Crystal Structure of Allophycocyanin from Red AlgaePorphyra yezoensis at 2.2-Å Resolution*

The crystal structure of allophycocyanin from red algae Porphyra yezoensis (APC-PY) at 2.2-Å resolution has been determined by the molecular replacement method. The crystal belongs to space group R32 with cell parameters a = b = 105.3 Å, c = 189.4 Å, α = β = 90°, γ = 120°. After several cycles of refinement using program X-PLOR and model building based on the electron density map, the crystallographic R-factor converged to 19.3% (R-free factor is 26.9%) in the range of 10.0 to 2.2 Å. The r.m.s. deviations of bond length and angles are 0.015 Å and 2.9°, respectively. In the crystal, two APC-PY trimers associate face to face into a hexamer. The assembly of two trimers within the hexamer is similar to that of C-phycocyanin (C-PC) and R-phycoerythrin (R-PE) hexamers, but the assembly tightness of the two trimers to the hexamer is not so high as that in C-PC and R-PE hexamers. The chromophore-protein interactions and possible pathway of energy transfer were discussed. Phycocyanobilin 1α84 of APC-PY forms 5 hydrogen bonds with 3 residues in subunit 2β of another monomer. In R-PE and C-PC, chromophore 1α84 only forms 1 hydrogen bond with 2β77 residue in subunit 2β. This result may support and explain great spectrum difference exists between APC trimer and monomer.

three or more core cylinders associated by APC discs, is in proximity of the reaction centers, whereas the rods are attached on the core and are composed of PC discs in the middle and PE or PEC discs on the tip. Light energy is transferred from PE or PEC via PC to APC and finally to the reaction centers (1).
The crystal structure of APC is very special compared with other phycobiliproteins. First, the spectrum difference between APC trimer and its monomer is very large. When APC monomers aggregate to trimer, the absorption spectrum has a 40-nm red shift; the CD spectrum also changes a great deal, and exiton interaction in the trimer of APC was suggested (10), whereas the spectrum difference between C-PC monomer and its trimer is not so large as in APC, although phycocyanin has the same ␣84PCB and ␤84PCB as APC.
Second, the functional unit of APC was thought to be a trimer, whereas the function unit of other phycobiliproteins were hexamer (␣␤) 6 or (␣␤) 6 ␥. Brejc and co-workers solved the structure of APC-SP from blue alga S. platensis (9) in the unit cell of APC-SP crystal; two trimers are associated in a "back to back" manner that might represent the assembly state of APC in nature. Red alga is higher than blue alga in evolution, so it would be interesting to know the packing of APC from red alga in the unit cell and in nature.
Third, in PE and PC, the two trimeric discs are superimposed along a 3-fold axis, but in PC and APC the two discs are connected perpendicularly. The pathway of energy transfer between PC and APC is still unknown.
The red algae Porphyra yezoensis is an algae that exists widely in nature. Its phycobilisomes contain R-PE, C-PC, and APC. In this paper we report the crystal structure of APC from P. yezoensis (APC-PY) at 2.2-Å resolution. The organization of APC trimers in the core cylinders of phycobiliproteins and the pathway of energy transfer were discussed.
Molecular replacement using program AMoRe (12) was carried out using the 2.3-Å structure of APC-SP as a model. Model cell parameters were a ϭ b ϭ c ϭ 150.0 Å, ␣ ϭ ␤ ϭ ␥ ϭ 90°, integrate radius was 30 Å, and rotation function calculation gave a rather high coefficient solution, * This work was supported by Chinese Academy of Sciences (KJ85-04-40) and the National Natural Science Foundation of China (39630090). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The structure was refined using X-PLOR (13). The consensus sequence was used for the initial model building. Fourier transform and electron density were first calculated in the resolution range of 10 Å to 3.5 Å. Residues that could not be fitted into the electron density map were omitted from the phase calculation in the next refinement cycle. After several cycles of rigid body, positional refinement, and manual model adjustment, the R-factor dropped to 25.4%, and a 2Fo-Fc Fourier map looked quite good. Then the resolution was extended to 2.2 Å. After the chromophores were fitted in the map and followed by several cycles of positional refinement and model adjustment, the electron density improved further. At this stage almost all side chains were well defined except those on the surface. Residue exchanges were carried out at this stage according to the omit map. After several cycles of positional refinement and model adjustment, the R-factor was converged to 24.0%, the individual B-factors were then refined, and the R-factor dropped to 21.5%. 169 water molecules were added to the model according to the Fo-Fc and 2Fo-Fc maps, and the final R-factor of the model was 19.3% (R-free factor was 26.9%) in the range of 10 Å to 2.2 Å.

