Dissecting the functions of conserved prolines within transmembrane helices of the D2 dopamine receptor

Abstract

G protein–coupled receptors (GPCRs) contain a number of conserved proline residues in their transmembrane helices, and it is generally assumed these play important functional and/or structural roles. Here we use unnatural amino acid mutagenesis, employing α–hydroxy acids and proline analogs, to examine the functional roles of five proline residues in the transmembrane helices of the D2 dopamine receptor. The well–known tendency of proline to disrupt helical structure is important at all sites, while we find no evidence for a functional role for backbone amide cis–trans isomerization, another feature associated with proline. At most proline sites, the loss of the backbone NH is sufficient to explain the role of the proline. However, at one site – P210^5.50 – a substituent on the backbone N appears to be essential for proper function. Interestingly, the pattern in functional consequences that we see is mirrored in the pattern of structural distortions seen in recent GPCR crystal structures.

INTRODUCTION

Proline stands apart from the other 19 canonical amino acids. Its cyclic side chain uniquely shapes protein structure and facilitates protein dynamics. As proline's side chain substantially restricts its backbone psi angle and removes a backbone hydrogen bond donor, proline disrupts the regular structures of both α helices and β sheets and can introduce flexibility into these structures.(1–4) Further, proline's backbone N–substitution biases its peptide bonds to the cis conformer, relative to other peptide bonds. The cis–trans isomerization of proline peptide bonds is well documented to mediate protein folding and mature protein function.(5) Here we examine which of the above properties contribute to function of the D2 dopamine G protein–coupled receptor (GPCR) at conserved proline sites within its transmembrane α helices (TMs), using conventional and unnatural amino acid mutagenesis.

Transmembrane proline residues are a characteristic feature of GPCRs and are found in five of the D2 receptor's seven TMs. These prolines are also highly conserved; three (P201^5.50, P388^6.50, and P423^7.50) are the most conserved residue of their respective helix among Class A GPCRs and the remaining two (P89^2.59, P169^4.59) are conserved among aminergic GPCRs. (Superscripts refer to Ballesteros–Weinstein numbering, in which the most conserved residue of helix X is denoted X.50).(6) A wealth of conventional mutagenesis studies has already established that these prolines play some significant functional role in GPCRs.(7–14) Proline kinks have long been hypothesized to mediate the helical movements involved in GPCR activation, acting as pivot points, hinges, and/or swivels to expose a G protein binding site at the intracellular end of the helical bundle.(15, 16) A recent crystal structure of the closely related D3 dopamine receptor confirms that the prolines investigated here are associated with helical kinks, especially prominent in TMs 2, 6, and 7 (Figure 1).(17) P169^4.59 sits only one helical turn from the N–terminus of TM4, although the pre–proline turn is indeed kinked relative to the rest of the helix. Interestingly in the D3 and other GPCR crystal structures, TM5 is only slightly kinked, but instead possesses a prominent bulge in the helical turn preceding P201^5.50.(17–22)

Figure 1.

Two views of the crystal structure of the D3 dopamine receptor (pdb: 3PBL), highlighting the prolines considered here.(15) Numbering is for the corresponding residues in the D2 receptor.

Proline's ability to distort helices most obviously comes from its lack of backbone hydrogen bond donor ability, which frees its would–be hydrogen bond acceptor, the backbone carbonyl of the residue i–4 from proline (Figure 2). Additionally, the proline side chain introduces a steric clash with the i–4 carbonyl. The net effect is usually manifested as a kink, which often frees the carbonyl i–3 to proline as well.(2, 23, 24)

Figure 2.

Hydrogen bonding patterns in an α-helix for (left to right) a typical amino acid; proline; an N–methyl amino acid; and an α-hydroxy acid.

Since the 19 other canonical amino acids are ill suited to probe the unique properties of proline, unnatural amino acid mutagenesis is an especially valuable tool to dissect the basis for a given proline's functional importance. As structural information on GPCRs continues to accrue, the subtle and high precision probes provided by unnatural residues will become increasingly valuable. Tests of detailed structural environments and specific functional roles are significantly facilitated by the unnatural amino acid methodology.

