Protein-Enhanced Small Molecule Disruptors of Ordered Membrane Domains

Abstract

Membrane order and fluidity influence many biological processes. However, tools to manipulate membranes under physiological conditions have been limited. In the process of high-throughput screening for molecules that shift the phase partitioning between ordered and disordered membrane phases of the tetraspan membrane protein peripheral myelin protein 22 (PMP22), we discovered two chemically similar compounds, VU0615562 and VU0619195, that shift PMP22 toward the disordered phase and destabilize the “lipid raft”-like ordered phase. Follow-up experiments showed that this latter activity is, counterintuitively, enhanced by the presence of PMP22, which normally stabilizes the ordered phase. Biophysical studies indicate that these compounds reduce raft stability through a mechanism that involves both direct interactions with proteins and the disruption of lipid packing. We further observed that acute treatment of live cells with VU0619195 modulated membrane fluidity and TRPM8 channel function while both compounds altered KCNQ1 channel activity, providing examples of practical applications for these compounds. These compounds reveal distinct lipid and protein-based forces that destabilize membrane order and represent new pharmacological tools for manipulating and probing the biological roles of ordered membrane domains in cells.

Introduction

A mid-2026 query of “lipid rafts” in PubMed yielded 7000 hits (over 1000 since 2020), reflecting an enormous interest in their properties and roles in biology. While long controversial, there now seems to be wide agreement that biological membranes do host raft-like membrane heterogeneities. These range from highly dynamic nanoscale fluctuations (1–3) to large and stable raft-like arrays, as occurs in specialized membranes such as myelin and the apical membrane of many epithelial cells (4, 5). Given the numerous proteins, pathways, and processes proposed to utilize or be modulated by lipid rafts (6–10), there is a need for pharmacological agents that can alter their formation or interactions with proteins. To discover small molecules with such properties, we conducted a screen for compounds that alter the ordered phase/disordered phase partitioning of human peripheral myelin protein 22 (PMP22). PMP22 has been quantitatively shown to preferentially partition into raft-like ordered domains in cell-derived giant plasma membrane-derived vesicles (GPMVs) (11). Here we describe and characterize an unexpected class of hits that resulted from this primary screen: compounds that reduce the partitioning of raft preferring proteins into rafts and reduce raft formation in a manner that is, counterintuitively, dependent on the presence of normally raft-promoting proteins. We believe these findings fit into updated models of the raft hypothesis. It has been proposed that cell membranes exist in a fluid phase very near the transition between liquid order (Lo) and liquid disorder (Ld) generating a highly responsive membrane that can be pushed toward a more Lo-like and/or Ld-like state depending on the cooperative forces between lipids and proteins (1, 2). The existence both of the protein-enhanced compounds described here and of a separate proteinagnostic class that we have previously reported (12), indicate that there are independent lipid- and protein-based forces contributing to lateral heterogeneities in membrane order.

Results

Discovery of compounds that alter the raft affinity of two membrane proteins

We conducted a high-throughput screen in GPMVs. A glycosylation-deficient mutant form (N41Q) of human PMP22 containing a non-perturbing Myc tag (for immunolabeling) was transfected into HeLa cells, which do not endogenously express significant quantities of PMP22. This non-glycosylated N41Q mutant form of PMP22 was used rather than wild type because the fraction of total N41Q PMP22 that traffics to the plasma membrane (i.e. the fraction of total PMP22 that reaches the plasma membrane vs that which is intracellularly retained) is 3x higher than wild type (13) while maintaining a WT-like preference for ordered domains (11), making this mutant easier to observe at the cell surface using fluorescence microscopy. GPMVs were then generated with paraformaldehyde (PFA) and dithiothreitol (DTT) (14) from N41Q PMP22-expressing cells and incubated at a room temperature, which is suitable to induce the formation of large and stable ordered and disordered phases that are easily observed following the addition of phase-specific dyes. GPMVs were then observed using a high-content imaging system allowing for thousands of GPMVs per condition to be imaged in an unbiased manner. The partitioning of PMP22 between raft-like ordered versus disordered phase domains (Pordered) was then quantitated from images using VesA software, as previously reported (15). Using this approach, we screened a chemically diverse 23,360 compound subset of the >100,000 compound Vanderbilt Discovery Collection, in a search for compounds that altered the fraction of PMP22 in the ordered phase (Pordered).

Hit compounds were re-ordered and their effects validated in triplicate (Fig. 1A). We then determined if hit compounds also altered raft formation in the GPMVs. Those compounds that altered both PMP22 partitioning and raft formation were then tested in GPMVs derived from HeLa cells that had not been transfected with PMP22, leading to the identification of compounds that alter both PMP22 partitioning and raft formation. GPMVs from untransfected cells were used to determine if any effects were proteinspecific. Compounds were next counter-screened against GPMVs made from cells expressing myelin and lymphocyte protein (MAL, fused to GFP) to test if their effect on protein partitioning is specific or more general. MAL is an unrelated tetraspan protein that also has a high affinity for ordered phase domains (16). Finally, hits were also tested in GPMVs derived from a second cell type (rat basophilic leukemia—RBL—cells).