RESULTS AND DISCUSSION
Amino Acid Sequence-Because the amino acid sequence of APC from P. yezoensis is still unknown, the following six APC amino acid sequences were used to get a consensus sequence for model building. Among these sequences, four are from cyanobacteria, Anabaena cylindrica (14), Calotrix PCC7601 (15), Fischerella PCC7603 (16), and Synechococcus PCC6301 (17), and two are from red algae, Aglaothamnion neglectum (18) and Cyanidium caldarium (19). The alignment of these six sequences is shown in Table I.
Quality of the Model-The final crystallographic R-factor for APC-PY model is 19.3%, in the range of 10.0 Å to 2.2 Å. The Luzzati plot gives a mean positional error of 0.26 Å (20). The r.m.s. deviations of bond lengths and bond angles are 0.015 Å and 2.9°, respectively. The quality of the final model is summarized in Table II. The Ramachandran plot shows that all dihedral angles fall into most favored or allowed regions with the only exception of ␤77Thr ( Fig. 1) (21), which has a conserved unusual dihedral angle in all known phycobiliprotein structures. In APC-PY, the N atom of ␤77Thr forms a hydrogen bond with OD (the oxygen atom in the ring D of chromophore)   oxygen atom of ␣84PCB. The electron density of this residue is well defined in APC-PY. The consensus amino acid sequence (Table I) was used to build the initial model and later modified according to the electron density map. In the 2.3-Å-resolution crystal structure of APC-SP (9), 28 residues were not well defined with 102 atoms of zero occupancy; these residues are ␣25 Asp, ␣35Glu, ␣36Arg, ␣49Glu, ␣50Arg, ␣53Lys, ␣54 Gln, ␣76 Tyr, ␣79 Asp, ␣120Lys, ␣127Glu, ␤2 Gln, ␤10Asn, ␤17Lys, ␤20 Asp, ␤25 Gln, ␤35Glu, ␤36Leu, ␤39Arg, ␤50Asn, ␤58Lys, ␤65 Asp, ␤68Arg, ␤116Lys, ␤117Glu, ␤131 Gln, ␤138Glu, ␤150Lys. The fit of these residues to our electron density map is better in APC-PY; most of them behave well at 1 density level (Fig. 2); others behave well at 0.7 density level except ␣76 in the loop, which has a density at 0.5 density level.
Comparing the crystallographic sequences of the final model of APC-PY and APC-SP, there are 37 nonidentical residues, 25 in the ␣ subunit and 12 in the ␤ subunit (Table III). In comparison with other APC sequences, the 37 residues of APC-PY are more conserved than those of APC-SP. For example, ␣52Val and ␣61Gln of APC-PY are identical to other known sequences. The electron density of these two residues in APC-PY are well defined.
Molecular Structure-The asymmetric unit of APC-PY contains ␣ and ␤ subunit. The ␣ subunit is composed of 160 residues, and the ␤ subunit contains 161 residues. Three ␣␤ monomers are arranged around a 3-fold axis to form a disc shaped (␣␤) 3 trimer of 30 Å in thickness and 110 Å in diameter with a cave in the center. The ␣ and ␤ subunits in the ␣␤ monomer have similar structures, with nine ␣-helices (X, Y, A, B, E, F, F', G, H) separated by irregular loops (Fig. 3).
In the APC-PY crystal, two trimers associate face to face into the (␣␤) 6 hexamer through a crystallographic dyad perpendicular to the triad. There are three (␣␤) 6 hexamers in a unit cell, locating at (0,0,0), (2/3, 1/3, 1/3), and (1/3, 2/3, 2/3) (Fig. 4a). The assembly of two trimers in this hexamer is completely different from that of APC-SP. In APC-SP crystal, the two trimers are associated loosely through ␤ subunits in a "back to back" manner (Fig. 4b) in the hexamer (9), but in APC-PY crystal, the two trimers in the hexamer contact through ␣ subunits, and the assembly of the two trimers is much tighter than that in APC-SP.  The assembly of the hexamer in the APC-PY crystal is similar to that of C-PC from Fremylla diplosiphon (C-PC-FR) and R-PE from Polysiphonia urceolata (R-PE-PU) hexamers; two (␣␤) 3 trimers associate face to face in the hexamer. The ␣ subunits provide the contacting surface, and the two trimers fit complementarily in the hexamer.
