Identification of a Binding Motif in the S5 Helix That Confers Cholesterol Sensitivity to the TRPV1 Ion Channel*

The TRPV1 ion channel serves as an integrator of noxious stimuli with its activation linked to pain and neurogenic inflammation. Cholesterol, a major component of cell membranes, modifies the function of several types of ion channels. Here, using measurements of capsaicin-activated currents in excised patches from TRPV1-expressing HEK cells, we show that enrichment with cholesterol, but not its diastereoisomer epicholesterol, markedly decreased wild-type rat TRPV1 currents. Substitutions in the S5 helix, rTRPV1-R579D, and rTRPV1-F582Q, decreased this cholesterol response and rTRPV1-L585I was insensitive to cholesterol addition. Two human TRPV1 variants, with different amino acids at position 585, had different responses to cholesterol with hTRPV1-Ile585 being insensitive to this molecule. However, hTRPV1-I585L was inhibited by cholesterol addition similar to rTRPV1 with the same S5 sequence. In the absence of capsaicin, cholesterol enrichment also inhibited TRPV1 currents induced by elevated temperature and voltage. These data suggest that there is a cholesterol-binding site in TRPV1 and that the functions of TRPV1 depend on the genetic variant and membrane cholesterol content.

The transient receptor potential (TRP) 3 family of ion channels is found throughout the animal kingdom and has been shown to subserve numerous functions. One extensively studied member of this family is the TRPV1 (Vanilloid 1) channel. Structurally, TRPV1 is thought to be a tetramer comprised of subunits each with six transmembrane domains (S1-S6), with the putative pore of the channel located between S5 and S6. It also contains large intracellular amino and carboxyl termini that have been shown to be involved both in channel gating and regulation (for review, see Ref. 1). Electrophysiological studies have shown that TRPV1 is an outwardly rectifying, non-selective, and calcium-permeable cation channel (2).
Although TRPV1 is found in many organs, it is predominantly expressed in primary sensory neurons where it acts as a sensor and integrator for noxious stimuli (2). In addition to being a temperature sensor and weakly voltage-dependent, TRPV1 is sensitive to a variety of chemicals (1). Among its many chemical agonists is capsaicin, the principal pungent compound in chili peppers. The intracellular binding pocket for capsaicin has been identified in a region that spans from S2 to S4 (for review, see Ref. 1). In addition to capsaicin, many complex amphiphilic molecules have been shown to activate or inhibit TRPV1. These include polyring compounds such as resiniferatoxin (3), quinazolinone (4), evodiamine (5), 17-␤-estradiol (6), as well as molecules with long acyl and amide chains such as anandamide (7,8), olvanil, and omega-3 polyunsaturated fatty acids (for review, see Ref. 1). Moreover, TRPV1 activity is regulated by the membrane lipid phosphatidylinositol 4,5-bisphophate (9 -11).
Another important membrane lipid in terms of TRPV1 activity is cholesterol (Fig. 1). Cholesterol is a major component of plasma membranes where it increases bilayer mechanical strength, thereby helping to prevent cell lysis (12). Importantly, cholesterol has been shown to modify the function of many classes of ion channels (for review, see Ref. 13). Cholesterol can modify channel activity indirectly by altering the thickness and elastic properties of the surrounding lipid bilayer (14). In addition, in recent years compelling evidence has shown a specific interaction between cholesterol and several protein channels (13,15). That is, based on experiments with structural analogs of cholesterol and channels with specific point mutations, it has been found that cholesterol binds to certain channel proteins, including some with structural similarities to TRPV1 (16). For either of these indirect or direct mechanisms, cholesterol modifies the energy difference between the open and closed states of the channel.
Previous studies of the effect of cholesterol on TRPV1 in whole cells showed that cholesterol depletion by incubation with methyl-␤-cyclodextrin (M␤CD) causes large decreases in capsaicin-evoked responses (17,18). Both studies suggest that these effects could be due to TRPV1 being functional only when present in cholesterol-rich plasma membrane microdomains (rafts) that can modulate the activity of some receptors and transport proteins. However, cholesterol depletion from whole cells also reduces TRPV1 concentration in the plasma membrane, thus making it difficult to determine whether the observed results are due to cholesterol effects on: 1) TRPV1 located in the plasma membrane or 2) TRPV1 trafficking in the cell (17).
Given this information, we thought it important to further explore the mechanisms by which cholesterol modifies TRPV1 activity. To avoid possible effects of TRPV1 cellular trafficking, we measured capsaicin-activated currents by patch clamping excised plasma membranes from HEK293 cells containing heterologously expressed rTRPV1. We found that the rTRPV1 currents were not changed by cholesterol depletion, but were markedly decreased by cholesterol enrichment, indicating that rTRPV1 function was modulated by cholesterol concentration in the plasma membrane. Therefore we also determined the membrane microdomain location of rTRPV1.