Figure 1. Hits reduce affinity of PMP22 and MAL for ordered domains.

a) Workflow from HTS to hit classification b) Hit compounds VU0615562 and VU0619195 (at 10 μM) decrease ordered domain partitioning of PMP22 (n = 5) and c) MAL (n = 3) in GPMVs. Bars are means ± SD. P-values are from ANOVA followed by Dunnett’s tests. EC₅₀ values are from non-linear, sigmoidal fits, ± SE. d) Representative images of GPMVs treated with 0.5 μM of the indicated compound. White arrows indicate that GPMVs with different levels of PMP22 overlap with the L_d maker. Scale bars are 50 μm.

Based on their effects on PMP22 Pordered, MAL Pordered, and phase separation, hits fell into several distinct categories. In this paper, we focus on one distinct and surprising class of compounds: those that decrease Pordered for both PMP22 and MAL and also disrupt raft formation in a manner that is enhanced by the presence of these normally raft-favoring membrane proteins.

VU0615562 and VU0619195 decrease P_ordered of PMP22 and MAL

We identified two compounds that reduced the raft affinities of both PMP22 and MAL. VU0615562 and VU0619195 decreased P_ordered for both PMP22 and MAL by ~18-20% with half-maximal effective concentrations in the ~1 μM range (Fig. 1B and C). This can be seen in representative images as an increase in the overlap of the fluorescence from labeled PMP22 and a marker for disordered membrane domains (Fig. 1D, white arrows). These experiments were conducted using GPMVs derived from HeLa cells, which were chosen both for their relative ease of transfection and because they phase separate near room temperature. To determine whether the effect was cell type dependent, we tested the effects of the two hits on P_ordered for PMP22 in RBL cells. In these experiments, both compounds significantly reduced the P_ordered observed for PMP22 in this cell line (Fig. S1A).

VU0615562 and VU0619195 destabilize ordered domains in a manner enhanced by raft proteins

We next examined whether the compounds influenced raft stability. The fraction of vesicles in which phase separation is observed was taken as an indicator for raft formation. Both VU0615562 and VU0619195 significantly decreased the fraction of phase separated vesicles compared to control from untransfected HeLa cells (Fig. 2a). The effects on the fraction of phase separation were also examined in RBL cells. We found that VU0615562 was not soluble at concentrations >10 μM in the GPMV buffer that was incubated with RBL cells. These experiments were therefore conducted at 5 μM. A reduction in phase separation was observed at this concentration, although it was not statistically significant (Fig. S1B). This may be due to the lower compound concentration or to the lower baseline levels of phase separation observed in RBL GPMVs compared to HeLa GPMVs: ~30% (Fig. S1A) vs. ~70% (Fig. 2A).

Figure 2. Hit compounds decrease phase separation in a manner enhanced by raft proteins.

a) Effects of 10 μM VU0615562 and VU0619195 on the fraction of phase separated GPMVs from untransfected cells. b) Effects of 10 μM compounds with PMP22 expression (magenta bars) or MAL (navy bars) (n = 3-5). c) Dose response data for phase separation from cell expressing PMP22, MAL or untransfected cells. EC₅₀ values are from non-linear, sigmoidal fits, ± SE. d) Effects of VU0615562 and VU0619195 on the fraction of phase separated GPMVs with TfR expression (orange bars) or from untransfected cells (gray bars) at 10 μM (n = 3). e) Fraction of phase separated GPMVs from untransfected cells with and without proteinase K treatment and compounds at 10 μM (n = 3). P-values are from unpaired t-tests.

Remarkably, the reduction in percentage of phase-separated vesicles was more pronounced in GPMVs where PMP22 or MAL were overexpressed (Fig. 2B). It is notable that the change was similar in effect with either protein, even though PMP22 has been shown to stabilize rafts while MAL (which preferentially partitions into rafts) has no effect on raft stability (See SI and Fig. S2a). A similar observation was made in dose-response experiments: as seen in Fig. 2C, the EC₅₀ curve from untransfected GPMVs does not approach 0 in contrast to results for GPMVs from cells expressing PMP22 and MAL. For both compounds the EC₅₀ values for this activity were in the vicinity of 1 μM. We recently characterized another class of raft modulating compounds. Notably, this class of compounds destabilized lipid rafts in a manner that was unaffected by the overexpression of PMP22 or MAL (12), providing evidence that the behavior described in this paper for VU0615562 and VU0619195 is distinctive.

To determine whether this enhanced raft-reducing effect is specific to raftpartitioning proteins such as PMP22 and MAL, we probed the impact of compounds VU0615562 and VU0619195 on phase separation in GPMVs with the transferrin receptor (TfR), a protein known to preferentially partition into the disordered phase (17). In contrast to the effects of PMP22 and MAL, TfR expression did not alter raft disruption induced by the compounds (Fig. 2D) even after doubling the amount of DNA transfected (fig. S2D). This suggests that VU0615562 and VU0619195 reduce raft stability via a mechanism that is amplified by the presence of raft-partitioning proteins such as PMP22 and MAL. We therefore define these compounds as protein-enhanced raft modulators.

