7 [14,15,18]. This enzyme binds CoA lipids and facilitates the first and rate limiting reaction of very long-chain fatty acid elongation [46]. Like TMEM120A, ELOVL7 contains a 6 TMD α-helical barrel where CoA binds [43]. These proteins share little sequence homology, but TMEM120A has two of the four histidine residues conserved with ELOVL7 that are catalytically important and interact with CoA [43] (Figure 2B). Mutating these residues (H196A & H197A) in TMEM120A shifts its ability to interact with larger CoA species (S-ethyl-CoA & acetyl-CoA) [43]. Alternatively, TMEM120A contains a tryptophan (W193A) in place of a third catalytically important histidine present in ELOVL7 (Figure 2B). This tryptophan has a π -π interaction with CoA and when mutated to an alanine significantly reduces the proteins binding affinity for CoA [16].

TMEM120A did not yield any enzymatic activity when it was assayed for ELOVL7’s reaction indicating a hypothetical substrate has not yet been identified [14]. ELOVL enzymes elongate long-chain acyl-CoAs through a condensation reaction [47]. ELOVL7 specifically elongates C16-C20 acyl-CoAs with highest activity measured using polyunsaturated C18-CoAs [46,47]. Although the highest activity was measured for polyunsaturated CoAs, using saturated CoAs was also sufficient for ELOVL7 activity [47]. Some ELOVLs do have substrate specificity for the degree of saturation as well as acyl-chain length [47]. Assaying ELOVL7’s reaction required that the protein was embedded in a membrane (microsomes from HEK293 cells [47,48], or proteoliposomes [46]) as no enzymatic activity was detected using solubilized protein [46]. Recombinant TMEM120A in proteoliposomes was assayed using a fluorescence-based enzyme-coupled reaction with saturated C18-CoA (stearoyl) as a substate [14]. No enzyme activity was detected for TMEM120A but ELOVL7 activity was present using the same method [14]. Additional substrates with various acyl-chain lengths and degrees of unsaturation would need to be screened to determine whether TMEM120A can facilitate an enzymatic reaction similar to ELOVLs.

TMEM120A regulates lipid metabolism

Beyond hypothesizing that TMEM120A can alter lipid metabolism based on structural similarities to an enzyme, it has been implicated in lipid storage and adipocyte development [19–21]. TMEM120A and TMEM120B (a closely related homolog) are both highly expressed in white and brown adipose tissue [19]. A knockdown of Tmem120A, Tmem120B, or both in premature adipocytes (3T3-L1 and OP9 cells) prevents triglyceride accumulation and gene expression indicative of maturation [19,21]. Reintroducing TMEM120A to knockdown cells fully rescues the phenotype, while TMEM120B has only a partial effect [19]. TMEM120A upregulation can also be induced in white adipose tissue via cold exposure [49], and by rosiglitazone (an agonist of the transcription factor PPARγ) in 3T3-L1 cells [50]. Furthermore, a murine model with a Tmem120A conditional knockout in mature adipocytes (Tmem120A-cKO) yielded a lipodystrophy phenotype [20]. The Tmem120A-cKO mice on a high-fat diet did not incur a similar increase in body weight as wild-type littermates, which was driven by excess fat-mass [20]. Female mice had a stronger phenotype than male mice which may be explained by females expressing fourfold the amount of TMEM120A in white adipose tissue [20]. These reports indicate TMEM120A is necessary for the development and maintenance of proper lipid metabolism in adipocytes.

How TMEM120A mechanistically affects lipids may depend on its debated sub-cellular localization. One group identified TMEM120A in the nuclear envelope and the endoplasmic reticulum (ER) [51], two found it in the ER [21,22], and three more claimed it was localized in the plasma membrane [9,17,24]. Those finding TMEM120A in the nuclear envelope identified altered gene expression of key regulators for adipocyte maturation [19] and metabolism [20] when it was knocked down or out, respectively. They also identified gene expression changes in heterologous systems where TMEM120A or other nuclear envelope transmembrane proteins were overexpressed [52]. The hypothesized mechanism is that TMEM120A affects the position of genetic material within the nucleus by interacting with lamins similar to other nuclear envelope proteins [20]. The location of several genes within the nucleus (labeled by fluorescent in situ hybridization) was different (closer or farther) from the nuclear envelope in white adipose tissue of Tmem120a-cKO mice compared to wild-type littermates [20].