To determine whether there were specific cholesterolbinding sites in rTRPV1 three additional series of patch clamp experiments were performed. First, epicholesterol (␣-3-OH-cholesterol epimeric form) was substituted for cholesterol (see Fig. 1), as previously done with other channels (16,19). Second, capsaicin-induced currents were measured for membranes transfected with rTRPV1s with specific point mutations in the S5 transmembrane helix, which has a sequence consistent with the cholesterol recognition amino acid consensus (CRAC) sequence found in several transmembrane proteins that bind cholesterol (20,21). Third, currents were measured for two common hTRPV1 variants with the same S5 sequence as wild-type rTRPV1 except for one amino acid residue in the hydrophobic region of the helix. In addition, experiments were performed to determine whether cholesterol enrichment also modified heat-activated and voltage-activated TRPV1. The results from these experiments indicate that there is a cholesterol-binding site in the TRPV1 S5 helix that when occupied by cholesterol prevents the channels from opening.

EXPERIMENTAL PROCEDURES
HEK293 Cell Culture and Capsaicin-induced Currents-HEK293 cells expressing large T antigen were transfected with wild-type and mutant pCDNA3-rTRPV1 and pIRES-GFP (BD Biosciences) with Lipofectamine (Invitrogen) following previously described methods (22,23). Inside-out and outside-out patch clamp recordings of TRPV1 were made using Ca 2ϩ -free symmetrical solutions consisting of 130 mM NaCl, 3 mM HEPES (pH 7.2), and 1 mM EDTA. Solutions were changed with an RSC-200 rapid solution changer (Molecular Kinetics, Pullman, WA). Unless otherwise indicated, all chemicals were purchased from Sigma.
Capsaicin-activation curves were measured at 120 mV and normalized to a saturating capsaicin concentration (4 M). Dose-response relationships were fitted with the Hill equation, where I is the mean current, I max is the maximum current, K 1/2 is the capsaicin concentration at I ϭ I max /2, n H is the Hill coefficient, and [cap] is the capsaicin concentration. Currents were low-pass filtered at 2 kHz and sampled at 10 kHz with an EPC 10 amplifier (HEKA Elektronik GMBH, Pfalz, Germany). For macroscopic current recordings, the following voltage protocol was used: patches were initially held at 0 mV for 10 ms, and voltage was then stepped from 0 to Ϫ120 to 120 mV in 10-mV increments for 100 ms, and then returned back to 0 mV for 10 ms. These recordings were performed at room temperature (19 Ϯ 3°C). For all experiments, leak currents in the absence of capsaicin were subtracted from currents in the presence of capsaicin.
Time courses of modification by cholesterol and epicholesterol were obtained by holding the membrane potential to 0 mV and then pulsing to ϩ120 mV for 100 ms every 10 min in the presence of 4 M capsaicin. Data were acquired and analyzed with PULSE software (HEKA Elektronik, Germany) and were plotted and analyzed with programs written using Igor Pro (Wavemetrics Inc., Portland, OR).
Noise Analysis-To determine the value of the open probability (P o ), the number of channels in the patch (N), and the single-channel current (i), we performed stationary noise analysis of macroscopic current traces activated by capsaicin at a voltage of 100 mV. Because currents activated instantaneously with voltage pulses, we changed the open probability by varying the concentration of capsaicin. For each patch, the mean (I) and variance (␦ 2 ) were determined from 50 to 80 current traces using the algorithm of Heinemann and Conti (24). Traces were recorded for increasing capsaicin concentrations of 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, and 4 M before application of M␤CD:cholesterol and again after 60 min in the presence of M␤CD:cholesterol. Mean versus variance plots were fitted to the function in Equation 2. The open probability was calculated from the relationship in Equation 3.
Single-channel currents were recorded at room temperature in inside-out patches using the same solutions as for macroscopic currents. Pipettes were pulled from borosilicate glass, covered with dental wax to reduce stray capacitance, and fire polished to a resistance of 10 -15 M⍀ (22,23). Currents were filtered at 2 kHz and sampled at 5 kHz. Thirty to fifty 500 ms pulses obtained at 60 mV were delivered every 200 ms and collected continuously in the presence 4 M capsaicin before and after exposure to cholesterol:M␤CD. Single-channel openings were detected with the 50% threshold crossing technique (25). Single channel traces were leak subtracted with a leak template formed by the average of 5-10 null sweeps. The open probability of each sweep was calculated and plotted as a function of time.
Temperature-and Voltage-induced Currents-For heat activation experiments, we used a temperature controlling recording chamber (Bipolar Temperature Controller TC-202, Medical Systems Corp, Greenvale, NY). The bath temperature was measured with a thermistor (Warner Instruments, Hamden, CT). Experiments were conducted at 52°C for initial and cholesterol:M␤CD-exposed currents (60 min). At this temperature the maximal thermally induced current is evoked.
Currents activated exclusively by voltage were recorded at room temperature in the presence of Ca 2ϩ -free symmetrical solutions consisting of 200 mM NaCl, 3 mM HEPES (pH 7.2), and 1 mM EDTA. The voltage protocol used was: patches were initially held at 0 mV for 10 ms, and voltage was then stepped from Ϫ120 to 200 mV in 20-mV increments for 50 ms, and then returned back to 0 mV for 10 ms.