To further investigate the role of proteins in enhancing the activity of compounds VU0615562 and VU0619195, we examined their impact on GPMVs treated with proteinase K. Proteinase K is a broad-spectrum serine protease capable of cleaving proteins preferentially after hydrophobic residues. It has previously been shown that GPMVs can be porous (18) so both extracellular loops and intracellular loops are likely be cleaved by the protease. We used GPMVs from untransfected cells for these studies because the protease activity prevents us from labeling or detecting PMP22 or MAL (by cleaving the myc tag or GFP fusion). Proteinase K treatment reduced the effects of the compounds on phase separation (Fig 2E). Treatment, however, did not completely eliminate the effects of the compounds, possibly due to the inability of the protease to completely digest all membrane proteins. These data indicate that the impact of compounds VU0615562 and VU0619195 on phase separation is strongly coupled to the presence of membrane proteins. Notably, the effects of a protein-independent raftdisrupting control compound originally reported in a previous study (12), VU0607402, were not altered by Proteinase K treatment (Fig. S2B), supporting the conclusion that the protein-enhanced activity of VU0615562 and VU0619195 is unique to these molecules.

We further explored the interaction between PMP22 and MAL and the compounds by changing the relative ratio of compound to protein then examining the effects on phase separation. To do this, we conducted experiments in GPMVs generated from cells transfected with different amounts of DNA to express different levels of protein while holding the concentration of the compounds constant at 0.1 μM (well below the EC₅₀ values, Fig. 1C). At this non-saturating concentration, as the amount of protein increases the relative amount of compound goes down. The concentration used here was informed by pilot experiments demonstrating that, at this concentration, some degree of phase separation was detectable across the full range of protein expression levels. We found that the impact of the compounds on phase separations was higher in GPMVs expressing less protein (Fig. S2C). This likely indicates that, at this concentration (<<EC50), there is more protein than the compounds can interact with. Thus, adding more protein to the system lowers the relative concentration of the compound reduces the effect. Again, these data support an interaction between PMP22 and MAL that alters the effects on phase separation.

It is known that during vesiculation, lipid asymmetry between the inner and outer leaflets is lost and that this loss contributes to the phase separation observed in GPMVs (19). Loss of asymmetry is typically greater when PFA and DTT are used as vesiculating agents (19). As a control, we were motivated to determine if expression of PMP22 or MAL alters lipid asymmetry as this could confound the interpretation of the apparent protein-enhanced effects of the compounds. When membrane asymmetry is lost, phosphatidylserine (PS) becomes exposed on the outer leaflet. Annexin V binds to exposed PS on the outer surface of membranes, enabling its detection. In this manner, the amount of outer leaflet PS detected by Annexin V labeling can be used as a reporter for loss of membrane asymmetry. Here we made GPMVs from untransfected cells and cells transfected with PMP22 or GFP-MAL. We stained the resulting GPMVs with Annexin V and measured their intensity via flow cytometry. As previously reported, Annexin V labeling was observed in all GPMVs (19). However, there were no significant differences in Annexin V staining between GPMVs that from untransfected cells or those with either protein (Fig. S2D). This rules out differences in asymmetry induced by the proteins as a contributing factor to the apparent protein-enhanced effects of the compounds.

To verify that the compounds were not remodeling the GPMVs we examined their sizes and the sizes of the ordered phase domains (both measured using the analysis software VesA) following treatment with compounds VU0615562 and VU0619195. Neither GPMV radii nor the relative sizes of ordered phase domains were changed by VU0615562 or VU0619195 (figs. S3A and B), even though they decrease the raft affinities of PMP22 and MAL, as well as reducing raft formation.

An examination of the chemical structures of VU0615562 and VU0619195 (Table 1) reveals a Tanimoto coefficient (a metric of chemical similarity ranging from 0-1) of 0.7778 (20). This chemical similarity is consistent with their similar effects on Pordered for PMP22 and MAL, as well as on raft formation. Notably, both compounds contain a 1,2,3-triazole group, which is a common motif in drugs and drug-like compounds that has been extensively reported to interact with various amino acids (21). This is consistent with the notion these compounds may directly interact with proteins.

Table 1.

VU0615562 and VU0619195 compound structures and parameters

^1.cLogP – calculated water-octanol partition coefficient, a measure of hydrophobicity

^2.FS3P – fraction of sp3 carbon atoms

^3.TSPA – topological polar surface area; <140 indicates ability to permeate cell membranes (75)

Compounds exhibit weak/nonspecific interactions with PMP22

To test for possible compound-protein interactions, we examined the interactions of compounds VU0615562 and VU0619195 with PMP22 using nuclear magnetic resonance spectroscopy (NMR). 2D 1H,15N-TROSY NMR spectra of PMP22 treated with VU0615562 (Fig. 3A left) show concentration-dependent shifts of the R159 peak and the gradual appearance of two new peaks, labeled α and β(note that the amide 1H and 15N NMR crosspeaks for PMP22 are only partially assigned). R159 is located at the C-terminus of PMP22, near the membrane interface. A similar perturbation of the R159 chemical shift was observed in the presence of VU0619195 (Fig. S3B left) and one new peak appeared (labeled γ) in the same region as the two new (α and β) peaks in the VU0615562 spectrum. Plots of the intensities of the new peaks or of the magnitude of the shifts of the R159 peak as a function of concentration suggest only weak or nonspecific interactions (Fig. 3A and B, right) of both compounds with PMP22. However, the appearance of new peaks during a ligand titration is normally associated with slow exchange on the NMR time scale, observation of which is very unusual for such thermodynamically weak binding. In our experience with NMR titrations of PMP22 with thousands of small molecules (22), this class of interaction with PMP22 is very rare. These data support a model in which compounds VU0615562 and VU0619195 interact only weakly with PMP22, but in a long-lived manner. Our data seem to be most consistent with the notion that these compounds interact with membrane-exposed or membrane-water interfacial sites of PMP22. To test if this interaction with PMP22 is unique to this class of compounds we used a recently described a control compound, VU0607402, that also destabilizes ordered domains but does so in a proteinindependent manner (12). Treatment with VU0607402 resulted in no chemical shift perturbations or appearance of new peaks (Fig. S4). This indicates that not all raft disrupting small molecules engage in weak/nonspecific interactions with PMP22.