Despite the similarity in assembly in APC-PY, C-PC-FR, and R-PE-PU hexamers, the superposition of the C ␣ atoms of APC-PY and C-PC-FR hexamers shows that the assembly of the two trimers in APC-PY hexamer is obviously looser than that in C-PC-FR and R-PE-PU hexamers. The calculated accessible areas between the trimers in C-PC-FR and R-PE-PU hexamers are about 5900 and 6900 Å 2 . In APC-PY hexamer, this value is about 3200 Å 2 , which is much bigger than that in APC-SP (600 Å 2 ); thus, the APC-PY hexamer can be described as a "loose hexamer." The interactions between the trimers in APC-PY hexamer are different from those in C-PC-FR and R-PE-PU hexamers.
First, the number of the residues involved in the interactions between the two trimers in APC-PY hexamer is smaller than that in C-PC-FR and R-PE-PU hexamers, indicating a weaker association. This is consistent with the calculated accessible areas between the two trimers in C-PC-FR, R-PE-PU, and APC-PY hexamers. The special polar network present in C-PC of Agmenellum quadruplaticum (C-PC-AQ), formed by residues 1␤46 Asn-6␣164 Asn-1␣21 Asn-6␣161 Glu-6␣33 Glu-6␣30 Arg (6) is not conserved in APC-PY hexamer. In addition, the electrostatic interactions between 1␣2Lys-6␣23Glu, 1␣17Arg-6␣108 Asp, and 1␣120Arg-4␣174 C-terminal carboxyl group, which were suggested to be involved in the hexamer formation in C-PC-FR (22), are also not present in APC-PY hexamer. Furthermore, the comparison of APC-PY with C-PC-FR and R-PE-PU reveals that all the conserved polar and ionic inter- actions between the two trimers in C-PC-FR and R-PE-PU hexamers are not present in APC-PY hexamer.
Chromophores ␣84PCB and ␤84PCB-In APC, two phycocyanobilins are covalently bound to cysteine residues at position ␣84 and ␤84 (Fig. 5). Both chromophores are well defined in APC-PY (Fig. 6). The geometry and protein environment of these two chromophores resemble those of APC-SP. The ␣84 PCB chromophores have a protein environment similar to that of ␤84PCB. The polar and ionic protein-chromophore interactions in ␣84PCB and ␤84PCB are shown in Table IV.
Chromophores ␣84PCB and ␤84PCB have similar hydrophobic environment; there are three aromatic residues close to ␣84, such as ␣90 Tyr, ␣91 Tyr, and ␣119 Tyr, and three close to ␤84, such as ␤90 Tyr, ␤91 Tyr, and ␤119 Tyr. In C-PC-FR, ␣90 and ␣91 are all Tyr, and ␤90 and ␤91 are all Ile. In R-PE-PU, ␣90 and ␣91 are His and Tyr, respectively, and ␤90 and ␤91 are all Ile. But in all known APCs, ␣90, ␣91, ␤90, and ␤91 are all Tyr. So the microenvironment of ␣84 and ␤84 in APC-PY is similar to that in C-PC-FR and R-PE-PU, indicating that ␣84PCB and ␤84PCB have similar conformation and spectrum character.
␤90 Tyr and ␤91 Tyr stabilize the ␤84PCB ring D conformation, which may make PCB have different spectrum characteristics in APC and C-PC.
Energy Transfer-There are 12 PCBs in APC-PY hexamer; the arrangement of these chromophores is shown in Fig. 7. The theory of shot-distance exiton interaction (23) and long distance dipole-dipole resonance mechanism (24) has been used to explain the energy transfer rate between chromophores.
Inside trimer of APC-PY, the distance between 1␤84PCB and 2␤84PCB in APC-PY is about 34 Å and that between 1␣84PCB and 2␤84PCB is about 21 Å. These values are similar to those in C-PCs. The chromophores are too far away to have exiton interaction. It is also difficult to explain why exiton interaction exists in APC but not in C-PC. Our study of chromophoreprotein interactions and comparison of microenvironments in R-PE-PU, C-PC-FR, and APC-PY show that almost all the chromophore-protein interactions exist within the same monomer (␣␤), the only exception being ␣84PEB in R-PE-PU, which  forms a hydrogen bond with ␤77 Thr in another monomer. In C-PC-FR, the situation is the same as in R-PE-PU. However, it is different in APC-PY; its ␣84PCB forms five hydrogen bonds with the residues in other monomer, such as ␣84PCB O2B-2␤62 Tyr OH, O1C-2␤62 Tyr N, O1C-2␤67 Thr OG1, O2C-2␤67 Thr OG1, and OD-2␤77 Thr N (see Table IV). We believe this difference may explain why the spectrum of APC changes greatly when its monomers associate to trimer. In APC-SP, distances of ␣84PCB O2B-2␤62 Tyr OH, O1C-2␤62 Tyr N, O1C-2␤67 Thr OG1, O2C-2␤67 Thr OG1, and OD-2␤77 Thr N are all within the distance of hydrogen bond formation. As we know, ␤62 Tyr and ␤67 Thr are close to chromophore ␤84PCB and may control the conformation of chromophore and bridge between ␣84 PCB and ␤84 PCB to make the exiton interaction occur.