Modeling of TRPV1-Cholesterol Interaction-The structure of TRPV1 used for docking is that proposed by Fernández-Ballester and Ferrer-Montiel (26), which is a homology model based on the x-ray crystal structure of Kv1.2 (27). All dockings were carried out with the program Autodock 4.2 (28) with a Lamarckian genetic algorithm running on a desktop PC. Autodock Tools (ADT) was used to prepare files. The calculation was performed with 2.5 million evaluation steps for each conformation. The resulting 200 low-energy structures were grouped into clusters with structural deviations below 2.0-Å root mean square deviations (28). A highly populated cluster is used by Autodock as a criterion to validate the possibility of a real interaction (28). Our calculations resulted in several docking sites with similar conformations, including a single cluster of 36 conformations with cholesterol bound to the CRAC site (20,21) in the S5 helix. Only the more populated clusters with lowest energy dockings were analyzed for protein-ligand contacts (LPC/CSU server) (see Ref. 29) (data not shown).
Cholesterol and Epicholesterol Treatments-To decrease or increase the membrane cholesterol (Avanti Polar Lipids, Alabaster, AL) content the patch was incubated with M␤CD (Sigma) or 10:1 M␤CD:cholesterol, respectively (30). For all experiments the concentration of M␤CD was 3.3 mg/ml and for M␤CD:cholesterol the concentration was 0.08 mg/ml. Epicholesterol (Fig. 1B) (Steraloids Inc., Newport, RI) was added to the membrane by incubating the patch with 10:1 M␤CD:epicholesterol (0.08 mg/ml) (31,32).
Site-directed Mutagenesis-Constructs were generated by introducing mutations into the wild-type rTRPV1 and into the hTRPV1-Ile 585 channels. Point mutations were constructed by a two-step PCR procedure as previously described (33).
Detergent Isolation-Detergent-resistant membranes (DRMs) and detergent-soluble membranes (DSMs) were separated by standard techniques (34). HEK293 cells were treated with 1% Triton X-100 for 30 min at 4°C and then applied to a discontinuous (5/35/45%) sucrose density gradient and analyzed by Western blots using a TRPV1 antibody obtained from Santa Cruz Biochemicals. Raft fractions were identified by blotting of the raft lipid GM1 with HRP-labeled cholera toxin (35,36). Detergent isolations were performed with no pretreatments, as well as with the same treatments with M␤CD:cholesterol and M␤CD used in the patch clamp experiments.
Statistical Analysis-Statistical comparisons were made with an analysis of variance test. p Ͻ 0.05 was considered statistically significant. All pooled data are presented as mean Ϯ S.E.

Increased Membrane Cholesterol Decreased Capsaicin-induced rTRPV1
Currents-Cholesterol is known to modify the trafficking of TRPV1 to the plasma membrane (17), as well as regulate the activity of some classes of channels (13). To isolate the effects of cholesterol on TRPV1 in the membrane from cellular trafficking processes, we measured currents from excised membrane patches of HEK293 cells containing heterologously expressed rat or human TRPV1. These patches were incubated with M␤CD or 10:1 M␤CD:cholesterol to modify membrane cholesterol content (30).
Depletion of membrane cholesterol with M␤CD had no effect on capsaicin-induced rTRPV1 currents (supplemental Fig. S1A). However, at room temperature, augmentation of cholesterol with M␤CD:cholesterol significantly decreased capsaicin-evoked currents (Fig. 2). Representative current traces at 120 mV ( Fig. 2A) showed that after a 60-min incubation with M␤CD:cholesterol the capsaicin-induced current decreased about 70% (Fig. 2B). Control experiments showed that even after longer incubation times in the absence of M␤CD:cholesterol, the currents did not markedly decrease (rundown) (Fig. 2B). In the presence of 10:1 M␤CD:cholesterol, the capsaicin-evoked current decreased exponentially with a time constant of ϭ 29 min (shown in Fig. 4C). Importantly, we found that the decrease in current seen upon cholesterol addition was reversible. That is, after a subsequent 10-min incubation of the cholesterol-enhanced patch with M␤CD alone, the current increased to 83 Ϯ 8% of control values ( Fig. 2E and supplemental Fig. S1B).