Whether TMEM120A directly prevents mispositioning of genetic material for proper expression may be a source of debate given its ability to cause similar changes from the ER [22]. TMEM120A seemingly promotes protective gene expression changes during the innate immune response against Zika virus infection (identified in a gain-of-function screen) [22,53]. TMEM120A’s C-terminal domain physically interacts with Stimulator of Interferon Genes (STING) in the ER promoting translocation to the ER-Golgi intermediate complex by way of COPII-coated vesicles [22]. STING can then facilitate the activation of transcription factors [22]. Regardless of TMEM120A’s subcellular location, it seems to have the capacity to alter gene expression.

Another hypothesis is that TMEM120A may directly produce excess lipids that are necessary building blocks in triglyceride synthesis [21]. Both TMEM120 (C. elegans homolog) and mouse TMEM120A overexpressed in COS-7 cells were identified to be sufficient for lipid droplet formation and triglyceride production [21]. This hypothesis is in line with a potential enzymatic action for TMEM120A similar to ELOVL7. In the case of ELOVL7, it facilitates the production of very long-chain CoA lipids contributing to triglyceride synthesis [46]. Interestingly, it was also postulated that TMEM120A may be promoting STING’s translocation by generating lipids necessary to produce COPII-coated vesicles during the innate immune response [53]. Whether TMEM120A regulates lipid metabolism and innate immunity by directly altering gene expression, enzymatic activity, or both remain unknown.

TMEM120A negatively regulates PIEZO2 & PKD2 channels

We reported that TMEM120A negatively regulates PIEZO2 channels [13]. Co-expressing TMEM120A with PIEZO2 in HEK293 or N2A-Piezo1 knockout (N2A-P1KO) cells, robustly decreased PIEZO2 peak current amplitudes and the number of cells responding to blunt probe indentation [13] (Figure 1A and B). HEK293 cells expressing PIEZO2 & TMEM120A that did have mechanically activated currents required greater indentation depths for activation (increased threshold) [13]. We then tested whether TMEM120A may have similar effects on other mechanically activated ion channels. We did not observe any alterations in PIEZO1 channel activity in the presence of TMEM120A in the whole-cell configuration (membrane deflection) or during cell-attached recordings (negative pressure membrane stretch) (Figure 1C and D) [13]. This was also true for the potassium selective TREK-1 channel which did not have a significant difference in peak current with or without TMEM120A [13]. Curiously, TMEM120B which is a closely related homolog of TMEM120A sharing 69% sequence identity & “nearly identical structures” [18], did not inhibit PIEZO2 when co-expressed [13]. Although TMEM120A and TMEM120B are structurally similar, their functions may not be identical in ion channel regulation [13] or adipocyte maturation [19].

An siRNA knockdown of Tmem120A in isolated mouse DRG neurons increased PIEZO2 mediated rapidly adapting current amplitudes by ~ 50% compared to sham-knockdown [13]. The mechanical threshold to elicit these currents was also significantly reduced in Tmem120a-KD neurons indicating disinhibition [13]. The effect of endogenous TMEM120A is smaller than in heterologous systems suggesting the amount of expression may be important. Using RNAscope to fluorescently label Piezo2 and Tmem120A transcripts, we found that > 95% of Piezo2 positive DRG neurons also expressed Tmem120A [13]. We detected distinct populations of neurons with high levels of Piezo2 & low levels Tmem120A and vice versa [13]. Low threshold mechanoreceptors (tyrosine hydroxylase positive neurons) had a relatively large population of high Piezo2 & low Tmem120A neurons while nociceptors (Trpv1 positive neurons) displayed the opposite [13]. Functionally, low threshold mechanoreceptors readily respond to mechanical stimuli while Trpv1 positive nociceptors are not dedicated to tactile sensing, have high mechanical thresholds, and limited response to mechanical stimulation [28,54]. This finding is consistent with another group who performed single-cell RNA sequencing on DRG neurons collected after patch clamp [55]. Neurons with PIEZO2 mediated rapidly adapting current had relatively high levels of Piezo2 and low levels of Tmem120A mRNA while mechanically insensitive neurons had the opposite expression pattern [55].

