Histidine tagging both allows convenient single-step purification of bovine rhodopsin and exerts ionic strength-dependent effects on its photochemistry.

For rapid single-step purification of recombinant rhodopsin, a baculovirus expression vector was constructed containing the bovine opsin coding sequence extended at the 3'-end by a short sequence encoding six histidine residues. Recombinant baculovirus-infected Spodoptera frugiperda cells produce bovine opsin carrying a C-terminal histidine tag (v-opshis6x). The presence of this tag was confirmed by immunoblot analysis. Incubation with 11-cis-retinal produced a photosensitive pigment (v-Rhohis6x) at a level of 15-20 pmol/10(6) cells. The histidine tag was exploited to purify v-Rhohis6x via immobilized metal affinity chromatography. Optimized immobilized metal affinity chromatography yielded a binding capacity of > or = 35 nmol of v-Rhohis6x per ml of resin and purification factors up to 500. Best samples were at least 85% pure, with an average purity of 70% (A280 nm/A500 nm = 2.5 +/- 0.4, n = 7). Remaining contamination was largely removed upon reconstitution into lipids, yielding rhodopsin proteoliposomes with a purity over 95%. Spectral analysis of v-Rhohis6x showed a small but significant red shift (501 +/- 1 nm) compared to wild type rhodopsin (498 +/- 1 nm). The pK alpha of the Meta I<==>Meta II equilibrium in v-Rhohis6x is down-shifted from 7.3 to 6.4 resulting in a significant shift at pH 6.5 toward the Meta I photointermediate. Both effects are reversed upon increasing the ionic strength. FT-IR analysis of the Rho-->Meta II transition shows that the corresponding structural changes are identical in wild type and v-Rhohis6x.

For rapid sin g le-step p u rifica tio n o f recom b in an t rh o dopsin, a b a cu lo v iru s ex p ressio n v ecto r was con stru cted co n ta in in g th e b o v in e o p sin cod in g sequence ex ten d ed at th e S'-end by a sh o rt seq u en ce en cod in g six h istid in e resid u es. R ecom b in an t b acu lo viru s-in fected Spodoptera frugiperda cells p rod u ce b ovin e opsin car ryin g a C -term inal h istid in e ta g (v-opshis6x). The p res en ce o f th is ta g w a s con firm ed b y im m u n ob lot analysis.

In cu b ation w ith 11-cis-retin al p rod u ced a p h o to sen si tiv e p igm en t (v-R hohis6x) a t a le v e l o f 15-20 pmol/106 cells. T he h istid in e ta g w a s ex p lo ited to purify v-R hohis6x v ia im m ob ilized m e ta l a ffin ity chrom atogra phy. O ptim ized im m ob ilized m eta l a ffin ity chrom atog raphy y ield e d a b in d in g ca p a city o f 5:35 nmol o f v-R h ohis6x p er m l o f r e s in and p u rifica tio n factors up to
500. B est sam p les w ere a t le a st 85% pure, w ith an average p u rity o f 70% (A280 nm/A500 nm = 2.5 ± 0.4, n = 7). R em aining co n ta m in a tio n w a s la rg ely rem oved upon r e co n stitu tio n in to lip id s, y ie ld in g rh od op sin proteoliposom es w ith a p u rity o v er 95%.
Spectral a n a ly sis o f v-R h oh is6ac sh o w ed a sm all but sign ifican t red sh ift (501 ± 1 nm ) com pared to w ild type rhodopsin (498 ± 1 nm ). T he pK a o f th e M eta I «-» Meta II equilibrium in v-R h oh is6x is d o w n -sh ifted ixom 7.3 to 6.4 resu ltin g in a sig n ific a n t sh ift at pH 6.5 tow ard the Meta I p h otoin term ed iate. B oth e ffe c ts are rev ersed upon in creasin g th e io n ic stren gth . FT-IR a n a ly sis o f th e Rho M eta II tra n sitio n sh ow s th a t th e co rresp o n d in g stru c tural ch anges are id e n tic a l in w ild ty p e and v-R hohis6x.
