Topological Model of the Rhodobacter capsulatus Light-Harvesting Complex I Assembly Protein LhaA (Previously Known as ORF1696)

Abstract

A theoretical topology of the Rhodobacter capsulatus membrane protein LhaA was formulated and evaluated by gene fusion experiments. The apparent topological locations of fusion enzymes were compared with the theoretically derived structure, and a model of LhaA is suggested that consists of 12 transmembrane segments, with the N and C termini residing in the cytoplasm.

Although purple photosynthetic bacterial integral membrane proteins have been crystallized as components of reaction center and light-harvesting (LH) complexes and used for X-ray diffraction and other structural analyses (14, 19, 25), other integral membrane proteins that are present in cells at lower concentrations have not been purified. Nevertheless, methods exist for evaluation of the membrane topology of proteins to produce two-dimensional structural models by use of cloned genes (16). For example, the PucC protein of Rhodobacter capsulatus, which is required to obtain LH complex II, was predicted to be an integral membrane protein on the basis of sequence analysis of the pucC gene, and gene fusion experiments supported a proposed membrane topology of 12 transmembrane segments (15).

The LhaA (formerly ORF1696) protein is a major factor in LH complex I assembly and shares 47% amino acid sequence identity in an alignment of translated genes with PucC (2, 28). Thus, LhaA and PucC are homologous and may be functionally related in the assembly of LH complexes, perhaps by facilitation of LH polypeptide membrane insertion, delivery of bacteriochlorophyll to LH apoproteins, or subunit oligomerization (28).

Protein structural information often provides insights into function, and so, as the first step toward elucidation of the LhaA higher-order structure, we present a two-dimensional LhaA membrane topology model. A detailed description of the experimental methods used and the data presented in this note is available elsewhere (27).

The LhaA amino acid sequence was evaluated by use of the GES (8), GvH1 (23), and KD (13) hydropathy scales of the TopPred II 1.1 software package (5) to assess the possible number and locations of transmembrane segments in the protein. The positive-inside rule (24) was considered along with the locations of predicted transmembrane domains to arrive at a representation of the LhaA protein, which consists of 12 transmembrane segments, as illustrated in Fig. 1. This theoretical model was used to identify regions of the LhaA protein as targets for the construction of lhaA′::pho′A translationally in-frame gene fusions.

FIG. 1.

LhaA membrane topology model. The primary amino acid sequence is given in single-letter code, transmembrane segments are encircled by ovoid shapes, and the phospholipid bilayer is indicated by horizontal lines. The LhaA amino acid numbers preceding protein fusion junctions are indicated, and extramembranous loops are labeled according to their predicted periplasmic (P) or cytoplasmic (C) location. Positively charged residues are designated with plus signs.

Activities of PhoA fusions.

The logic behind using the phoA gene for studying membrane protein topology is that PhoA (alkaline phosphatase; AP) fusions to regions of a membrane protein that result in a periplasmic location of the AP moiety typically have relatively high activities and stabilities, whereas those in which the AP moiety is located in the cytoplasm should have low activities and stabilities. However, sometimes aberrant AP activities are obtained, and so, topological models based on the results of gene fusion experiments should be viewed critically (1, 6, 7).

Construction of lhaA′::pho′A fusions was done by digestion of the cloned lhaA gene in plasmid pCY1800 (27) with restriction enzymes having unique cleavage sites (NaeI, BstXI, and BsaAI), followed by treatment with Bal 31 nuclease and ligation (20) of the resulting truncated segments of lhaA into plasmid pUC19::phoA (4). Three of the fusions (Gln-30, Ser-244, and Glu-464) were made by subcloning restriction fragments into plasmid pUI310 or pUI320 (22), and one (Met-56) was made by PCR (20, 27). The ligated products were transformed into Escherichia coli phoA mutant CC118 (17), and plasmids of selected transformants were DNA sequenced across the fusion joint and assayed for AP activity as described previously (15). Table 1 summarizes the locations of the fusion joints and AP activities of lhaA′::pho′A translationally in-frame fusions. We designated AP activities of greater than 30 U high, activities between 9 and 30 U intermediate, and activities below 9 U low. The AP activities of most of the theoretically predicted periplasmically located fusions were between approximately 130-fold and 3-fold higher than AP fusions to predicted cytoplasmic domains of LhaA. Notable exceptions were the high AP activity of the most N-terminal fusion (Gln-30), the relatively low activities of the predicted periplasmic fusions between transmembrane segments 11 and 12 (Ala-413 and Leu-442), and the intermediate activities of the predicted cytoplasmic C-terminal segment fusions (Leu-452 and Glu-464).

TABLE 1.