Here we compare the effect of replacing each of 5 TM prolines of the D2 receptor with unnatural α–hydroxy acids, and an N–methyl amino acid (Figure 2). We also consider cyclic proline analogs and conventional amino acids that correspond to the α–hydroxy acids. Unnatural α–hydroxy acids introduce a backbone ester that, like the proline peptide bond it replaces, is not a hydrogen bond donor (Figure 2). N–Me–Ala, effectively proline lacking its side chain C_γ, shares N–substitution (and thus lack of hydrogen bond donor ability) with proline, but has greater conformational freedom. Finally, proline analogs that vary the size of the ring or introduce substituents can probe tolerance for subtle changes to the proline side chain as well as cis–trans isomerization. Pipecolic acid (Pip) and azetidine–2–carboxylic acid (Aze), six– and four–membered ring analogs of proline, respectively, have a greater cis bias relative to proline, while 2–methylproline (2–Me–Pro) is substantially more trans–biased.(5)

Previously, we have characterized three distinct phenotypes for functionally important proline residues in ligand–gated ion channels. In the M1 transmembrane helix of both the α subunit of the muscle–type nicotinic acetylcholine receptor and the 5–HT_3A receptor, a conserved proline in the middle of the helix can be replaced by a range of α–hydroxy acids, giving essentially wild–type behavior.(25, 26) In contrast, incorporating any canonical amino acid other than proline gave a nonfunctional receptor. A second phenotype was seen in the M2–M3 loop of the 5–HT_3A receptor.(27) In this system, cis–trans isomerization of a proline peptide bond is critical for receptor gating (see also (28, 29)), as revealed by a series of cyclic proline analogs with varying cis preferences. Finally, we have characterized an important proline–aromatic interaction in the Cys loop of the muscle–type nicotinic acetylcholine receptor, in which the hydrophobicity of a Phe side chain preceding proline is important for receptor function.(30) Using the various unnatural amino acids discussed above, we sought to determine which of these phenotypes is most applicable to the transmembrane prolines of the D2 receptor.

RESULTS AND DISCUSSION

Experimental Approach

We utilized a recently optimized electrophysiological readout of the D2 receptor that yields quantitative dose–response relationships from receptor activation of coexpressed GIRK1/GIRK4 channels in Xenopus laevis oocytes and is amenable to unnatural amino acid mutagenesis by nonsense suppression.(31) Here we probe five new sites of the D2 dopamine receptor with an array of unnatural residues, establishing the generality of this protocol for evaluating GPCR function.

All unnatural amino acid and hydroxy acid mutations as well as the valine mutants noted in Table 1 were generated by nonsense suppression, while all other mutants were expressed conventionally. A wild–type rescue experiment (incorporating Pro by nonsense suppression) at each proline site yielded an EC₅₀ within two–fold of the wild–type value (see Supporting Information), validating the ability to perform nonsense suppression at all residues considered here.

Table 1.

Conventional amino acid, hydroxy acid, and N–Me–Ala mutations^a

Site	EC50 (nM)		EC50 (nM)
wild type		39
Pro892.59	Gly	35	N–Me–Ala	46
	Ala	65	Aah	140
	Valb	53	Vah	47
	Leu	110	Lah	140
Pro1694.59	Gly	180	N–Me–Ala	40
	Ala	1300	Aah	59
	Valb	1100	Vah	76
	Leu	>2000c	Lah	100
Pro2015.50	Gly	>2000c	N–Me–Ala	17
	Ala	600	Aah	180
	Val	>2000c	Vah	330
	Leu	NDd	Lah	NDd
Pro3886.50	Gly	67	N–Me–Ala	92
	Ala	230	Aah	58
	Valb	>2000c	Vah	190
	Leu	1000	Lah	35
Pro4237.50	Gly	61	N–Me–Ala	22
	Ala	140	Aah	21
	Valb	170	Vah	45
	Leu	NDd	Lah	NDd

^aReproducibility of the data is discussed in text. Errors for the goodness of fit to the Hill equation are presented in the Supporting Information.