Figure 3. Compounds show weak/nonspecific interactions with PMP22.

a) (left) Four overlaid ¹H-¹⁵N TROSY spectra of PMP22 at varying concentrations of VU0615562 showed the appearance of two new peaks, labeled α and β, and movement of the peak corresponding to R159. (right) Plots of the peak height (for the new peaks) and of the chemical shift perturbation, Dd, (for R159), as a function of the concentration. b) (left) Six overlaid ¹H-¹⁵N TROSY spectra of PMP22 at varying concentrations of VU0619195 showed the appearance of a new peak, labeled γ, and movement of the peak corresponding to R159. (right) Plot of either the peak height (for the new peak) or chemical shift perturbation, Δδ, (for R159), as a function of the concentration of VU0619195 is linear, suggesting that the binding is non-specific or very weak.

To further evaluate the mechanism of action of compounds VU0615562 and VU0619195, structure-activity relationship (SAR) studies were conducted on chemically similar compounds available in the VU Discovery Collection (fig. S5). For VU0615562, seven additional compounds with similar structures (similarity integer value of 97) were tested. For VU0619195, six additional compounds with similar structures (similarity integer value of 85) were tested. All contained a 1,2,3-triazole group. None of the compounds tested had a greater impact on PMP22 ordered partitioning or raft stability than VU0615562 and VU0619195 (fig. S5). It is interesting, however, that for both parent compounds, analogs were found that maintained the capacity of the parent to reduce the raft affinity of PMP22 but largely lacked the ability to reduce raft formation, indicating that these two activities are incompletely coupled.

Compounds VU0615562 and VU0619195 alter membrane fluidity in vesicles and live cells

To shed light on how VU0615562 and VU0619195 reduce raft formation, we next examined additional effects on membranes. As an initial test, we examined their effects on membrane fluidity. This approach also provided an opportunity to connect GPMV findings with raft behavior in both model membranes and in plasma membranes of living cells. For this, we used the environmentally sensitive dye Di-4-ANEPPDHQ (Di-4) to report on membrane fluidity (23). Increased membrane fluidity causes a redshift of the Di-4 emission spectrum, whereas blue-shifted spectra arise from decreased fluidity.

We first treated GPMVs from HeLa cells incubated with each compound (10 μM) and Di-4 and then measured emission spectra. For both VU0615562 and VU0619195, a pronounced redshift from vehicle was observed in micrographs (Fig. 4A) and in the emission spectra (Fig. 4B). To quantify these changes, generalized polarization (GP) values were calculated (Fig. 4C). GP values provide a relative means of comparing spectral shifts. For Di-4 emission, lower values correspond to a more fluid (redshifted) environment. Decreased GP values were detected for both VU0615562 and VU0619195, indicating that they significantly increase membrane fluidity (Fig. 4C). Furthermore, we detected a difference in activity of the two compounds: VU0619195 increased fluidity more than VU0615562.

Figure 4. Compounds alter membrane fluidity in GPMVs, live cells and LUVs.

a) Representative spectral images (cropped from 40X fields) of GPMVs stained with Di-4 and treated with DMSO or 10 μM VU0619195. b) Representative Di-4 emission spectra of GPMVs treated with 10 μM hit compounds. c) Generalized polarization values calculated from Di-4 emission spectra shown in B. Bars are means ± SD (n = 3). P-values are from ANOVA followed by Dunnett’s multiple comparisons tests. d) Absolute value of change in GP from DMSO control in Lo and Ld LUVs. P-values are from unpaired student’s t-tests, n = 3. e) Representative spectral images of live HeLa cells stained with Di-4 and treated with DMSO or 10 μM VU0619195. Scale bar = 50 μm. f) Generalized polarization values calculated from Di-4 emission intensities calculated from individual cells as shown in D. Bars are means ± SD, 10-15 cells (technical replicates) per treatment were measured for each of 4 biological replicates. P-values are from ANOVA followed by Dunnett’s multiple comparisons tests.