We also measured the effects of cholesterol on capsaicin dose-response curves and current-voltage (I-V) relationships. Dose-response relationships showed that the cholesterol-induced reduction in current at 30 and 60 min incubation with M␤CD:cholesterol occurred over a range of capsaicin concentrations (Fig. 2C). The capsaicin dose-response curves were fit to the Hill equation (Equation 1) and the K 1/2 values were 317 Ϯ 33 nM (mean Ϯ S.E.) before M␤CD:cholesterol treatment, 323 Ϯ 23 nM after a 30-min M␤CD:cholesterol treatment, and 349 Ϯ 35 nM after a 60-min M␤CD:cholesterol treatment, respectively. The Hill coefficients (n H values) were 1.6 Ϯ 0.1 before M␤CD:cholesterol treatment, 1.5 Ϯ 0.1 after 30 min M␤CD:cholesterol treatment, and 1.5 Ϯ 0.3 after 60 min M␤CD:cholesterol treatment, respectively. The observation that these values of K 1/2 and n H were not statistically changed (p Ͼ 0.05) provides evidence that the binding of capsaicin to its sites on TRPV1 was not markedly altered by the addition of cholesterol. In the absence of additional cholesterol, the observed TRPV1 current-voltage relationship was, as expected, outwardly rectifying with a reversal potential of near 0 mV (Fig.  2D). Although the addition of cholesterol decreased the capsaicin-evoked current over 60 min, it did not appreciably alter the outwardly rectifying nature of the I-V relationship (Fig. 2D), as the scaling of the I-V relationship at 30 and 60 min gave shapes similar to the initial I-V relationship (see red dotted line representing normalization of data for 60 min of cholesterol exposure to its own maximal current value). Thus, the addition of cholesterol reversibly decreased the capsaicin-induced current, without changing the binding of capsaicin to the channel or the outwardly rectifying nature of the current-voltage relationship.
To further explore the underlying cause of these observed cholesterol decreases in capsaicin-induced rTRPV1 currents, we performed noise analysis on rTRPV1 channels (37) in the presence and absence of cholesterol. Fig. 3A presents the experimental results at varying concentrations of capsaicin, showing good parabolic fits to current-variance versus the mean current both before and after M␤CD:cholesterol incubation. Note that the initial slope of the curves before (black) and after M␤CD: cholesterol incubation (gray) were almost the same, suggesting that the open channels were unchanged. Values of P o , i, and N obtained from the fits shown in Fig. 3A and the relationship P o max ϭ I/iN showed that after incorporation of cholesterol both i (6.85 Ϯ 1.41 pA and 7.96 Ϯ 1.75 pA, before and after cholesterol, respectively) and P o max (0.95 Ϯ 0.17 and 0.89 Ϯ 0.33 before and after cholesterol, respectively) remained nearly constant, whereas N decreased about 70% after a 60-min incubation (N after /N before ϭ 0.32 Ϯ 0.06) (Fig. 3B). The noise data suggests that the effect of cholesterol is to reduce the number of functional channels in the patch. In single-channel recordings we confirmed that cholesterol enhancement did not alter the single-channel current and that there were not appreciable changes in open probability (Fig. 3, C and D), until the channel suddenly disappeared from the recording after a few minutes in the presence of M␤CD:cholesterol (Fig. 3, C and D).
Thus, the incorporation of cholesterol into the membrane markedly and reversibly decreased capsaicin-induced rTRPV1 currents. We next considered if these current decreases were due to a specific cholesterol interaction with rTRPV1 by measuring currents for: 1) membranes augmented with epicholesterol and 2) membranes containing rTRPV1 with point mutations in the S5 helix at the interface with the plasma membrane.

Epicholesterol Had No Effects on Capsaicin-induced rTRPV1
Currents-The incubation of a membrane patch with M␤CD: epicholesterol produced no significant decreases in capsaicin-evoked currents (Fig. 4, A and B) over the same time course used for cholesterol treatment (Fig. 4C). Moreover, the observed capsaicin dose-response relationship overlapped in  the absence and presence of epicholesterol (Fig. 4D), giving K 1/2 ϭ 260 Ϯ 16 nM, n H ϭ 1.5 Ϯ 0.4, and K 1/2 ϭ 234 Ϯ 34 nM, n H ϭ 1.6 Ϯ 0.11 after M␤CD:epicholesterol incubation for 30 and 60 min, respectively. That cholesterol and epicholesterol had significantly different effects on capsaicin-induced rTRPV1 currents (Fig. 4) suggests that there is stereospecificity in the ability of cholesterol to inhibit the capsaicin-evoked currents.
Currents from rTRPV1 with Point Mutations in Transmembrane Helix S5-We next measured capsaicin-induced currents from control and M␤CD:cholesterol-treated patches containing heterologously expressed rTRPV1 with point mutations at different positions in transmembrane helix S5 where we had found a CRAC sequence (Fig. 7A). Fig. 5 shows the effects of changing the positively charged arginine (Arg) with the negatively charged aspartic acid (Asp) at residue 579 (near the edge of the transmembrane region), and the aromatic phenylalanine (Phe) with the polar amino acid glutamine (Gln) at position 582 (in the hydrophobic region). Cholesterol augmentation reduced the measured currents for both rTRPV1-R579D and rTRPV1-F582Q (Fig. 5, B-D). Both mutants gave similar capsaicin dose-response curves (Fig. 5E). The coefficients before M␤CD:cholesterol treatment were K 1/2 ϭ 192 Ϯ 14 nM, n H ϭ 1.6 Ϯ 0.25 and 237 Ϯ 49 nM, n H ϭ 1.7 Ϯ 0.3 for rTRPV1-R579D and rTRPV1-F582Q, respectively, and after treatment with M␤CD:cholesterol K 1/2 ϭ 218 Ϯ 32 nM, n H ϭ 1.5 Ϯ 0.15 and 250 Ϯ 32 nM, and n H ϭ 1.7 Ϯ 0.23 for rTRPV1-R579D and rTRPV1-F582Q, respectively. Moreover, when residue 579 was changed to an electically neutral amino acid (R579Q), the effects of cholesterol on capsaicin-activated currents were augmented with respect to the R579D mutation, thereby pointing to the importance of the presence of the charge at this site (Fig.  5, A and D).