^bConventional mutant generated by nonsense suppression

^cAttempts to measure large values of EC₅₀ are compromised by background responses to dopamine at concentrations >100 μM.

^dNo dopamine–induced current detected for dopamine concentrations up to 1 mM

Because of the indirect nature of the assay, interpretation of EC₅₀ requires some caution. A detailed discussion of this issue has been presented elsewhere.(32) For present purposes, we consider EC₅₀ values that differ by a factor of greater than 2 to be distinguishable in this assay.

General Observations

We began our investigation of the D2 receptor's transmembrane prolines by incorporating α–hydroxy acids at each site. For the most part, incorporation of an α–hydroxy residue is not strongly perturbing; all but one of the hydroxy acid mutants we could characterize gave EC₅₀s within 5–fold of wild type (Figure 3, Table 1). In addition, each conventional amino acid residue is typically much more perturbing than its unnatural α–hydroxy acid analog. This trend suggests the importance of lacking a hydrogen bond donor at these sites. N–Me–Ala provides an alternative way to remove the backbone NH group, and it is also well–tolerated at all sites. Given the results for α–hydroxy residues and N–Me–Ala, it is not surprising that various cyclic proline analogs, which are much more similar to proline, are well–tolerated at all transmembrane proline sites, yielding EC₅₀s within 5–fold of wild type (Table 2).

Figure 3.

Results for incorporating α-hydroxy acids; their amino acid analogs; Gly; and N–Me–Ala. At four sites, indicated by a ≈, EC₅₀ is too large to determine accurately. * = no current detected.

Table 2.

EC₅₀s (nM) for proline analog mutations^a

Site	Pip	Aze	2–Me–Pro
Pro892.59	26	26	64
Pro1694.59	19	25	37
Pro2015.50	34	41	160
Pro3886.50	82	47	130
Pro4237.50	80	15	130

^aReproducibility of the data is discussed in text. Errors for the goodness of fit to the Hill equation are presented in the Supporting Information.

Mutations of proline to the conventional amino acids Ala, Val, and Leu range from modestly perturbing to dramatically perturbing (EC₅₀ shifts of 3.5–fold to >2000–fold from wild type), with the exception of P89^2.59, discussed below. Generally, Gly is less perturbing than the other natural amino acids, with the glaring exception of P201^5.50G. These conventional mutants provide a useful reference to which we can compare the α–hydroxy mutants.

The P89^2.59 site, despite >70% conservation among aminergic receptors, accepts diverse mutations with only modest functional consequences. All mutations made, including conventional mutations, give wild–type EC₅₀s or relatively small shifts from wild type. Evidently this site, which is located in a helix not intimately involved in ligand binding nor conformational changes, does not require the unique side chain and hydrogen bonding properties conferred by proline, and we will not consider it further here.

P169^4.59, P388^6.50, and P423^7.50

For the TM4, 6, and 7 proline sites, we observe hydroxy acid mutations (to Aah, Vah, or Lah; “ah” signifying α–hydroxy) in all cases to be less perturbing than the corresponding conventional amino acid mutations (to Ala, Val, or Leu) (Figure 3, Table 1). In fact, most of the hydroxy acid mutants for these sites (P169^4.59Aah, P169^4.59Vah, P388^6.50Aah, P388^6.50Lah, P423^7.50Aah, and P423^7.50Vah) have EC₅₀s that are not meaningfully different from wild type (i.e., within the margin of error we can expect from this assay). As hydroxy acids and proline both lack backbone hydrogen bond donors, this property alone may account for proline's functional role at these sites. However, hydroxy acids do introduce additional perturbations; they weaken the hydrogen bond acceptor strength of the preceding residue's backbone carbonyl and create an electrostatic repulsion between the introduced main chain oxygen and the i–4 carbonyl.