GPMVs consist of both protein and lipid components. To investigate the effects of the compounds on lipid-only bilayers, we also performed Di-4 fluidity experiments in large unilamellar vesicles (LUVs). This provided an opportunity to answer questions regarding the GPMV phase separation experiments. First, the protein-enhanced effect was observed for two raft proteins but not for a non-raft protcein. This leads to the question of whether compounds preferentially interact with lipids in ordered domains or with raft proteins. Second, while experiments with GPMVs show the effects of the compounds are enhanced by raft proteins, they do not inform on whether the compounds affect lipid-only bilayers. To test this, we generated L_o and L_d LUVs consisting of mixtures of POPC, sphingomyelin (SM), and cholesterol at concentrations known to form liquid ordered (L_o, 25% POPC, 35% SM, 40% cholesterol) or liquid disordered (L_d, 70% POPC, 25% SM, 5% cholesterol) phases. We then examined the absolute difference in GP values in the presence and absence of the compounds in the L_o versus L_d LUVs. Both compounds altered GP in both ordered and disordered LUVs, indicating they enhance membrane fluidity (Fig. 4D). GP measurements at different lipid to compound ratios can be seen in Fig. S6. These data suggests that the compounds interact with both L_o and L_d membranes and that they have effects in the absence of proteins. Strikingly, however, the effect of VU0619195 was significantly larger in the ordered LUVs compared to the disordered LUVs (Fig. 4D). We note that the fluidity was not significantly different in ordered membranes treated with VU0615562 despite its high chemical similarity to VU0619195. We speculate these differences are likely due to the chemical substitutions on the peripheral phenyl groups of the two compounds, resulting in different interactions with membrane lipids. In summary, raft-destabilizing compounds VU0619195 and VU0615562 increase membrane fluidity in both GPMVs and LUVs, and these effects are specifically enhanced in L_o membranes for VU0619195. We note that, while we do not know the compound-to-lipid ratio for GPMVs, the change in GP in GPMVs (~0.22) was more than double the change in GP observed at a 10:1 lipid:compound in LUVs (~0.08) (Fig. 4C and Fig. S6).

We repeated the fluidity measurements for VU0615562 and VU0619195 with Di-4 in live cells. HeLa cells were treated with compounds ~15 min before a brief incubation with Di-4 to limit internalization of the dye and ensure primarily plasma membrane labeling. Cells were then imaged via laser scanning confocal microscopy using a spectral detector, allowing images from multiple emission wavelengths to be simultaneously acquired (Fig. 4E and fig. S7A). GP values were then calculated from the confocal microscopy data. We found that the changes in GP were very similar to those from GPMVs (Fig. 4F). Compound VU0619195 produced a statistically significant increase in membrane fluidity. As a control, we added methyl-β-cyclodextrin (MBCD) to HeLa cells to deplete membranes of cholesterol. Compared with the effects of VU0619195, the effect of MBCD treatment was modest. Although not statistically significant, the effects of VU0615562 on Di-4 emission in live cells showed a change similar to that observed in the GPMV Di-4 experiments (Fig. 4, compare C and F).

It is notable that the MBCD-treated cells appeared rounded (previously reported (12)) (fig. S7A, bottom row) compared to the compound-treated cells, which retained the well spread appearance of untreated cells. This implies that compounds VU0615562 and VU0619195 are better tolerated by live cells than MBCD treatment. Trypan blue cell viability experiments confirmed neither VU0615562 nor VU0619195 were cytotoxic in the conditions used in the live cell experiments of this study (fig. S7B) while MBCD showed toxicity as we previously showed (12). This suggests these compounds could serve as improved tools for manipulating lipid rafts in living cells.

Compound VU0619195 alters TRPM8 channel activity

There are no functional assays for the activity of PMP22 or MAL. To test if compounds VU0615562 and VU0619195 could alter a reported lipid raft-dependent biological system, we examined their effects on the human TRPM8 ion channel function. TRPM8 is a cold- and menthol-sensing ion channel (24, 25). Many ion channels, including TRPM8, are believed to partition into and be regulated by ordered membranes (26–28). TRPM8 activity has been previously shown to be sensitive to levels of membrane cholesterol suggesting it is sensitive to membrane order (29). Automated patch clamp (APC) electrophysiology experiments were conducted (Fig. 5A, fig. S7A) to determine if treatment with VU0619195 altered TRPM8 function. VU0619195 was selected due to its significant effects on membrane fluidity in live cells. MBCD was again included for comparison to previously published work. APC was chosen for its applicability to cells that are patched for longer periods of time. TRPM8- expressing cells were equilibrated to 30 °C (29) and treated with menthol, the canonical TRPM8 agonist, before and after 15 minutes of continuous microfluidic perfusion with the lipid raft-modulating compounds. Comparisons of the currents pre- and postmenthol-stimulation were used to evaluate the effects of the compounds on TRPM8 function. Treatment with VU0619195 and MBCD reduced TRPM8 activation in response to menthol stimulation (Fig. 5B). These results suggest that TRPM8 function is sensitive to bulk changes in membrane order whether they are induced by this small molecule tool or by altering lipid content.

Figure 5. Effects of membrane modulators on TRPM8 and KCNQ1 activity.

a) Average current traces of TRPM8 menthol response before and after exposure to the compounds in HEK293T cells expressing human TRPM8 from automated patch clamp experiments. Each cyan dot indicates when the pulse program is applied. 100 μM menthol was applied for 75 seconds 4 times before perfusing 0.03% DMSO (control) or 10 μM VU0619195 or 10 mM MBCD for 15 min. 100 μM menthol was then applied with compound twice for 75 seconds. One n is the sum of 20 cells per amplifier on the plate. b) Average ratio of menthol response from each compound in TRPM8 HEK293T cells. The ratios were calculated from data in panel a, where the last two menthol responses before compound application were averaged and compared to the two menthol responses after compound application. P-values were determined by ANOVA followed by Dunnett’s tests. Bars are means ± SD. c) Normalized current at 100 mV following the perfusion protocol (Fig. S8B). Points are means ± SEM. d) (left) Superimposed ¹H-¹⁵N BEST-TROSY spectra of hTRPM8-VSLD titration with VU0619195 from an NMR-detected titration. (right) Global analysis of the titration data as a function of VU0619195 concentration. e) Averaged KCNQ1/KCNE1 currents generated by a 60 mV voltage step after acute exposure to DMSO or hit compounds at 1, 3, and 10 μM concentrations. Compounds were washed onto cells five minutes prior to establishing the whole-cell recording configuration. P-values from ANOVA followed by Dunnett’s tests (n = 37-69 cells, error = 95%CI).