Representative current traces (Fig. 6A) and normalized currents (Fig. 6B) from membrane patches containing rTRPV1-L585I showed that the capsaicin-induced current was essentially unchanged after M␤CD:cholesterol treatment, as were the capsaicin dose-response curves (Fig. 6C), giving K 1/2 ϭ 382 Ϯ 16 nM, n H ϭ 1.6 Ϯ 0.03 before treatment and K 1/2 ϭ 396 Ϯ 5 nM, n H ϭ 1.5 Ϯ 0.02 after M␤CD:cholesterol for 60 min. Thus, this single point mutation at residue 585 in S5 essentially abolished the large inhibitory cholesterol effects of wildtype rTRPV1 (Fig. 2). Consistently, the addition of cholesterol did not noticeably alter the shape of the I-V relationship in this mutant (data not shown).
Comparisons of rTRPV1 and hTRPV1 Variants-Human TRPV1 has two common variants containing natural point mutations at residue 585, with isoleucine (variant 1) or valine (variant 2) at that residue (38) (Fig. 7A). Cholesterol augmentation with M␤CD:cholesterol had quite different effects for variant 1 and for this variant mutated at position 585 to have the same S5 sequence as variant 2 (hTRPV1-I585V). Specifically, cholesterol augumentation had no appreciable effect on capsaicin-induced currents for membranes containing hTRPV1-Ile 585 (Fig. 7, B and D), whereas cholesterol augumentation reduced by about 35% the currents for membranes with hTRPV1-I585V (Fig. 7, C and D). However, the decrease in currents produced by cholesterol augmentation was greater for wild-type rTRPV1 than for either hTRPV1 variant (Fig. 7D). The values of K 1/2 were not significantly different for the two hTRPV1 variants before (217 Ϯ 28 and 232 Ϯ 26 nM for hTRPV1-Ile 585 and hTRPV1-I585V, respectively) and after treatment with M␤CD:cholesterol (185 Ϯ 30 and 204 Ϯ 37 nM for hTRPV1-Ile 585 and hTRPV1-I585V, respectively) (Fig. 7E). Incubation with M␤CD:cholesterol did not alter the shape of the I-V relationship for either hTRPV1-Ile 585 or hTRPV1-I585V (Fig. 7, F and G).
We next measured capsaicin-induced currents and cholesterol sensitivity for hTRPV1-I585L, which was of interest because it has the same S5 sequence as wild-type rTRPV1 (Fig.   (Fig. 8A) to about the same extent and with the same time course as wild-type rTRPV1 (Fig.  8B). Moreover, the Hill parameters were unchanged from that of rTRPV1 (Fig. 8C) and so was the I-V relationship (data not shown). These electrophysiological data demonstrate the importance of residue 585 in S5 for the cholesterol sensitivity of TRPV1. Importantly, the inhibitory effects of cholesterol enhancement on capsaicin-induced rTRPV1 currents were also found in the absence of capsaicin when rTRPV1 was activated thermally at 52°C (supplemental Fig. S2, A and B) as well as by voltage (supplemental Fig. 2, C and D).
Model of Cholesterol-binding Site-The three residues we show here to be important for cholesterol binding are predicted to be in the S5 transmembrane segment and would be facing the same side of this ␣-helix. In other proteins (20) a cholesterol consensus motif has been proposed to be formed by a cleft in the membrane protein located at a membrane interfacial region and containing at least a positively charged residue and an aromatic residue.