An alternative way to remove the backbone NH of a peptide bond is with N–Me–Ala, which contains a fragment of the proline ring and does not significantly perturb the backbone carbonyl. At the TM4, 6, and 7 proline sites, N–Me–Ala is not largely perturbing, being essentially equivalent to Aah and producing EC₅₀s within 2.5–fold of wild type (Figure 3, Table 1). This result suggests that the carbonyl mutation associated with α–hydroxy residues is not a large factor in these helices.

Given the tolerance for α–hydroxy and N–methyl residues, it is not surprising that the proline analogs Aze and Pip are also well tolerated in TM4, 6, and 7 (Table 2). As described in previous work, Aze and Pip show different intrinsic cis–trans preferences than Pro.(5) The minimal impact of these mutations indicates that cis–trans isomerization is not an essential component of receptor function. While varying ring size does not substantially alter function, introducing a methyl substituent at proline's α carbon (2–Me–Pro) does produce measurable EC₅₀ shifts at the 388^6.50 and 423^7.50 sites, but not at 169^4.59.

Taken together, the mutagenesis results above indicate that the key function of proline at these sites is to disrupt the α helix by removing the key backbone NH. Any strategy that accomplishes this, including cyclic residues, α–hydroxy residues, or N–methyl residues, produces receptors with essentially wild type function.

Interestingly, Gly is less perturbing than the other conventional amino acids at these sites (Figure 3, Table 1). Glycine is well known to have an especially low propensity for helical structures relative to other amino acids (second–lowest only to proline),(33, 34) a property believed to stem both from its high conformational entropy and from its minimal burial of solvent accessible surface area in helices.(35) The good tolerance of Gly at these sites suggests a generic functional requirement for a helix breaker at these sites. Proline appears to accomplish this by lack of a backbone hydrogen bond donor, but Gly can achieve the same end by different means. We do note that while the P423^7.50G mutant gave a near–wild–type EC₅₀, it produced signals that were generally small, suggesting either poor expression or diminished receptor efficacy, possibilities we cannot differentiate with our assay.

Because the α–hydroxy residues do not seriously compromise receptor function at these sites (the way their α–amino analogs do), we can use the α–hydroxy data to consider the effects of side chain variation. The largest perturbations are seen for Lah at P169^4.59 and Vah at P388^6.50. This suggests that the 169^4.59 site is sensitive to size, while 388^6.50 is sensitive to β branching. While no P423^7.50 mutants produced dramatic EC₅₀ shifts, cells injected with appropriate mRNA and tRNA for P423^7.50L and P423^7.50Lah mutations gave no response in our assay, suggesting this site is especially sensitive to steric bulk. Supporting this notion, the P423^7.50 side chain points directly at TM1 in the D3 receptor crystal structure, while P169^4.59 and P388^6.50 are directed more toward the lipid bilayer.(17)

P201^5.50

The 201^5.50 site shows a qualitatively distinct pattern from the other prolines considered here. As before, α–hydroxy residues are less perturbing than their α–amino analogs (Figure 3, Table 1). Unlike the other proline sites, however, no hydroxy acid mutation to P201^5.50 yields a wild–type EC₅₀. P201^5.50Aah and P201^5.50Vah are loss of function by 4– and 8–fold, respectively (we were unable to characterize the P201^5.50Lah mutant). Interestingly, N–Me–Ala is minimally perturbing, perhaps even gain of function, and the proline analogs Pip and Aze are wild type (Table 2). Taken together, these data suggest that simply deleting the hydrogen bonding ability of the backbone NH is not sufficient to produce a maximally functional receptor. A substituent on the N is also necessary, and it can be either a ring (Pro, Pip, Aze) or a methyl (N–Me–Ala).