Next, to assess shorter time scales with compound treatment, whole-cell manual patch-clamp electrophysiology was conducted. Experiments using VU0619195 show its inhibitory effect on TRPM8 activation occurs within 1 minute of treatment and that this effect is completely reversed within 1 minute of washing out the compound (Fig. 5C, Fig. S8B, C, and D), indicating a relatively rapid and reversible effect for this compound. This reversibility, combined with the data showing that compounds VU0615562 and VU0619195 are not toxic at the concentrations used, supports the idea that these compounds are not exerting their effects by disrupting membrane integrity.

To verify that the effects of VU061919 were not due to the compound changing the levels of TRPM8 at the cell surface we quantified TRPM8 at the plasma membrane via flow cytometry. We found that VU061919 did not alter cell surface levels of TRPM8 (fig. S8E). Interestingly, as we previously reported, MBCD increased the level of surface TRPM8, which may be a consequence of it inhibiting endocytosis (12, 30).

We next asked if VU0619195 directly interacts with TRPM8 using NMR. NMR titrations of the compound to the TRPM8 voltage sensing-like domain (VSLD) were conducted. TRPM8-VSLD is the main regulatory domain of TRPM8, containing the canonical menthol binding site (31). We observed multiple TRPM8-VSLD chemical shift perturbations in response to compound titration without saturation being observed, indicating weak/nonspecific interactions with TRPM8-VSLD (Fig 5D, fig. S9A and B) similar to those observed with PMP22 (Fig. 3). A larger number of perturbed resonances were observed with TRPM8-VSLD than for PMP22. These data indicate that these weak/nonspecific interactions with VU0615562 and VU0619195 are not unique to PMP22 but occur with at least some other membrane proteins.

Compounds VU0615562 and VU0619195 alter KCNQ1 channel activity

KCNQ1 is another ion channel that has been proposed to associate with ordered regions of membranes depending on its binding to different KCNE subunits (32). Here we tested the activity of KCNQ1 with accessory protein KCNE1, which has previously been shown to purify with caveolin enriched fractions suggesting the KCNQ1/KCNE1 complex associates with ordered membranes (32). Automated patch clamp recording experiments were conducted. We found that, like TRPM8, the activity of KCNQ1/KCNE1 was inhibited by acute treatment with order-disrupting compounds. Both VU0615562 and VU0619195 inhibited activation of KCNQ1/KCNE1 relative to the vehicle-treated condition (Fig. 5E, fig. S10A-D).

Compounds VU0615562 and VU0619195 do not alter EGFR activity

To test the generality of these findings, we also examined the impact of compounds on the function of a third membrane-based molecular system: epidermal growth factor receptor (EGFR) signaling in response to the epidermal growth factor (EGF). Previous studies indicated changes in EGFR activation following cholesterol depletion by MBCD treatment (33–35). We treated HeLa cells with VU0619195 for 15 minutes (as in the TRPM8 experiments) prior to a brief (1 min) treatment with EGF. Cells were then lysed and levels of EGFR phospho-tyrosine 1173 and phospho-ERK (phosphorylated downstream of EGFR phosphorylation) were quantified via Western blot analysis. We did not see significant effects on either EGFR or ERK phosphorylation (Fig. S11), indicating that these signaling processes are not sensitive to short-term changes in membrane fluidity or raft modulation. These data also indicate that not all plasma membrane signaling events are sensitive to the changes in lipid raft stability and fluidity induced by VU0619195.

Discussion

Here we present first-in-class small molecules that destabilize raft formation and decrease protein partitioning at micromolar concentrations in a manner enhanced by two proteins that normally preferentially partition into rafts. Both molecules increased fluidity in GPMVs and LUVs, indicating they affect both lipid-only vesicles and complex biological membranes. In live cells, VU0619195 induced a greater increase in membrane fluidity than cholesterol depletion by MBCD. Both compounds were observed to be non-toxic to model cell lines. VU0619195 also altered the signaling of a raft-sensitive ion channel, TRPM8, while both compounds inhibited KCNQ1/KCNE1 activation. These compounds represent promising pharmacological tools for manipulating lipid raft formation and membrane proteins in biophysical and cell biological studies.

Proposed mechanism of VU06156562 and VU0619195

Both PMP22 and MAL have been quantitatively shown to preferentially partition into ordered membrane domains over disordered domains (11, 16). Both MAL and PMP22 are tetraspan membrane proteins but have no sequence homology and are not believed to be structurally or functionally related. Thus, our results showing that the preferential raft partitioning of these two proteins is dramatically reduced by the presence of VU0615562 and VU0619195 suggest some degree of non-specificity in how they interact with proteins to exert this activity. However, this effect was not observed in the presence of a disordered phase preferring membrane protein, the transferrin receptor.