To get an idea of the physical appearance of the CRAC sequence in TRPV1, we examined the two available homology models of the transmembrane domains of TRPV1 (26, 39). In both cases, amino acids Arg 579 , Phe 582 , and Leu 585 are facing the lipid-exposed side of the S5 helix and the sequence is very similar to the cholesterol consensus motif in ␤-adrenergic receptors (20). We carried out a docking experiment using the model of Fernández-Ballester and Ferrer-Montiel (26) as a docking template and cholesterol as the ligand. Fig. 9 shows the 5 best structures of cholesterol docked in the S5 of TRPV1. These correspond to a larger cluster (36 of 200 results with root mean square deviations below 2 Å) interacting at the site identified by mutagenesis, calculated with the Autodock suite. In this model of cholesterol binding, the OH group makes a possible electrostatic interaction with Arg 579 and the flat ␣-face of cholesterol is stacked against Phe 582 , which is an interaction seen very often in crystal structures of proteins bound to cholesterol and other sterols (20,40). The important interaction with residue Leu 585 seems to be mediated by the aliphatic tail of cholesterol, which enters an incipient cavity formed by Leu 585 and other hydrophobic residues. When Leu 585 is mutated to an Ile residue, the cavity is reduced in size, suggesting an explanation for the lack of effect of cholesterol in L585I and the Ile 585 human isoform. Modeling also suggests a reason why epicholesterol might not be able to favorably interact with TRPV1. Epicholesterol can be docked in the same general site as cholesterol with poor clustering results (9/200 below 2-Å root mean square deviations, see "Experimental Procedures") (supplemental Fig. S3). The aliphatic tail of epicholesterol makes an interaction with Leu 585 , but the position of the OH group on the ␣-face (Fig. 9) increases its van der Waals volume, as compared with cholesterol, making it difficult for Phe 582 to interact with the ␣-face of epicholesterol. The increased volume also makes the simultaneous interaction with all three residues more energetically unfavorable.
TRPV1 Was in Detergent-soluble Membranes-Because previous electrophysiological studies have suggested that rTRPV1 is primarily localized in membrane raft microdomains (17,18), we performed studies to determine whether rTRPV1 extracts with DSMs or DRMs from HEK293 cells. DSMs, which contain relatively low concentrations of cholesterol, and DRMs, which have higher cholesterol concentrations, are thought to corre- spond to putative non-raft and raft microdomains, respectively (34). Both before and after M␤CD or M␤CD:cholesterol treatment, rTRPV1 was found in DSM fractions and there was no detectable rTRPV1 in DRM fractions (supplemental Fig. S4).

DISCUSSION
In these studies we have shown that cholesterol, an important component of cell plasma membranes, modulates the function of the cation-selective channel TRPV1. Although depletion of membrane cholesterol had no effect on capsaicininduced TRPV1 currents, augmentation of cholesterol markedly reduced these currents in both wild-type rat TRPV1 and one of the major human variants of TRPV1. Specifically, we found that the addition of cholesterol inhibits TRPV1 by binding to specific sites along the S5 helix having a putative CRAC motif. These data suggest the extent cholesterol alters the function of TRPV1 is dependent on the species and membrane cholesterol content.
All of the electrophysiology experiments reported in this article were performed on patches excised from HEK293 cells with heterologously expressed rTRPV1 channels. Therefore, the observed effects of membrane cholesterol depletion or augmentation had to be due to changes within this excised membrane, rather than to possible modifications in cell trafficking. We found that cholesterol depletion had no effect on the currents (Fig. 2), which contrasts to electrophysiological data from whole cells where cholesterol depletion reduced capsaicin-induced currents (17,18). We argue that this means that the observed results in whole cells upon M␤CD treatment were due to the role of cholesterol in membrane trafficking in the cell, consistent with the observation that cholesterol depletion decreased the concentration of rTRPV1 in the plasma membrane (17). Moreover, this interpretation is consistent with the results of Liu et al. (41) who found for whole cells that cholesterol removal caused no appreciable change in temperatureinduced rTRPV1 currents over a range of temperatures (44 to 50°C).
In contrast to the non-effect of cholesterol depletion, cholesterol augmentation significantly and reversibly reduced capsaicin-induced rTRPV1 currents (Fig. 2). In terms of molecular mechanism, the similarity in shape of the dose-response curves before and after cholesterol augmentation (Fig. 2C) suggests that cholesterol addition had no appreciable effect on the affinity of capsaicin to its binding site. The similar shapes of the I-V curves (Fig. 2D) suggest that there was little change in the voltage-sensing domain. Importantly, the noise analysis experiments revealed that the decrease in current with cholesterol addition did not arise from a significant change in the conducting channels as reflected in no appreciable changes in the single channel current (i) or open probability (P o ), but rather a decrease in the number of agonist-responsive channels in the patch (Fig. 3B). Our single-channel recordings also suggest that cholesterol sequesters the channel in a non-conducting state. These data could be interpreted to mean that cholesterol impedes channel opening (see Refs. 19 and 42).