Also in contrast to the TM4, 6, and 7 sites, Gly is highly disruptive at 201^5.50, more so than Ala (Figure 3, Table 1). Together with the fact that no hydroxy acid tested gave a wild–type EC₅₀ at this site, these findings suggests a more specific need for proline here, not merely a generic requirement for a helix–breaking residue.

Phenotypes Correlate with Helix Distortions

In addition to sharing a functional phenotype, the TM4, 6, and 7 prolines all seem to play a similar structural role (Figure 1). In various structures of GPCRs, TM4, 6, and 7 all display a significant kink associated with the proline.(17–21) Movements of helices 6 and 7 are considered to be critical to receptor activation, with the kinks playing a prominent role.(16) Indeed, in a recent crystal structure thought to represent an active conformation of the β2 adrenergic receptor, displacement of TM6 is accomplished by a slight “unwinding” of the helical turn preceding the proline at position 6.50.(36) The movement originates at residue 6.46, the would–be backbone hydrogen bond acceptor to position 6.50, potentially providing a direct link between proline 6.50's lack of hydrogen bond donor ability and functionally important helix flexibility.

The 201^5.50 site gave a unique functional phenotype and indeed, it is clear from Figure 1 that the structural perturbation associated with the TM5 proline is distinct from what is seen in helices 2, 4, 6, and 7. Instead of a prominent kink, helix 5 has a bulge. It has been proposed that this bulge is functionally significant, as residues in the bulge directly contact ligands in the agonist binding site. Specifically, the side chain of S197^5.46, which is i–4 relative to P201^5.50 and located at the center of the bulge, is thought to contribute a hydrogen bond that is important to agonist binding.(37, 38) Evidently, the steric effect of proline's N–substitution is key to producing this structural bulge, consistent with the functional requirement for N–substitution we found at this site.

Conclusions

We have probed five conserved proline sites located in the transmembrane region of the D2 receptor. Transmembrane prolines are not uncommon, and are frequently considered to play important functional roles by introducing structural perturbations and/or increased conformational flexibility to the helix.(4, 15)

In all functionally important transmembrane proline sites we have investigated to date – the TM4, 5, 6, and 7 prolines in this study and the M1 transmembrane proline of two different ligand–gated ion channels – we find that lack of a backbone hydrogen bond donor is important for function.(25, 26) A general phenotype has emerged in which loss of function caused by mutations to conventional amino acids is mitigated by the corresponding α–hydroxy acids. In contrast to this transmembrane proline phenotype, α–hydroxy residues produced nonfunctional human 5–HT₃ receptors when replacing a proline that appears to undergo cis–trans isomerization.(27) This proline lies not in an α–helix, but in a loop/turn region, where cis–trans isomerization seems much more plausible.

We do observe variations on the general transmembrane proline phenotype. Unlike the D2 receptor's TM4, 6, and 7 prolines, the requirement for lacking a backbone hydrogen bond donor is absolute at the ion channel M1 proline sites; conventional amino acids, including glycine, produced completely nonfunctional channels.(25, 26) The TM5 proline of the D2 receptor additionally requires N–substitution for fully wild–type function. Taken together, these studies establish the power of unnatural amino acid methodology for dissecting out the various consequences of the proline residue's unique structural features.

METHODS

Molecular Biology

In these experiments, the cDNA for GIRK1 and GIRK4 was in pBSMXT plasmids and for the D2 receptor (human long form) was in the pGEMhe plasmid. Site–directed mutagenesis was performed using the Stratagene QuikChange protocol to generate the appropriate codon. For unnatural amino acid mutants and conventional mutants generated by nonsense suppression, the site of interest was mutated to the TAG stop codon. Plasmids were linearized with the appropriate restriction enzymes (the GIRK plasmids with SalI and the D2 receptor with NheI or SbfI). Receptor mRNA was prepared by in vitro runoff transcription using the Ambion T7 mMessage mMachine kit and GIRK1 and GIRK4 mRNA was prepared with the T3 kit.