Rhodopsin is the major component of the outer segments of the vertebrate rod photoreceptor cell. This visual pigment con sists of an integral membrane protein to which a chromophore, 11-cis-retinal, is covalently linked via a protonated Schiff base. Rhodopsin triggers the conversion of photon energy (light) into a graded membrane potential. The absorption of a photon leads to a number of discrete conformational changes in the protein moiety of the pigment (sequel of photointermediates -» photo cascade), finally resulting in the exposure of G-protein binding sites at the cytoplasmic surface of the protein. In the past decade, research has focused on analyzing the relationship between the structure of the receptor and its functional prop erties. Heterologous expression of the protein in combination with site-specific mutagenesis has become an attractive way to study this relationship. Several expression systems capable of in vitro biosynthesis of opsin have been described (1)(2)(3)(4)(5). Ex pression levels in these systems are usually quite low compared to total cell protein (<0.5%) and even to total membrane pro tein. For most analyses, recombinant rhodopsin therefore has to be extensively purified. Several methods have been de scribed for the purification of recombinant rhodopsin (1, 4, 6). These methods often have the disadvantage that the obtained samples are still contaminated to a various extent with pro teins derived from the cells, used for recombinant protein pro duction (4, 6). Quite pure preparations can be obtained using immunoaffinity chromatography (1). However, this approach is expensive (monoclonal antibody, peptides for elution), labori ous (antibody production, purification, and coupling), and fairly inefficient (low column capacity, recovery only in the order of 50%). Hence, this procedure is not very suitable for production of larger amounts (1-10 mg of purified protein) required for structural studies (crystallization, FT-IR spectroscopy,1 NMR spectroscopy).
The recombinant baculovirus-based expression system is ari excellent system for the production of larger amounts of recom binant bovine rhodopsin (6, 7). Thus far, we have used concanavalin A-Sepharose affinity chromatography to purify rho dopsin produced in this system (6), which, combined with reconstitution into proteoliposomes, yields reasonably pure preparations (60-80%). However, it is quite inefficient. Be cause of contaminating viral glycoproteins, the column capacity for recombinant rhodopsin is small, and a laborious elution profile has to be applied. Here we report on an alternative approach, using a his Li dine tag engineered onto the C terminus of bovine opsin. Histidine tagging for purification of recombi nant proteins by means of IMAC has been used for a number of proteins both in prokaryotic (8-10) and eukaryotic (11) expres sion systems, including recombinant baculovirus (12)(13)(14). None of these proteins, however, belonged to the superfamily of heptahelical G-protein-coupled membrane receptors of which rho dopsin is a member. Membrane proteins have to be solubilized with the help of specific detergents which may impair affinity techniques. Hence, we have evaluated IMAC for purification of recombinant rhodopsin. Here we will demonstrate that an adapted IMAC allows relatively simple, highly efficient single- 1 The abbreviations used are: FT-IR, Fourier transform infrared; IMAC, immobilized metal affinity chromatography; v-Rho, wild-type regenerated opsin produced in vitro by recombinant baculovirus; v-Rhohis0x, C-terminally histidine-tagged regenerated opsin produced in vitro by recombinant baculovirus; DoM, dodecyl-^-l-maltoside; PIPES, 1,4-piper azinediethanesulfonic acid; MES, 4-morpholineethanesulfonic acid; HEPPS, 4-(2-hydroxyethyl)-l-piperazinepropanesulfonic acid; ds, double-stranded; NTA, nitrilotriacetic acid; PAGE, polyacryl amide gel electrophoresis. step purification of histidine-tagged rhodopsin with an excel lent purification factor (>500). The procedure we developed is directly applicable to other membrane (receptor) proteins. In terestingly, the C-terminally located histidine tag slightly in fluences spectral properties (3 nm red-shift) and photocascade (downshift of the pKa of the Meta I ** Meta II equilibrium) of rhodopsin, however, without perturbing the structural changes accompanying the photocascade. The effects are fully reversible at higher ionic strength and probably represent unexpected electrostatic effects on selected rhodopsin properties.