VU06156562 and VU0619195 are hydrophobic, with cLogP values of 3.93 and 4.49 respectively (meaning 99+% of the compounds will partition into the membrane), but below the threshold of cLogP = 5 for ~100% sequestration into the interior of membranes or fat. This is supported by our results showing these compounds interact with LUVs (Fig. 4D). The molecular architecture of these molecules is not lipid-like (Table 1). Each compound has aromatic moieties with attached polar groups. Aromatic groups containing a nitrogen (pyridine-like) or with directly attached hydrogen bonding moieties preferentially partition into membranes in the headgroup/interface region, not deep in the membrane interior (36). This observation suggests that these molecules interact with the membrane near the water-aqueous phase interface, allowing their apolar (mainly aromatic moieties) to extend into the membrane surface below the polar headgroups of the lipids, while their more polar moieties are likely located in the hydrated headgroup region of the membrane.

VU06156562 and VU0619195 also share a polar 1,2,3-triazole group. 1,2,3- triazoles have been shown to be highly interactive with proteins. A study of 220 PDB structures co-crystallized with 1,2,3-triazole containing compounds found that this group is a potent hydrogen bond acceptor, interacting with many polar amino acids (21). Our NMR studies of the interactions of PMP22 with both compounds revealed them to undergo a weak/nonspecific interaction with residue R159 located at the interfacial C-terminus of the protein and at least a couple of other residues (as reflected by the appearance of new unassigned peaks). Since the non-specific interactions of VU0615562 and VU0619195 with PMP22 were similar it logically follows that these interactions are based on PMP22 interacting with the moiety shared by the two molecules, the 1,2,3-triazole core. It is notable that R159 is part of a putative cholesterol binding (CRAC) motif, a common motif found in many membrane proteins. Indeed, this motif has previously been shown to be involved in cholesterol interactions with PMP22 (37). We hypothesize that the compounds likely interfere with PMP22-cholesterol interactions, disfavoring its partitioning in the ordered phase. While not demonstrated definitively, studies with MAL in cholesterol-depleted GPMVs suggest it may also bind cholesterol (16). Both MAL and PMP22 are believed to play important roles in cholesterol homeostasis in certain cell types (38–41)

NMR results with the TRPM8 VSLD also indicate a weak/nonspecific interaction with VU0619195. The general similarity of its NMR-monitored response to titration by VU0619195 to the analogous response of PMP22 indicates the mechanism by which VU0615562 and VU0619195 shift partitioning of otherwise raft-preferring proteins toward the disordered phase may be generalizable to other membrane proteins. We do note that there were a larger number of perturbed chemical shifts for TRPM8-VSLD compared with PMP22. TRPM8-VSLD is highly dynamic, and it was recently shown that ligands bind to this voltage sensor-like domain via conformational selection (31), indicating strong coupling between compound binding, modulation of TRPM8-VSLD dynamics, and regulation of channel function.

Membrane fluidity experiments in GPMVs and LUVs demonstrated that both compounds also increase membrane fluidity. The effects on fluidity of VU0619195 were increased in Lo phase LUVs with higher levels of SM and cholesterol, while the activity of VU061556 was not significantly different in Lo vs Ld LUVs. We hypothesize that these differences are due to the structural differences in the two molecules (see Table 1). Finding that VU06156562 and VU0619195 altered the fluidity of LUVs also demonstrated that the compounds do interact with and alter the fluidity of protein-free bilayers, allowing us to conclude that the compounds act in a manner that is enhanced by raft proteins but not entirely dependent on them.

The above considerations lead us to propose a model for how VU06156562 and VU0619195 promote a shift from preferential ordered phase domain partitioning of PMP22 towards the disordered phase and reduce raft phase formation (Fig. 6). According to this model, the peripheral parts of the two compounds preferentially disrupt lipid-lipid interactions in the ordered phase to promote the disordered phase while the core 1,2,3-triazole of the molecule interacts non-specifically with PMP22 and other raft-normally raft-preferring membrane proteins through structural features that are common to membrane proteins, particularly the presence of polar residues near the membrane-aqueous phase interface. These non-specific interactions may interfere with the lipid-protein interactions that favor partitioning of the protein in the ordered phase and that stabilize the raft phase. As a consequence, normally ordered phase-preferring proteins such as PMP22 not only shift out of the ordered phase domain and into the disordered phase, but also destabilize the ordered phase when they do so. While we use PMP22 here as a model, our NMR findings with TRPM8-VSLD indicate that this mechanism may be extended to other (normally) raft-preferring membrane proteins. On the other hand, there was little impact on ordered domain formation when TfR was expressed, as expected since it prefers to partition into the disordered phase regardless of whether VU06156562 or VU0619195 is present. While we think the above proposed mechanism best explains the data, it remains possible that expression of PMP22 and MAL induce other changes in the protein and lipid profiles of cells which enhance the effects of these compounds.

Figure 6. Proposed mechanisms of compounds.

Compounds form weak/nonspecific interactions with membranes proteins via the core 1,2,3-triazole. Proposed interaction with the PMP22 CRAC motif illustrated. Compounds also interact with the membrane, likely through the peripheral aromatic groups to promote disorder.