Of note, another group that knocked down Tmem120A in mouse DRG neurons did not observe an increase in maximum PIEZO2 mediated currents (regardless of stimulation strength) but did see a non-significant increase in this metric when only using IB4+ neurons [7]. Mechanical threshold and current amplitudes at individual membrane indentation depths for RA type currents were not reported [7]. In our hands, we observed a more pronounced effect at smaller indentation depths between Tmem120A knockdown and sham-siRNA treated DRGs [13]. Neither group observed a change in the overall proportion of RA currents [7,13]. Taken together, this indicates that endogenous TMEM120A at basal expression levels is sufficient to modify the threshold required to activate PIEZO2 without substantially compromising its function.

How TMEM120A negatively regulates PIEZO2 channels remains unclear. We did not detect a significant difference in PIEZO2 cell surface localization with or without TMEM120A co-expression using total internal reflection fluorescent (TIRF) microscopy [13]. It has been reported that TMEM120A inhibits PKD2 channels through physical interaction [17]. TMEM150C also directly interacts with mechanosensitive channels (e.g. PIEZO1&2) prolonging their inactivation kinetics [11]. However, we did not detect any substantial colocalization (Pearson Correlation Coefficient) of fluorescently labeled TMEM120A & PIEZO2 in cells overexpressing these constructs which makes physical interaction being the mechanism of channel inhibition unlikely but does not necessarily rule it out [13]. Furthermore, we did not observe any considerable alterations of cortical actin or microtubules [13] which can modulate mechanosensitive channel activity [56,57].

Although no clear mechanism for TMEM120A mediated PIEZO2 inhibition has been resolved, several intriguing possibilities remain including TMEM120A’s potential lipid modifying function. Several lipids have been identified to regulate PIEZO1 and PIEZO2’s activity (polyunsaturated fatty acids [57,58], margaric acid [57–59], PI(4,5)P₂ [54], PI(3,5)P₂ [60], sphingomyelin [61], etc.). Supplementing cells with margaric acid, a 17-carbon saturated fatty acid, robustly inhibits PIEZO1 [58,59] and PIEZO2 [57]. However, upregulation of margaric acid by TMEM120A is unlikely given that this lipid more efficiently inhibits PIEZO1 [57,59] and TMEM120A had no effect on that channel [13]. Polyunsaturated fatty acids finetune both channels’ inactivation kinetics also making them unlikely mediators of the effect of TMEM120A [57,58,62,63]. However, the idea of a membrane-bound lipid modifying enzyme modulating the function of a mechanosensitive ion channel is not novel. Sphingomyelinase (localized in the plasma membrane) was an identified of as a key regulator of PIEZO1 gating in vascular tissue necessary for blood flow sensing [61]. Alternatively, TMEM120A’s apparent ability to regulate gene expression could play a role in our observed PIEZO2 inhibition [52]. MAPK activity is necessary for PIEZO2 sensitization by Gi-coupled receptor activation [64] implying the expression of some genes can affect channel activity. Potentially, there may be more direct regulators of PIEZO2 requiring upstream signaling by TMEM120A to induce their expression. How exactly TMEM120A affects PIEZO2 remains a mystery but its ability to specifically inhibit PIEZO2 implies that elucidating this mechanism could reveal a specific factor negatively regulating this ion channel.