EXPERIMENTAL PROCEDURES
Materials-Dodecyl"/3-l-maltoside (DoM) and nonyl-j3-l-glucoside were prepared as described previously (15). PIPES, HEPPS and MES buffers were purchased from Research Organics Inc,, Trolox from Al drich, and leupeptin from Sigma. AgCl windows were obtained from Fisher Scientific Co. Regen eration of v-opshisfix into v-Rhohis(}x was accomplished in total cellular membrane preparations, Briefly, cell pellets were resuspendcd in buffer A (6.5 mM PIPES, 10 mM EDTA, 5 mM dithioerythritol, and 100 ng/ml leupeptin, pH 6.5) at 10" cells/ml and lysed upon homogenination (P ot ter-Elvehjem-tube), Homogenized cells were centrifuged (30,000 x g 3 20 min, 4 °C), and the pellet was resuspended in buffer B (20 mM PIPES, 130 mM NaCl, 10 mM KC1, 3 mM MgCl^, 2 mM CaCl2, 0,1 mM EDTA, and 100 ng/ml leupeptin, pH 6.6). All subsequent manipulations were per formed in a nitrogen atmosphere and under dim red light (Schott-Jen a, RG 645 cut-off filter). Retina lipids were added in 100-fold molar excess over v-opshis(ix, followed by ll-c/s-retinal in a small volume of dimethylformamide (10-fold molar excess over v-opshisfix) and solid dodecyl-j8-1-maltoside (DoM) to 0.5 mM, Samples were rotated for 2 h at room temperature and centrifuged (30,000 x gt 20 min, 4 °C). v-Rhohis(Jx was solubilized from the pellet in the same buffer that was used to bind v-Rhohis(ix to the Ni2'1 ' nitrilotriacetic acid agarose resin (NP' -NTA; Qiagen). Several buffers were evaluated for the IMAC, Buffer C con sisted of 20 mM HEPPS, 0,5 M NaCl, 0,1 mM phenylmethylsulfonyl fluoride, 100 ng/ml leupeptin, 1 mM imidazole, 50 ju ,m Trolox, and 20 mM DoM (pH 8,0), Buffer D was identical with buffer C, except for 10 mM imidazole instead of 1 mM. Buffer E was identical with buffer C, oxcept for buffer and pH: PIPES (pH 6,5) was used instead of HEPPS (8,0). The regenerated pellets were incubated with these buffers (1 ml/10H cells) for 2 h at room temperature to solubilize v-Rhohiafix. Extracts were centrifuged (30,000 X g t 20 min, 4 °C) to remove insoluble material, and glycerol was added to a concentration of 15% (w/v). The solubilized v-Rhohis8x was then applied to the NiB >,'-NTA column at a flow rate of 1-2 ml/la for 16 h at 4 °C under continuous recycling. This flow rate was maintained throughout the entire procedure. Typically, we used a col umn volume of 1 ml for the v-RhohisG x extract from 1-2 x 10n infected insect cells. Unbound material was eluted from the column with the same buffer that was used for loading of the sample except that it was supplemented with 15% (w/v) glycerol and 20 mM nonylglycose instead of 20 mM DoM (10 column volumes). This was followed by a second rinse using the same buffer, but now containing 25 mM imidazole (4 volumes). The bound v-Rhohis(Jx was then eluted from the column using the same buffer (C, D, or E) supplemented with 100 mM imidazole and containing 20 mM nonylglucose instead of DoM and 130 mM instead of 500 mM NaCl (6 volumes).

Construction of the Histidine Tag Transfer Vector (pAcJAC2)-
Reconstitution of Purified v-RhohisC tX into Retina Lipids-Retina lip ids were extracted from illuminated bovine retinae using standard proce dures (19), The lipids were stored at -8 0 °C in the dichloromethane/ methanol extract. Their concentration was determined by a modified Fiske-Subbarow phosphate assay (20). The aliquot required for reconsti tution is dried with argon gas, dissolved in a small volume of methanol, and then diluted with buffer B containing 20 mM nonylglucose.
Purified v~Rhohisr>x was concentrated to 3-10 nmol/ml on an Omega 30K filter (Filtron, Northborough, MA) and mixed with 1 volume of retina lipid extract (100-fold molar excess). This mixture was layered on top of a sucrose step gradient (10%, 20%, and 45% (w/w) in 3 X diluted buffer B) and spun overnight (100,000 X g } >16 h, 4 °C). Reconstituted v-Rhohis(ix was collected from the 20% and/or 45% interface. Collected proteoliposomes were either diluted with the required buffer, and used for analysis, or with distilled water, pelleted (80,000 x^, 1 h, 4 °C), and stored at -8 0 °C until further use.