AP and β-gal activities of LhaA′::Pho′A and LhaA′::Lac′Z fusions

Position of fusion jointa	AP activity (U)b	AP relative activity (%)	Normalized AP relative activity (%)c	β-gal activity (U)b	Predicted locationd
Gln-30	56.5	60.0	77.8	ND	C
Met-56	35.3	37.5	NDe	1.4	P
Leu-108	2.4	2.5	1.0	12.4	C
Gly-136	43.6	46.3	17.6	0.7	P
Ala-163	3.5	3.7	2.6	17.0	C
Gln-202	37.1	39.4	21.0	0.3	P
Ser-244	0.7	0.7	0.6	ND	C
Gly-288	52.2	55.4	54.2	0.7	P
Leu-351	94.2	100.0	100.0	1.9	P
Ala-413	11.3	12.0	9.0	15.8	P
Leu-442	9.1	9.7	22.0	9.4	P
Leu-452	18.9	20.1	ND	34.2	C
Glu-464	20.0	21.2	21.4	ND	C

^aThe last LhaA amino acid prior to the start of fusion sequences is given. 

^bEnzyme activity is reported as units of A₄₂₀ per minute per 10⁶ cells. 

^cActivities are normalized to the rate of Leu-351 hybrid protein synthesis in pulse-labeling experiments. 

^dC, cytoplasmic; P, periplasmic. Locations are based on theoretical analyses. 

^eND, not determined. 

Western blot analysis of AP fusion proteins.

The steady-state levels of fusion proteins may be visualized as anti-AP immunoreactive segments of AP fusion proteins. Cytoplasmically located fusions frequently are unstable, and the AP segments are degraded rapidly, and so, hybrid proteins that are transported into the periplasm often yield relatively strong AP signals in immunoblots because they are more stable due to proper folding (15, 21). Results of Western blot analysis of a subset of the LhaA′::Pho′A fusion proteins of intact cells are shown in Fig. 2. Several faint bands were correlated with the theoretical molecular masses of hybrid proteins and are likely to represent the full-length proteins of the Glu-464, Leu-442, Leu-351, Gly-288, and Gln-30 fusions (Fig. 2, lanes 1 to 4 and 10). It is possible that fusion protein bands are present in other lanes but are obscured by aggregation of degradation products. Fusion proteins predicted to have AP located in the periplasm showed larger amounts of a 47-kDa immunoreactive segment, presumably released by cleavage of the fusion protein in the vicinity of the LhaA′::Pho′A junction (the arrow in Fig. 2 indicates the full-length AP protein of E. coli phoR mutant CC149 [18], the positive control), than most of the cytoplasmic fusions (compare the P and C lanes in Fig. 2). These results are consistent with the idea that the AP moiety folds properly in the periplasm, making it more resistant to proteolytic degradation than AP segments in the cytoplasm, which appear to be in proteolytically sensitive conformations. Based on the intensity of the AP band (at 47 kDa) in each lane of Fig. 2, the relative amounts of the fusion proteins detected are consistent with their predicted periplasmic or cytoplasmic locations, and thus, these results support the theoretical topology model (Fig. 1).

FIG. 2.

Western blot of LhaA′::Pho′A fusion proteins. The letter C or P above a lane indicates the predicted cytoplasmic or periplasmic localization, respectively, of the AP moiety. Molecular masses (M) are in kilodaltons. The arrow marks the position of the 47-kDa AP protein, and asterisks indicate bands having molecular masses similar to those predicted for the full-length fusion proteins. Lanes: −, strain CC118; +, strain CC149; 1, Glu-464; 2, Leu-442; 3, Leu-351; 4, Gly-288; 5, Ser-244; 6, Gln-202; 7, Ala-163; 8, Gly-136; 9, Leu-108; 10, Gln-30 (see Table 1 and Fig. 1 for locations).

Pulse-labeling experiments.

It was proposed that AP fusions to membrane proteins may yield AP activities that differ from each other because of differential synthesis of AP, resulting from different translation rates of fusion mRNAs, as opposed to the AP protein segment being located periplasmically or cytoplasmically (21). To evaluate the rates of synthesis of the AP moieties of the LhaA′::Pho′A fusions, we used [³⁵S]methionine pulse-labeling, immunoprecipitation, sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and quantification of radioactivity to normalize the AP activities of a subset of the LhaA′::Pho′A fusions (16, 21, 27). No full-length fusion proteins (based on predicted molecular masses) were revealed by this technique, although 47- and 40-kDa protein bands were observed and assumed to be remnants of proteolytic degradation of the full-length AP fusion proteins. The intensities of these bands were used to obtain normalized AP activities (relative to the activity of fusion Leu-351) as summarized in Table 1 (27). Fusions Leu-108, Gly-136, and Gln-202 yielded lower normalized than relative AP activities, suggesting that they were synthesized relatively rapidly, whereas fusion Leu-442 exhibited higher normalized than relative AP activity, indicating that this fusion protein was synthesized at a low rate.