As previously mentioned, TMEM120A has been reported to negatively regulate PKD2 channels [17]. The association between PKD2 and TMEM120A was first identified in a proteomics screen looking for proteins that physically interact with PKD1&2 [65]. Functionally, this interaction with TMEM120A reduces the PKD2 channel activity [17]. TMEM120A co-expression was sufficient to inhibit wild-type PKD2 measured using single-channel recordings, and a constitutively active PKD2 mutant (F604P) in the whole-cell configuration [17]. The mutant PKD2 was used for whole-cell recordings since wild-type channels insufficiently express at the plasma membrane in heterologous systems to detect current in this mode [17]. The authors performed a series of truncation experiments which indicated that TMEM120A’s S6 transmembrane segment interacts with the PKD2’s transmembrane regions resulting in channel inhibition [17]. TMEM120A’s S1 segment also interacts with the pore domain of PKD2 but this does not affect channel activity [17].

Co-expressing TMEM120A with PKD2 also seemed to affect the mechanosensitivity of CHO cells [17]. It has been reported that PKD2 in complex with PKD1 increases intracellular calcium content when a fluid shear stress stimulation is applied [66]. CHO cells expressing only PKD2 or TMEM120A did not alter the open probability or single-channel amplitudes when negative pressure was applied through the patch pipette [17]. However, cells co-expressing PKD2 and TMEM120A did appear to have increasing open probability and single-channel amplitudes upon increasing negative pressure stimuli [17]. Of note, the authors did not report any mechanically induced currents from CHO cells that were non-transfected or expressed GFP [17]. Other reports indicate that CHO cells do produce some endogenous mechanically activated currents [9].

Endogenous PKD2 and TMEM120A were identified to be co-expressed in three renal cell lines in their primary cilia (Madin-Darby canine kidney, inner medullary collecting ducts, and Lilly Laboratories Culture-Porcine Kidney 1) [17]. Loss of function mutations of PKD2 result in a disease state called Autosomal Dominant Polycystic Kidney Disease (ADPKD) where cysts accumulate in the kidneys [17]. The authors asked the question of whether TMEM120A overexpression would be sufficient to produce an ADPKD like phenotype in zebrafish [17]. Even though TMEM120A and PKD2 co-expression in the same cells of zebrafish was not shown, the combination of a partial Pkd2 knockdown and Tmem120A overexpression in zebrafish lead to cyst generation, and tail curling (an indication of PKD2 loss of function) [17]. This experiment indicates that manipulating expression levels of these two proteins in vivo results in a PKD2-dependent disease state. However, it does not implicate TMEM120A in ADPKD and a physiological role for its ability to negatively regulate PKD2 remains undetermined.

Another relevant aspect of this study is the determination that the N-terminus of TMEM120A is necessary for oligomerization [17]. Expressing a truncated peptide consisting only of TMEM120A’s N-terminus (K135X) with full-length TMEM120A, and PKD2 reduced the interaction between the two proteins and no channel inhibition was observed [17]. The rationale for this effect is that TMEM120A has a greater affinity for the K135X peptide than PKD2 thus reducing the amount of full-length protein available for the inhibitory interaction. Whether using this truncated peptide in other contexts to manipulate TMEM120A’s function has yet to be demonstrated.

TMEM120A affects mechanical pain behavior

TMEM120A’s ability to inhibit PIEZO2 channels indicates it can affect mechanosensation even if it is unlikely to be a channel itself, and there are several reports that it somehow contributes to mechanical pain sensing. TMEM120A is highly expressed in DRG neurons innervating the skin which respond to somatosensory stimuli including those that are painful. Multiple animal models reducing TMEM120A’s expression appear to reduce pain behaviors in response to noxious mechanical stimuli. A conditional knockout of Tmem120A in mouse MrgPrd-positive DRG neurons (small diameter C-fibers involved in noxious mechanical pain sensing [28]) resulted in a reduced nocifensive response defined by paw withdrawal and licking upon high force von Frey filament stimulation [9]. A similar reduction in nocifensive responses was observed using a broad knockdown of Tmem120A by injecting a sciatic nerve bundle with shRNA in AAV2/6 viral particles [9]. In confirmation of the nonspecific knockdown, sciatic nerve bundles of Tmem120A-flox mice were injected with AAV2/6 viral particles containing Cre recombinase with a synapsin promoter [9]. Like the other two modes of reducing TMEM120A expression, this broad conditional knockout in DRG neurons also reduced nocifensive responses in these mice [9]. Another group reported that knocking down Tmem120A in rats by intrathecal injection of antisense oligodeoxynucleotides raised the mechanical nociceptive threshold using the Randall-Selitto paw withdrawal test [23].