Immunoblot Analysis-Proteins were separated on a 12.5% SDSpolyaery 1 amide gel (21), transferred to nitrocellulose (Schleicher and Schull BA85, pore size 85), and probed with monoclonal or polyclonal antibodies (17). The anti-opsin antibodies Rho-lD4 (monoclonal) and CERNJS858 (polyclonal) have been described before (22, 23) and were used in 1:1000 dilution. The anti-histidine tag antiserum was obtained from Cappel, Organon Teknika N.V., Turnhout. Belgium and was also used 1:1000, UV/Vis Spectroscopy and Photolysis-UV/Vis spectroscopy to deter mine absorbance band shape and A,mix was routinely performed at pH 6.5 in buffer B supplemented with 20 mM DoM using a Perkin-Elmor A15 recording spectrophotometer. Detergent was added to prevent light-scattering artifacts. The spectrum of detergent-solubilized sam ples was recorded, after addition of 1 m hydroxylamine to 50 mM, before and after illumination (5 min, 300-watt light bulb, KGl. heat filter, Schott, Mainz, FRG). For accurate determination of the Anmx, difference spectra were used (illuminated spectrum subtracted from the "dark' one).
Fourier Transform Infrared (FT-IR) Difference. Spectroscopy-FT-IR spectra of the rhodopsin -* Meta II transition wore recorded in principle as described previously (25,26). Briefly, 1 nmol samples were deposited on AgCl windows by isopotential spin-drying (27). The resulting films were hydrated with 3 /xl of MES buffer (pH 5.5), covered with another AgCl window, sealed with Teflon tape, and mounted into a variable temperature cell (Specac P/N215G0, Kent, UK). FT-IR spectra were recorded on a Mattson Cygnus 100 spectrometer (Madison, WI) equipped with a liquid nitrogen-cooled, narrow band MCT detector. Sample hydration was monitored by the ratio between the peak absorb ance near 3400 cm" 1 (OH stretching mode of water) and near 2900 cm 1 (CH stretching mode of protein and lipids). Rhodopsin -> Meta II difference spectra were recorded at 10 °C. Spectra were taken at 8 cm 1 resolution in 5-min blocks before and after illumination at (1280 scans/ spectrum). Each sample was illuminated for 30 s in the spectrometer under computer control using a 12-V 20-watt halogen lamp equipped with Schott KGl and GG530 cut-off filters. Difference spectra were computed by subtracting spectra before illumination from spectra taken after illumination, using the EXPERT-IR software (Mattson),

Construction of the Histidine Tag Transfer Vector pAcJAC2
and Generation of Recombinant Virus-In order to introduce a tag consisting of six histidines at the C terminus of bovine opsin, the transfer vector pAcJAC2 was constructed (Fig. 1A). pAcJAC2 is derived from the baculovirus transfer vector pAcDZl (28), A short synthetic sequence encoding a histidine tag followed by a stop codon was introduced at the C terminus using two complementary oligonucleotides (Fig. IB). As a re sult, the two alanine residues at the C terminus of opsin (Ala-346 and Ala-348) were substituted by an aspartic acid and the first histidine residue, respectively (Fig. IB), These substitu tions concern residues, which are not highly conserved. Also, the aspartic acid residue introduced should partly counterbal ance the additional charge carried by the histidine tag. In pAcJAC2, opsin biosynthesis is controlled by the poly hedrin promoter while the small heat shock promoter of hsp70, drives the biosynthesis of /3-galactosidase, which functions as a reporter enzyme (17, 28).

Histidine Tagging of Recombinant Bovine Rhodopsin
Expression of v-opshis6x using Recombinant Virus AcNPV/ ops his-Protein samples derived from recombinant virus-infected Sf9 cells were analyzed by immunoblot, using a poly clonal antiserum elicited against bovine opsin ( Fig. 2A). Wildtype virus-infected Sf9 cells are used as a negative control, and native bovine opsin and wild-type v-ops as positive controls (Pig. 2A, lanes 1,2, and 4). The presence of the histidine tag in v-opshis6x was confirmed by the following observations: 1) only v-opshis6x is recognized by the histidine tag antibody (see be low), 2) the apparent molecular mass of v-opshis6x is larger (by about 2 kDa) than that of native opsin and v-ops (Fig. 2A, lane   3 versus lanes 2 and 4), 3) v-opshis6x does not react with the monoclonal antibody 1D4 (Fig. 2B, lane 1 versus 2), since the C-terminal histidine tag extension eliminates the epitope for this antibody (29).