Unlike cholesterol, epicholesterol did not modify the capsaicin-evoked current (Fig. 4), indicating that the effect of cholesterol is not due to a nonspecific membrane effect, but rather to a stereospecific binding site. Similar results with other channels, including inward-rectifier K ϩ channels and large-conductance voltage/Ca 2ϩ -gated K ϩ channels, have also been interpreted to mean that the effect of cholesterol are not due to an indirect modification of lipid bilayer properties, but rather to a stereospecific interaction at the surface of the protein (19,32). We consequently explored rTRPV1 for possible cholesterolbinding sites. One established sequence selective for binding cholesterol is the putative CRAC motif, first identified in the FIGURE 7. Cholesterol enhancement of responses for two human TRPV1 variants. A, sequence alignment of the S5 transmembrane segment of rTRPV1 and two common hTRPV1 variants, variant "1" with Ile 585 and variant "2" with Val 585 . The CRAC motif is noted. The red box highlights amino acid 585. Representative current traces from inside-out excised patches elicited by 4 M capsaicin before (black) and after 60 min (red) incubation with 10:1 M␤CD:cholesterol for B, hTRPV1-I585, and C, hTRPV1-I585V. D, bar chart of average currents elicited by 4 M after 60 min of exposure to M␤CD:cholesterol for hTRPV1-Ile 585 (0.97 Ϯ 0.05, n ϭ 9) and hTRPV1-I585V (0.65 Ϯ 0.04, n ϭ 8) variants and rTRPV1 (0.28 Ϯ 0.02, n ϭ 31). M␤CD:cholesterol data were normalized to the initial currents obtained in the presence of 4 M. * denotes significant differences with respect to hTRPV1-Ile 585 (p Ͻ 0.05) and ** denotes a difference of rTRPV1 with respect to both human variants (p Ͻ 0.01). E, capsaicin sensitivities before (black, K 1/2 ϭ 217 Ϯ 28 nM, n H ϭ 1.4 Ϯ 0.08 and 232 Ϯ 26 nM, n H ϭ 1.4 Ϯ 0.07 for hTRPV1-Ile 585 and hTRPV1-I585V, respectively) and after treatment with M␤CD:cholesterol (red, K 1/2 ϭ 185 Ϯ 30 nM, n H ϭ 1.5 Ϯ 0.08 and 204 Ϯ 37 nM, n H ϭ 1.4 Ϯ 0.09 for hTRPV1-Ile 585 and hTRPV1-I585V). n ϭ 5 for each case. F and G, current to voltage relationships for hTRPV1-Ile 585 and hTRPV1-I585V variants, respectively, before (black symbols) and after (red symbols) a 60-min M␤CD:cholesterol treatment. Dotted curves are the data after cholesterol application normalized to their maximal value.
benzodiazepine receptor (43) and later identified in many other proteins and peptides (40,44,45). This motif is defined as a sequence pattern of -(Leu/Val-(X) 1-5 -Tyr-(X) 1-5 -Arg/Lys)-, in which (X) [1][2][3][4][5] represents between one and five residues of any amino acid (40,45). Cholesterol interacts with the CRAC motif with both attractive van der Waals interactions between hydrophobic surfaces and electrostatic interactions between the positively charged Arg or Lys residue and the cholesterol -OH group (40,45). In a study of a peptide from the fusogenic gp41 protein of HIV-1, Epand et al. (40) found that the single Leu to Ile substitution in the CRAC motif (from LWYIK to IWYIK) resulted in no preferential interaction with cholesterol (33). Another cholesterol binding motif, called the cholesterol consensus motif, also found in many proteins, has three sites (Trp/ Tyr)-(Ile/Val/Leu)-(Lys/Arg) on one helix and one site (Phe/ Tyr/Trp) on an adjoining helix (20). For example, in the crystal  structure of the human ␤ 2 -adrenergic receptor, Hanson et al. (20) found that cholesterol binds in a shallow groove between these helices. As with the CRAC motif, key features of the cholesterol consensus motif are the van der Waals interactions of the cholesterol with the aromatic residues and electrostatic interactions with positively charged residues (20).
A cholesterol-binding motif in rTRPV1 is located in the S5 helix, which contains a CRAC sequence from residue 579 to 586 (Fig. 7A). Our experiments showing different sensitivities to cholesterol augmentation of specific rat mutants (Figs. 5 and 6) and human variants (Figs. 7 and 8) provide strong evidence for the presence of a cholesterol-binding site in helix S5. Two points here are worth discussing. First, the abrogation of most of the sensitivity to cholesterol augmentation by the rTRPV1-L585I mutation (Fig. 6) and the restoration of this sensitivity by the inverse mutation in hTRPV1-I585L (Fig. 8), shows the importance of hydrophobic interactions for this cholesterol effect. Given that the Leu to Ile mutation is rather subtle, its effect also highlights the importance of steric factors in the interaction. Second, given that the CRAC sequence contains a positively charged (Arg or Lys) residue, it might seem surprising that the R579D mutation (positive to negative charge with different side chain volumes) retained sensitivity to cholesterol augmentation, albeit reduced (Fig. 5B). However, the interaction with the -OH dipole can occur with charges of either sign (20) in a manner that may be enough to stabilize the interaction of cholesterol with the protein. Our data imply that in the case of rTRPV1, the positive charge Arg 579 contributed a fraction of the stabilization energy because charge neutralization (R579Q) rendered TRPV1 more sensitive to cholesterol than charge reversal (R579D) (Fig. 5).
Recently, Rosenhause-Dankster et al. (46) have further characterized the interaction of Kir2.1 channels with cholesterol. By using a combination of electrophysiological recordings and mutagenesis as well as modeling based on the crystal structure of Kir2.1 they have concluded that cholesterol does not appear to have a typical binding pocket in Kir2.1. Rather they found that a more complex belt-like structure is formed to contain the apex of the flexible G-loop near the interface with the transmembrane region, affecting channel gating in a manner that stabilizes the closed state of the channel (46). The question remains whether other regions of TRPV1 also interact with the S5 CRAC motif to form a more complex cholesterol-stabilizing structure.