Activities of β-gal fusions.

Although the evaluations of the predicted LhaA topology in the AP fusion analyses described above generally supported the theoretical model, the exceptions prompted us to investigate LhaA topology by an alternative method. Construction of lhaA′::lac′Z fusions was done by replacing the pho′A allele of some of the lhaA′::pho′A fusions with the lacZ allele from plasmid pSP72::lacZ (15) by using standard subcloning procedures (20, 27). Since β-galactosidase (β-gal) fusions usually are most active when joined to a cytoplasmic domain, fusions located at cytoplasmic regions should have relatively high activities compared to fusions located at periplasmic segments (16). As shown in Table 1, most of the β-gal activities were relatively low for predicted periplasmically located fusions and relatively high for predicted cytoplasmic fusions. This pattern is reciprocal to the AP activity data and therefore generally supportive of the LhaA theoretical structural predictions. Two exceptions are β-gal fusions Leu-442 and Ala-413, which were predicted to be located in the periplasm but exhibited β-gal activities that were intermediate between the extremes of the data. However, the LhaA′::Lac′Z fusion at the Leu-452 site had the highest β-gal activity of all of the β-gal fusions, indicating a cytoplasmic location for the C terminus of LhaA.

Summary and discussion.

The topology model of the LhaA protein consists of 12 transmembrane segments, seven cytoplasmic domains enriched in the positively charged amino acids Arg and Lys, and six periplasmic loops (Fig. 1). Extramembranous domains range in length from 10 (loop C5) to 35 (loop C3) amino acid residues, and the N and C termini are located in the cytoplasm.

We were unable to obtain fusions to the C4 and C5 domains, but we suggest that these regions are located in the cytoplasm for the following reasons. The C4 domain is likely to reside in the cytoplasm in accordance with the positive-inside rule (24), since it contains four Arg residues as opposed to one positive residue in each of the predicted flanking periplasmic loops, P4 and P5 (Fig. 1), and the predicted membrane-spanning segments, 8 and 9, flanking C4 were strongly indicated by the hydropathy algorithms. The AP fusion activities for loops P4 and P5 were high, whereas β-gal fusion activities were low, confirming their periplasmic location, and so this topological arrangement indicates a cytoplasmic location for the C4 domain. Similarly, loop C5 is preceded N terminally by strongly predicted membrane-spanning segment 10 and experimentally verified periplasmic loop P5, and so it seems that C5 is located in the cytoplasm.

A “putative” transmembrane domain predicted by the TopPred program to occur within the first 30 amino acids of LhaA, which are present in the Gln-30 fusion, may act as an ersatz signal sequence in isolation from the rest of the protein and result in transport of the AP segment of the Gln-30 fusion protein into the periplasm and the observed high AP activity. However, the absence of a Gln-30 47-kDa band in Fig. 2 indicates a cytoplasmic location, and aberrant PhoA activities were reported for analogous N-terminal fusions (7). The strongly predicted location of transmembrane segment 1, the greater numbers of positively charged amino acids in the predicted N-terminal domain (compared to P1), and the Western blot results argue for a cytoplasmic location for the LhaA N terminus.

The four positively charged residues located near the C terminus of the LhaA protein (Arg-454, Lys-461, Lys-462, and Arg-469; Fig. 1) could help to anchor the C terminus in the cytoplasm and, if they did, would be important topological determinants. Since these residues are replaced with AP or β-gal moieties in fusions Ala-413, Leu-442, and Leu-452 and partially in fusion Glu-464, variable topologies for these fusions could arise and lead to fusion enzyme activities that do not reflect the topology of the native LhaA protein. We assume that the P6 segment is a periplasmic loop and that residues near the C terminus help to establish the LhaA topology suggested in Fig. 1.

The LhaA protein was recently reported to enhance LH complex I assembly in R. capsulatus, and it was proposed that one function of LhaA might be in the delivery of bacteriochlorophyll molecules to the LH complex I α and β proteins (28). Homologues of LhaA have been discovered in other purple bacteria, Synechocystis strain PCC6803, and Prochlorococcus marinus (3, 9–12, 26). We suggest that the members of this family of proteins have a function in common, namely, tetrapyrrole delivery for LH complex assembly in all of these species. It will be interesting to test the LhaA and PucC structural models presented here and elsewhere (15) to see if aspects of these proposed structures relate to a function in tetrapyrrole transfer into or across the cytoplasmic membrane for delivery to LH complexes.