Hence, recombinant virus AcNPV/opshis directs expression of a fully intact histidine-tagged v-ops. The expression level varies between 20 and 30 pmol/106 cells, which is comparable to wild-type v-ops (6).  Table I. Lane 1 in Fig. 3, A, 23, and C, presents the total extract of Sf9 membranes after regeneration. Only by immu noblot analysis can glycosylated v-opshis(JX be identified (ar row) together with a minor amount of unglycosylated species {arrowhead). The nonbound, flow-through fraction shows a very similar protein pattern {lane 2), except that most v-Rhohisfix has bound, since only minimal amounts of the gly cosylated form are detected in this fraction. Additional wash ings with extraction buffer (buffer C) elute a complex protein population (Fig. 3, lanes 3 and 4), probably representing aspecifically or very weakly bound proteins. The nonglycosylated v-opshis6x is already strongly present in these fractions (Fig, 3,  B and C, lanes 2 -4 , arrowhead) and apparently is not very well retained by the column. Protein contamination, weakly inter acting with the column, could be eluted by raising the imidaz ole concentration to 26 mM (Fig, 3, lane 5). Under these condi tions, a minor amount of v-Rhohisflx was eluted. Most of the specifically bound v-Rhohis0x, however, only was eluted upon raising the imidazole concentration to 100 mM (Fig. 3, lane 6). This fraction also contains a minor amount of nonglycosylated species and is still contaminated by several other minor bands (Fig. 3A, lane 6, open arrowheads).

Regeneration and Purification ofv-Rhohis6x-To
The amount of nonglycosylated v-opshis(> x in the DoM extract and column fractions varied between different experiments, but it was always present at immunodetectable levels (data not shown). In addition, immunoblots showed the presence of a third opsin species (Fig. 3£, lanes 1-4) which on SDS-PAGE gels migrated between the glycosylated and nonglycosylated v-opshis6x. This species did not react with the anti-histidine tag antibody (Fig. 3C, lanes 1-4), We were unable to detect this band in the final purified v-Rhohis^ fraction (Fig. SB, lane 6').
An overview of IMAC results obtained under two conditions, as determined by spectroscopic analysis, is given in Table I. Nearly complete binding of v-Rhohis6x to the Ni2'* -NTA agar ose was attained. At pH 8.0 (buffer C), a pH at which IMAC is

Fiii, 3. T ypical p u r ifica tio n p ro file o f v-R h oh is0x u sin g IMAC, Fractions were analyzed by SDS-PAGE (A, Cooniassio blue staining) and immunoblot (B and C). Immunoblots were screened with the poly clonal anti-rhodopsin antibody OERNJS858 (B) and the anti-histidine tag antibody (C). Lane 1 represents the Sf9-ccll membrane extract loaded onto the Ni^'-NTA-agarosc column. The nonbound fraction is shown in lane 2. Lanes 3 and 4 represent early fractions eluted upon washing with buffer C, whereas lane 5 contains the proteins that eluted from the column after supplementing butter C with 25 mM imidazole. The strongly bound fraction, eluted from the column using buffer C supplemented with 100 mM imidazole, is shown in lam 6. Lanes 7 and 8 in A represent purified, reconstituted native bovine rhodopsin (lane 7) and v-Rhohi9(Jx (lane 8). Position of glycosylated (large arrow) and nonglycosylated (small, filled arrowhead) v-Rhohisflxl as well as the intermediate do-liistidine-tagged form (large arrowhead) and some mi nor contaminants in the purified fraction (small open arrowheads) are indicated. The immunopositive bands (B and C) and the Goomasaiepositivo band (A, lane S) around 67 kDa represent a dimer of v-Rhohistt)e.