Although there are many variants of human hTRPV1 subunits (47), two very common alleles (I and V) were identified at position 585 (see Fig. 7A). In a genetic and functional screening of hTRPV1 variants obtained from a human population, Cantero-Recasens et al. (48) found that at amino acid 585, 51% are heterozygous (Ile/Val), 15% are homozygous for Val/Val, and 34% homozygous for Ile/Ile. They found these variants could be activated by capsaicin, which we have confirmed. Moreover, we found that the two variants had different responses to cholesterol addition (Fig. 7, B-D), which represents, to our knowledge, the first demonstrated functional difference between these two variants. Given that rTRPV1-L585I was insensitive to cholesterol augmentation (Fig. 6), it was gratifying to find that hTRPV1-Ile 585 , with the same S5 sequence as rTRPV1-L585I, was also insensitive to cholesterol augmentation (Fig. 7B). Thus, a single amino acid substitution dramatically altered the sensitivity of TRPV1 to cholesterol among species.
It follows that the observed decreases in capsaicin-induced currents by cholesterol augmentation were due to specific interactions with a cholesterol-binding site in the S5 transmembrane helix of TRPV1. We have identified critical amino acids on the S5 helix that greatly affected how TRPV1 responds in the presence of cholesterol. Based on the similarities of capsaicin dose-response and I-V curves before and after cholesterol addition, the cholesterol effects are not due to changes in capsaicin binding or channel selectivity. Rather we argue that cholesterol binding must cause a conformational change in TRPV1 that stabilizes the closed state(s) (41) of the channel. A similar conclusion was reached in a noise analysis study of Kir2.1 channels (19).
The time required to inhibit rTRPV1 with M␤CD:cholesterol incubation was greater than the time for recovery of the current with the same concentration of M␤CD ( Fig. 2 and supplemental Fig. S1). Let us consider possible reasons for this observation. First, the membrane bilayer contains many more cholesterol-binding sites (each monolayer with its raft and nonraft regions) than do the TRPV1 channels embedded in the bilayer. Thus, from mass action principle, it follows that cholesterol would initially preferentially partition into the bilayer sites and the channel sites would take longer to fill. The TRPV1 channel is formed from tetramers, and cholesterol has to partition between subunits to get to its sites within the protein. Moreover, it may take all four sites to be occupied for the channels to fully close. We argue that this latter process may be the rate-limiting step. One possibility for the faster recovery time is that the removal of cholesterol from only one (or two) of the subunits may be sufficient for the channel to reopen.
Our result (see supplemental Fig. S4) that rTRPV1 is primarily located in cholesterol-poor DSMs (thought to be related to non-raft microdomains) is similar to results for TRPC3 and TRPC6 (49), but in contrast to other TRP channels, such as TRPC1, TRPC4, and TRPC5 (49) and TRPM8 (42), that have been associated with cholesterol-rich DRMs. Although the reasons for these different microdomain locations are not clear, the different locations can impact on the TRP channels' response to cholesterol. Thus, the cholesterol-rich raft microdomain location of TRPM8 helps to explain why M␤CD treatment increases cold-evoked TRPM8 currents in cells (42), whereas the non-raft localization of rTRPV1 (data seen in supplemental Fig. S4) helps explain why M␤CD treatment does not modify the capsaicin-induced current of this channel in HEK293 cells (Fig. 2 and supplemental Fig. S1).
TRPV1 is known to serve many physiological functions and is also involved in pathological conditions (for review, see Ref. 1). Here we consider the physiological significance of the observations that the activity of both rTRPV1 and hTRPV1 can be diminished by cholesterol. This is relevant because different cells have different cholesterol concentrations, with the cholesterol unequally partitioned between plasma and organelle membranes. Moreover, many plasma membranes contain transient cholesterol-rich raft and cholesterol-poor non-raft microdomains. Here we have shown that rTRPV1 is located in DSM (non-raft) fractions (supplemental Fig. S4). We speculate that the activity of TRPV1 could be regulated by expression of the Val/Val or Ile/Ile variants in cells with different cholesterol concentrations, or by transferring them between non-raft (low cholesterol) and raft (high cholesterol) microdomains within a given plasma membrane. For example, after exposure to bradykinin TRPM7 channels have been shown to relocalize from cholesterol-poor (non-raft) to cholesterol-rich (raft) domains (50). In the future, these variants may prove to display a physiologically important role in humans containing both variants of TRPV1 under conditions of hypercholesterolemia, for instance. Our data also argue in favor of a species-specific sensitivity of TRPV1 to cholesterol as happens for the Kir6 channel, which shows differences in cholesterol sensitivity in a porcine model (51) as compared with a rabbit model (52). These findings and our own suggest that the same channel can be differentially regulated by cholesterol in different species.