Hydroxy or amino acids, all commercially available, were appended to the dinucleotide dCA and enzymatically ligated to truncated 74mer THG73 tRNA as previously described.(39, 40) The 74mer tRNA was prepared using the Ambion T7 MEGAshortscript^TM kit by transcription from a DNA oligonucleotide template with its 5' terminal two nucleotides bearing methoxy groups at the ribose C2' position, as described in the literature to enhance RNA transcript homogeneity.(41) Crude tRNA–amino acid or tRNA–hydroxy acid product was used without desalting, and the product was confirmed by matrix–assisted laser desorption ionization time–of–flight mass spectrometry on a 3–hydroxypicolinic acid matrix. Deprotection of the NVOC group on the tRNA–amino acids or NB group on tRNA–Aah was carried out by 5–min photolysis on a 1 kW xenon lamp with WG–335 and UG–11 filters immediately prior to injection. tRNA–Vah and –Lah were unprotected and were injected directly.

Oocyte Preparation and RNA Injection

Stage V–VI oocytes of Xenopus laevis were harvested and injected with RNAs as described previously.(39) For nonsense suppression experiments, each cell was injected with 15 ng each of GIRK1 and GIRK4 mRNA approximately 64 hours before recording, then 4 – 30 ng receptor mRNA and approximately 25 – 60 ng appropriate tRNA approximately 48 hours before recording. Mutants yielding small responses required a second injection of receptor mRNA and appropriate tRNA 24 hours before recording.

For wild type experiments, each cell received a single injection of 0.16 ng receptor mRNA and 10 ng each of GIRK1 and GIRK4 mRNA approximately 48 hrs before recording. Conventional amino acid mutants (except those generated by nonsense suppression) were prepared identically, except 1 ng receptor mRNA was required. Small responses were obtained for P423^7.50G, which required 10 ng receptor mRNA. Injection volumes for each injection session were 25 – 75 nL per cell.

As a negative control for suppression experiments at each site, unacylated full length tRNA was co–injected with mRNA in the same manner as charged tRNA. These experiments yielded negligible responses for all sites. Wild–type recovery conditions (injecting tRNA–Pro and appropriate mRNA) were injected alongside mutant nonsense suppression conditions to control for data variability.

Electrophysiology

Oocyte recordings were made in two–electrode voltage clamp mode using the OpusXpress™ 6000A (Axon Instruments). Recording buffers were ND96 (96 mM NaCl, 2 mM KCl, 1 mM MgCl₂, 5 mM HEPES, 1.8 mM CaCl₂, pH 7.5) and high K⁺ ringer (96 mM NaCl, 24 mM KCl, 1 mM MgCl₂, 5 mM HEPES, 1.8 mM CaCl₂, pH 7.5). Solution flow rates were 2 mL min⁻¹ and drug application flow rates were 2.5 mL min⁻¹. Initial holding potential was −60 mV. Data were sampled at 125 Hz and filtered at 50 Hz. Cells were subjected to a ND96 pre–wash for 10 s, a high K⁺ application for 50 s to establish basal currents, and dopamine application in high K⁺ ringer for 25 s, followed by high K⁺ and ND96 washings for 45 s and 90 s in duration, respectively. Dopamine–induced currents were measured over the basal K⁺ current as described previously.(31) Dopamine (Sigma–Aldrich) solutions in high K⁺ ringer were prepared immediately before recording by dilutions from a 1 M stock in ddi water. Dose–response data were obtained for a minimum of eight concentrations of dopamine, for a minimum of two cell batches, and for a minimum of 8 cells total. Dose–response relations were fitted to the Hill equation, I_Norm = 1/(1+(EC₅₀/A))^nH, where I_Norm is the normalized current peak at [dopamine] = A, EC₅₀ is the concentration of agonist that elicits a half–maximum response, and n_H is the Hill coefficient. EC₅₀ values were obtained by averaging the I_Norm values for each cell at a given dose and fitting those averaged I_Norm data to the Hill equation.