usually performed, an average recovery of 84% was achieved with a purification factor of at least 450. An attempt to reduce contamination by weakly binding proteins, by applying the membrane extract in the presence of a higher imidazole con centration (10 mM; buffer D), resulted in a much higher loss of v-Rhohis0x and in fact lowered the purification factor (not shown). Finally, the performance of the procedure we devel oped at pH 8.0, was evaluated at pH 6.5, since rhodopsin and most of its mutants are thermally much more stable at the T able I General characteristics of v-Rhohisffx purification by IMAC as determined by UVjVis spectroscopy Values given are mean ± S.D. of three experiments. The percentage solubilization is related to the amount solubilized under standard con ditions (6), which was taken as the 100% value. latter pH. This actually improved the purification factor to at least 500, without significant reduction in recovery (Table I, buffer E). At pH 6.5, best v-Rhohis6x samples had a purity of 80-85% (A280/A500 ratio of 2.0-2.1), and the average purity of combined fractions was 70-75%. The maximal capacity of the Ni2_h-NTA agarose column for v-Rhohis6x has not been deter mined exactly but is at least 35 nmol/ml bed volume. The efficiency of this IMAC procedure is spectrally illustrated in Fig. 4A. Curve 1 represents the total membrane extract applied to the column, while curve 2 presents the spectrum of the combined purified v-RhohisG x fractions. The contamination in the v-Rhohis6x fraction obtained after IMAC resides in several minor bands. These are largely re moved upon subsequent reconstitution of v-Rhohis6x into ret ina lipid proteoliposomes, which represent a more native-like environment, we routinely use to analyze functional properties of recombinant rhodopsin (6). To simplify reconstitution, the detergent used during regeneration (DoM) is exchanged during IMAC for nonylglucose, which is more easily exchanged for phospholipids (30). The resulting v-Rhohis6x proteoliposomes contain at least 95% rhodopsin on a protein base and are suitable for all functional analyses (Fig. 3A, lane 8) Spectral Pi'opei'ties and Photocascade of v-Rhohis6x-Curve 2 in Fig, 4A is a typical absorbance spectrum of the combined v-RhohisSx fractions obtained after IMAC purification. Illumi nation of v-Rhohis6x "bleaches" the main absorbance band at 500 nm, and we used difference spectra to calculate the Amax more accurately. Unexpectedly, the absorbance band of v-Rhohis6x turns out to be slightly red-shifted (Amax = 501 ± 1 nm, n -7) relative to wild-type (Amax = 498 ± 1 nm) (Fig. 4B). This slight but significant red shift has been observed in all samples produced so far, both before and after IMAC purifica tion. The shift is independent of the presence of 10 mM Ni2* ions (complexed histidine tag) or 10 mM EDTA (free histidine tag).
Analysis of the later part of the photocascade of reconstituted v-Rhohis6x presents another subtle effect of histidine tagging: at pH 6.5, the Meta I «■> Meta II equilibrium is clearly shifted toward Meta I (Fig. 5JB) in comparison to wild-type rhodopsin (Fig. 5A). At pH 5,5, v-Rhohis6x again behaves similar to wildtype (Fig. 5C:

little Meta I formed, normal formation of Meta III). Again, this effect is not influenced by excess Ni2+ ions or complexing agent (EDTA). Analysis over a larger pH range, in fact, demonstrates that the pKa of the Meta
Meta II equi librium is down-shifted about 1 pH unit from 7.3 in wild-type to 6.4 (Fig. 6).
Hence, we reasoned that these subtle changes in v-Rhohis properties might be due to the additional surface charge intro duced by the histidine tag. Increasing the ionic strength by addition of KC1 up to 1 m concentration indeed reverted Amax and photocascade of v-Rhohis6x to wild-type behavior (Table II).
Structural Analysis of the Rho -> Meta II Transition by FT-IR Spectroscopy-FT-IR difference spectroscopy provides detailed structural information on the transitions in the rho dopsin photocascade (e.g. Refs. 25-27, 31, and 32). We have used this approach to investigate whether histidine tagging would also exert influence on the conformational changes ac companying Meta II formation. Comparison of the FT-IR dif ference spectra for the rhodopsin to Meta II transition in native rhodopsin and v-Rhohis6x shows that these spectra are highly similar (Fig. 7). All major bands characteristic of the formation of Meta II (numbers in cm"1) are present in v-Rhohis6x as well and within experimental error (±2 cm"1) at the same fre quency as wild-type. Hence, the histidine tag does not detectably influence the conformational changes accompanying pho to excitation of rhodopsin.

Purification of Recombinant Membrane
Receptors-G-protein-coupled membrane receptor proteins have been expressed in all common in vitro expression systems (mammalian cell lines, insect cell lines, yeast, Escherichia coli)} but the levels of expression are always low (<0.5% of cellular protein). Hence, for more elaborate functional studies as well as for structural studies, a very efficient purification procedure is needed. The most selective and frequently used approach is to exploit re ceptor affinity for an (ant)agonist or a monoclonal antibody. For instance, the most popular way to purify recombinant rhodopsins is by immunoaffinity chromatography over immobi lized 1D4 (1, 33).