That proteins play a role in the stability of membrane domains is not a new concept. Molecular dynamics simulations have shown that the presence of transmembrane proteins can nucleate domain formation in mixed membranes (42). It has also been suggested that transient homodimers of GPI-anchored proteins stabilize rafts (43). The actin cortex on the other hand has been shown to limit domain formation and slow the diffusion of membrane components (44). CTxB, the membrane binding subunit of cholera toxin, a tool commonly used in lipid raft studies, promotes phase separation by clustering the GM1 ganglioside lipid molecules (45, 46). It has also been theorized that observed sphingolipid biding motifs on proteins may promote their inclusion in and assembly of lipid rafts (47, 48).

In addition to the protein-enhanced effect of these compounds to reduce ordered phase formation, VU06156562 and VU0619195 have an intrinsic disordered phase-disrupting activity that seems to be based on disrupting lipid-lipid interactions, which are more pronounced in the ordered phase than in the disordered phase (49), an effect which energetically favors the disordered phase relative the ordered phase. Ordered domains have long been thought to be stabilized by both lipid-lipid and lipid-protein interactions (50–52). Here we show that compounds VU06156562 and VU0619195 appear to act not so much by their impact on bilayer properties and protein-lipid interactions in the disordered phase, but rather mainly by preferentially destabilizing both lipid-lipid and lipid-protein interactions in raft-like ordered phase domains. This dual mode of action appears to be unique to this class of compounds. We recently described a separate class of compounds that modulate phase separation in a protein-independent manner (12). These previously described compounds were found to significantly alter GPMV phase separation, but this effect was no different in GPMVs with or without PMP22 or MAL. The activity of those compounds was unaffected by proteinase K, unlike VU06156562 and VU0619195 presented in this work. The protein-independent compounds also demonstrated no interactions with PMP22 by NMR. Due to these differences, we categorized the protein-independent and protein-enhanced compounds as two different classes of membrane modulators.

Proof-of-principle application of VU0619195 in membrane biology

We tested the use of VU0619195 in studies with TRPM8, KCNQ1/KCNE1, and EGFR. Prior studies with MBCD and cholesterol-depleting drugs implicated membrane order in regulating the activities of TRPM8, KCNQ1/KCNE1, and EGFR (33, 34, 53, 54). Our results showed that micromolar levels of this compound significantly decreases TRPM8 signaling. In our proposed mechanism for VU0619195, we suggest that they interfere with cholesterol binding to TRPM8 to exert their effects. TRPM8 has several highly conserved cholesterol binding motifs (including three in the VSLD domain) that these compounds may interfere with (55). It is also possible that there are additional weak/nonspecific interactions of VU0619195 with full-length TRPM8 beyond those reflected by our NMR-monitored titration of the VSLD. The finding that the effects of VU0619195 on TRPM8 signaling are both rapid and reversible (Fig. 5C) highlights its utility as a tool for functional studies. Our results also show that acute treatment of the KCNQ1/KCNE1 channel complex with either VU06156562 or VU0619195 reduces channel function, consistent with the previous data showing this complex natively occupies raft-like membrane domains (32).

Despite reducing raft formation, neither compound affected EGFR phosphorylation. This contrasts with the effect of MCBD, a known lipid raft disruptor, on EGFR signaling (33, 34, 56). These data demonstrate functional differences in the response of the membrane to MBCD and our newly described compounds. They also show that not all “raft dependent” signaling events at the plasma membrane are sensitive to these compounds, opening the possibility of further discriminating the role of rafts in these events. Further studies are required to confirm the proposed mechanisms by which the compounds reported in this work exert their effects. We also do not conclude from this work that the alterations in membrane order and fluidity induced by the compounds will only affect the activity of raft-associated proteins. Additional follow-up studies examining broad classes of membrane proteins that are thought to be raft-dependent vs raft-independent using these compounds will provide additional mechanistic insights as well as answer long-standing biophysical questions.

Conclusions

Overall, the discovery of this class of raft-modulatory molecules in this work presents new pharmacological tools for interrogating lipid rafts and raft proteins in cells. Furthermore, the discovery of this class of compounds sheds light on the fact that both protein-lipid and lipid-lipid interactions collude to stabilize lipid rafts (51, 52) and, as supported by our previous work (12), that these forces can be independently manipulated.

Materials and Methods

Details of all methods for this paper are presented in Supporting Information. This includes cell culture and transfection procedures for all cell lines used, GPMV formation, labeling and imaging, high-throughput screening and data analysis procedures, and compound purchasing information. We detail NMR experiments with PMP22 and TRPM8, fluidity assays, cell viability, and electrophysiology experiments with TRPM8 and KCNQ1/KCNE1, and immunoblotting with EGFR. Statistical methods are described as well.

Significance.

Membrane order plays a critical role in cellular function, yet tools to manipulate it under physiological conditions are limited. We report two small molecules, VU0615562 and VU0619195, that destabilize ordered membrane domains through protein-enhanced mechanisms. These compounds shift raft-preferring proteins like PMP22 and MAL into disordered phases and increase membrane fluidity in both model and live cell membranes. They also modulate TRPM8 and KCNQ1 ion channel activity, demonstrating functional utility. Unlike cholesterol-depleting agents, these molecules act reversibly and are non-toxic, offering control over membrane organization. This work introduces a new class of pharmacological tools for probing lipid raft biology and highlights the interplay between lipid–lipid and protein–lipid interactions in membrane domain stability.

Data and materials availability:

All data are available in the main text or the supplementary materials.