Another important question is whether TMEM120A has a similar effect during inflammatory and neuropathic mechanical pain. This includes hyperalgesia where the response to noxious stimuli is heightened, and allodynia where innocuous stimuli elicit pain behavior. TMEM120A’s involvement in neuropathic and inflammatory mechanical hyperalgesia was assessed in rats by knockdown using the intrathecal injection of antisense oligodeoxynucleotides [23]. Mechanical nociceptive threshold was quantified using the Randall-Selitto paw withdrawal test [23]. Tmem120A knockdown rats had significantly attenuated pain responses when various proinflammatory mediators were injected intradermally (prostaglandin E2, tumor necrosis factor-α, and low molecular weight hyaluronic acid) [23]. Knocking down Tmem120A also mitigated the acquisition of mechanical hyperalgesia in a systemic inflammation model (intraperitoneal lipopolysaccharide [LPS] injection) [23]. Interestingly, neuropathic hyperalgesia induced by chemotherapy drugs (oxaliplatin & paclitaxel) was not different between Tmem120A knockdown rats and controls [23]. These experiments indicate the expression of TMEM120A is necessary for the acquisition of inflammatory but not chemotherapy-induced mechanical hyperalgesia. Determining whether TMEM120A is upregulated in DRG neurons during inflammation has not been tested. However, TMEM120A was observed to be upregulated in trigeminal ganglion (TG) neurons when LPS soaked cotton balls were inserted into the pulp chambers of rat molars (a pulpitis model) [26]. This study also determined TMEM120A expression was necessary for the acquisition of pain associated with pulpitis in rats [26]. Whether TMEM120A is also necessary for the maintenance of inflammatory mechanical hyperalgesia has not been assessed. This is an important question as it will have implications for when administration of potential TMEM120A inhibitors may be effective at reducing pain, e.g., prior to the onset of inflammatory pain and/or during.

TMEM120A may play a role in mechanical allodynia as well. An E3-ubiquitin ligase called PARKIN has been proposed to regulate the subcellular localization of TMEM120A in DRG neurons [24,25]. Knockdown of Parkin increased the amount of TMEM120A at the plasma membrane, and overexpressing Parkin had the opposite effect [24,25]. Importantly, the total amount of TMEM120A was not altered when Parkin was overexpressed or knocked down [24]. In a co-immunoprecipitation assay, it appeared PARKIN and TMEM120A physically interact leading the authors to hypothesize this is how PARKIN modifies TMEM120A localization [24,25]. Knocking down Parkin in mice resulted in decreased nociceptive mechanical thresholds in von Frey tests indicative of a potential necessity for TMEM120A to be at the plasma membrane to affect mechanical pain [24]. In the context of inflammation, PARKIN is significantly downregulated in mice treated with Complete Freund’s adjuvant (CFA) [24]. Through this lens, it was proposed that inflammation downregulates PARKIN reducing its capacity to regulate TMEM120A’s subcellular localization causing increased membrane trafficking and thus mechanical allodynia [24]. This pathway was manipulated by feeding mice Hen Egg Lysozyme (HEL) resulting in PARKIN upregulation and normalization of TMEM120A distribution within DRG neurons [25]. Mice fed HEL were resistant to mechanical allodynia induced by CFA, vincristine, and spinal nerve ligation [25]. Of note, alterations to mechanical allodynia were only observed when using static stimuli (von Frey filaments) but not those that are dynamic (brushstrokes) [24,25].