However, this approach has several disadvantages: it suffers from rather low recoveries (^50% in our hands), low column capacity (1-2 nmol/ml bed volume), and expensive exploitation (monoclonal antibodies, peptide for elution). It is therefore not very suitable to purify larger amounts of recombinant protein.
As an alternative, we adapted lectin affinity chromatography over concanavalin A-Sepharose to the recombinant baculovirus system (6, 17). This technique gives good recoveries (3:80%) and is fairly inexpensive, but due to the large contamination with viral glycoproteins, a complex elution profile has to be applied and the column capacity for rhodopsin binding is low (approximately 1 nmol/ml bed volume). As a consequence, it only affords rather low purification factors (50-100) and is quite laborious in the case of larger batches. We therefore searched for other alternatives and IMAC in combination with histidine tagging looked like a good candidate (8-13).
IMAC Purification of Histidine-tagged Rhodopsin-We se lected the C-terminal of rhodopsin as a suitable site for append ing a 6x histidine tag since 1) this site allows immuno affinity purification and therefore should be well accessible and 2) C-terminal extension does not significantly influence func tional properties of the protein (33, 34).2 Recombinant histidine-tagged bovine opsin could be success fully, functionally expressed in the recombinant baculovirus/ insect cell expression system. Regeneration with 11-cis-retinal resulted in a pigment (v-Rhohis6x) that could be purified almost to homogeneity in a single step using IMAC under optimized conditions. This purification procedure is rapid, fairly inexpen sive, yields good recoveries (£=80%) with an excellent purifica tion factor (^500), and can handle, thanks to the high column capacity (2:35 nmol/ml bed volume), relatively large batches of rhodopsin. Interestingly, the histidine tag at the C terminus was found to have a subtle influence on the spectral and photolytic properties of this visual pigment, but these effects could be reversed by increasing the ionic strength.
The presence of the histidine tag at the C terminus of bovine opsin was first confirmed by immunoblot analysis. The modi fied C terminus was no longer recognized by the monoclonal antibody Rho-lD4 (29), and v-opshis6x was clearly identified by a polyclonal antibody elicited against a hexahistidine peptide. In addition, the apparent molecular mass of v-opshis6]C on the immunoblots has increased relative to v-ops. Our histidine tag should increase the molecular mass by approximately 0.7 kDa. The observed shift of the apparent molecular mass, seen on the immunoblots, is at least 2 kDa, however. This might reflect a direct or indirect (SDS binding) effect due to the additional charge introduced by the histidine residues, which would change the migration behavior on the gel. Such a large effect has not been reported before but it might be more pronounced in the case of membrane proteins, which by themselves usually show aberrant behavior in SDS-PAGE.
Ultimate proof for the presence of the histidine tag in v-opshis6x of course lies in the successful application of IMAC using Ni2'1 " chelation. With this technique we could combine a high column capacity with high purification factors and good recoveries. The A280/A5{)() ratio indicates that, under optimized conditions, the combined purified fractions on average contain at least 70% v-Rhohis6x. However, peak column fractions have been obtained in which this was as high as 80-85%. Immuno blot analysis shows that the purified v-Rhohis6x samples con tain some nonglycosylated v-opshis6x. Lack of glycosylation was reported to reflect impaired protein folding and loss of regeneration capacity (17,35). Therefore, one of the contami nants present in the v-Rhohis6x samples is v-opshisflx. Protein staining of PAGE gels reveals in addition to non glycosylated v-opshis6x two minor bands at ca 25-30 kDa, Hence, these three proteins are the major contaminants after IMAC. Upon subsequent reconstitution into proteoliposomes, this remaining contamination is also largely removed. In the case of opsin this is due to its relative low stability in detergents and tendency to aggregate, which impairs correct reconstitution into a lipid matrix. Indeed, absolute FT-IR absorption spectra, where the amide I (1620-1690 cm*"1) and amide II (1530-1560 cm"1) bands are extremely sensitive to protein secondary structure (36), do not suggest the presence of any (partially) misfolded opsin,3 while this was clearly observed for the mutants E134D and E134R (31).