The distances of chromophores between the two trimers in APC-PY hexamer are similar to those in C-PC-FR and R-PE-PU hexamers. Based on the 1.9-Å resolution crystal structure, the possible pathway of energy transfer within and between the two trimers of R-PE-PU were discussed (25). There are three pairs of short distance interactions between two trimers, such as 1␣84 3 4␣84,1␣140a 3 6␤155, and 1␤155 3 6␤155. ␣84PEB is on the inner surface of R-PE-PU, and 1␣84PEB 3 4␣84PEB may be the dominant energy transfer pathway between the two trimers. Similar energy pathways (1␣84PCB 3 4␣84PCB) also exists in C-PC-FR hexamers (26). In C-PC-FR, R-PE-PU, and APC-PY hexamers, the distances between 1␣84 (C10 atom) and 4␣84 (C10 atom) are 27.5, 28.7, and 30.3 Å, respectively, which are comparable. Therefore, the distance between the two chromophores of APC-PY seems adequate for effective energy transfer.
In addition to the energy pathway composed of chromophores, the aromatic pathway formed by aromatic residues may play an important role in energy transfer. The energy transfer from chromophores to aromatic residues vice versa can be explained by exiton interaction mechanism, because the distances between some chromophores and aromatic residues, such as ␣84PCB-␣90 Tyr, ␣84PCB-␣91 Tyr, ␤84PCB-␤90 Tyr, ␤84PCB-␤91 Tyr are very short (ϳ4 Å). Förster's dipole-dipole resonance transfer can occur between different aromatic residues rather than between chromophores and aromatic residues, because the overlap integral between chromophore absorption ( max Х 650 nm for fluorescence spectrum) and aromatic residue emission ( max Х 300 nm for fluorescence spectrum and max Х 400 nm for phosphorescence spectrum) is quite small. In APC-PY there are two areas abundant in aromatic residues as shown in Fig. 8. One is close to chromophore ␣84PCB and composed of ␣164 Phe, ␣165 Tyr, ␣166 Phe, ␣168 Tyr, ␣90 Tyr, ␣91 Tyr, ␣97 Tyr, ␣119 Tyr, and ␤18 Tyr. Another is close to chromophore ␤84 and composed of ␤165 Tyr, ␤166 Phe, ␤168 Tyr, ␤90 Tyr, ␤91 Tyr, ␤94 Tyr, ␤97 Tyr, ␤119 Tyr, and ␣18 Tyr. ␣164 Phe is involved in the hydrophobic interactions between the two trimers. The aromatic residues in the ␣ subunit and the ␤ subunit have high homology and similar locations. Other aromatic residues are on the periphery of the disc, such as ␣76 Tyr, ␣60 Phe, ␤76 Tyr, ␤62 Tyr, ␤30 Tyr, ␤31 Phe, ␣30 Phe, and ␤81 Tyr; among them ␣76 Tyr and ␤62 Tyr may mediate the energy transfer between 1␣84PCB and 2␤84PCB. Because PC and APC are connected perpendicularly in vivo, aromatic residues on the periphery may mediate the energy transfer from the chromophores of PC to those of APC.
Functional Unit-In the core cylinders of phycobilisomes, several APC trimers are close together, but the association manner of these APC trimers is still unknown. Based on dissociation experiments, it was suggested that allophycocyanin does not form hexamers (27), because almost all the residues involved in the trimer-trimer aggregation in C-PC-AQ and C-PC-FR hexamers are not conserved in APC. Similar conclusions were reported later (6,9). In the APC-PY hexamer, all the interactions involved in the formation of C-PC-AQ and C-PC-FR hexamers and all the conserved polar and charged interactions in C-PC-FR and R-PE-PU hexamers are not present, but APC-PY can still associate face to face to form a hexamer, which is maintained by some polar and charged in- teractions, different from those in C-PC-FR and R-PE-PU. Because the distances of chromophores between the two trimers in this hexamer are also adequate for effective energy transfer, we assume that the loose hexamer may represent the basic unit of APC in physiological conditions. It is possible that linker proteins may help to stabilize the loose hexamers.