Several independent labs have reported that TMEM120A is necessary for various mechanical pain states, but how can this reckon with what we know regarding its function? It remains unlikely TMEM120A is directly conferring tactile pain stimuli into ionic current given the significant doubt about its channel function [7,13–18]. However, it has been suggested that TMEM120A must reach some threshold of accumulation in the plasma membrane to contribute to mechanical pain sensing [24,25]. TMEM120A’s apparent ability to negatively regulate PKD2 and PIEZO2 at the plasma membrane [13,17], it is possible that it can affect other receptors or channels. Whether TMEM120A acts analogously to sphingomyelinase in endothelial cells by modifying the function of plasma membrane proteins through lipid production is not known [61]. Lipids such as lysophosphatidic acid have been implicated in the acquisition of pain [67–70]. Furthermore, TMEM120A’s ability to alter gene expression has not been explored in sensory neurons. Determining whether TMEM120A acts in this capacity to affect the mechanosensitivity of DRG neurons during inflammation is a logical line of investigation. Of note, all behavioral experimentation regarding TMEM120A’s involvement in mechanical pain was performed using only male animals [9,23,25,26]. Given the sexual dimorphic TMEM120A expression in adipocytes and effect on lipodystrophy [20], it would be pertinent to assess any differences in pain behavior between male/female Tmem120A-null animals.

One final unanswered question is how TMEM120A’s ability to inhibit PIEZO2 fits into mechanical pain sensing. PIEZO2 is necessary for mechanical allodynia [5], but loss of the channel only modestly affects hyperalgesia [6] and does not alter noxious mechanical pain sensing [5,71]. It has been reported that knocking out Piezo2 in a subset of mouse DRG neurons paradoxically increased mechanical pain responses [72]. This is hypothesized to be associated with gate control theory of pain where innocuous tactile stimulation relieves pain through inhibitory spinal circuits [72,73]. Whether TMEM120A can influence this circuitry by inhibiting PIEZO2 channels to facilitate noxious pain responses is speculative. The need for both TMEM120A and PIEZO2 for mechanical allodynia poses another paradox. However, assessing any changes to the interplay of TMEM120A and PIEZO2 during inflammation in DRG neurons has not been examined.

Conclusion

There are many unanswered questions regarding TMEM120A’s structure, function and role in mechanosensation, lipid metabolism, and STING regulation (Figure 3). Assessing whether TMEM120A can alter gene expression, lipid content, or both in tissues other than adipocytes could provide mechanistic insights into its function. Whether TMEM120A can act as a lipid modifying enzyme is unclear. It has structural homology to a lipid modifying enzyme (ELOVL7) but no direct measurements of enzymatic activity have been reported and its potential substrates remain unknown [14]. TMEM120A’s ability to alter gene expression in heterologous systems and adipose tissue [19,20,52] suggests it could be doing the same in sensory neurons, but this also has not been tested. Regardless, determining how TMEM120A affects DRG and TG neurons during inflammation could have substantial ramifications on our understanding of mechanical pain. Although TMEM120A may not be the slowly adapting mechanically activated ion channel in DRG neurons, its effect on several physiological processes leaves important questions for continued research.

Figure 3. Summary of TMEM120A’s electrophysiological analysis, structure, and putative regulation of PIEZO2 & PKD2 channels, mechanical pain, and lipid metabolism. Generated with BioRender.

Disclosure statement

No potential conflict of interest was reported by the authors.

Data availability statement

Figure 2 was generated from publicly available pdb files, available at: PDB DOI: 10.2210/pdb7F3T/pdb, PDB DOI: 10.2210/pdb6Y7F/pdb. Source data for Figure 1 is available at DOI: 10.1085/jgp.202213164.

Additional information

Funding

Work in the Rohacs lab is supported by NIH grants NS055159, GM093290, and GM131048.