In addition to glycosylated and nonglycosylated (rhod)opsin, a third anti-opsin immunoreactive band was detected, primar ily in the IMAC wash fractions. The latter band migrates between the first two bands and is not recognized by the anti histidine tag antibodies. This opsin species is only formed in minor amounts (<5% of total), and, presently, it is not clear where it derives from. It could be due to chemical cleavage of the histidine tag on the column, e.g. in the Ni2 '-complexed from, or derive from endogenous carboxy peptidase activity pres ent in the Sf9 cells. Both reactions might remove up to the entire tag, which would explain why this species has no or only low affinity for the Ni2' * -NTA agarose and is not recognized by the anti-histidine tag antibodies. Thus far, no one has de scribed a similar observation or any carboxypeptidase activity in Sf9 cells although a C-terminal histidine-tagged protein has been purified from this expression system before (13). However, we only did observe this (partial) removal of the histidine tag from v-Rhohisflx because of the relative large difference in apparent molecular mass between v-Rho and v-Rhohis6x and the availability of the anti-histidine tag antiserum. In the earlier reports, no large difference in molecular mass between the native and histidine-tagged proteins was observed, and no antiserum against the histidine tag was available. On the other hand, carboxypeptidase activity has been previously put to good use to remove a polyhistidine peptide fused to dihydrofo late reductase after purification from expression in E . coli (9). Hence, if cleavage of the histidine tag does become a serious problem for certain proteins, it is recommendable either to in clude carboxypeptidase inhibitors during extraction and purifi cation or to protect the histidine tag with a C-terminal proline residue or to insert/append it elsewhere into the protein.
Functional Properties of v-RhohisGx-The expressed v-opshisBx smoothly recombines with the chromophore, 1 1 -ctsretinal, into a functional photopigment. However, unexpect edly, the absorbance spectrum of v-Rhohis(ix shows a small but significant red shift (Amnx -501 nm)  II equilibrium. In addition, electrostatic effects (either local or more general like an increase in the protein dipole) could be involved. Both effects would indeed be suppressed by an in crease in ionic strength. It would be quite exciting if electro static effects could at least partially be responsible for the observed phenomena, since to our knowledge such "long range influence" has not been documented yet. We are presently investigating this in more detail by mapping bulk-pH depend ence, measuring surface pH, establishing reversal by removal of the histidine tag, and insertion of the histidine tag in other

T able II Reversal of histidine tag effects by an increase in ionic strength
The Amnx was determined from difference spectra (cf Fig. 4B), The percentage of Meta I remaining in the Meta I «-> Meta II photoequilib rium at pH 6.5 is calculated from the absorbance difference at 480 nm (spectrum 2 minus 4 in Fig. 5). Mean ± S.D. of n determinations is given. This increase in ionic strength does not significantly affect this proportion in bovine rhodopsin.

KC1
Meta I remaining v-Rhohisfìx 0.1 501 ± 1 (n 44 ± 5 (n = 4) 1.1 499 ± 1 (n = 2) 15 ± 3 (n « 2) v-Rho 0.1 498 ± 1 (n ss 15) 19 ± 4 (n - 5) positions. In order to establish whether the effect of the histidine tag would also penetrate at the structural level, FT-IR analysis was performed. The band shape of the amide I band indicates a very similar secondary structure composition for rhodopsin and v-Rhohis6xJ and the FT-IR difference spectra correspond ing to the rhodopsin to Meta II transition are also highly similar. All characteristic bands of this transition (at 970,1562,  1643, 1686, 1712, 1728, 1750, and 1768 cm"1; the latter two of which have recently been assigned to the C = 0 stretching mode of aspartic acid 83 (31, 32)) are also present in the difference spectrum obtained with v-Rhohis6x, Hence, according to anal ysis by FT-IR, no significant differences exist in the structural alterations accompanying receptor activation in rhodopsin or v-Rhohis6x.
Conclusion-The use of IMAC in combination with histidine tagging has considerably improved and simplified the purifica tion of recombinant rhodopsin. This will now permit us to further scale up the cell culture volume and recombinant pro tein production, which thereby will yield sufficient protein for more detailed structural studies, requiring modified protein (FT-IR, NMR, crystallization).
The adaptations we have introduced in the IMAC (use and exchange of detergents, lower pH, stepwise imidazole elution) should make it applicable for membrane proteins in general. It cannot be foreseen whether the subtle influence of the histidine