Abstract
Prochlorococcus, the most abundant genus of photosynthetic organisms1, owes its remarkably large depth distribution in the oceans to the occurrence of distinct genotypes adapted to either low- or high-light niches2, 3. The pcb genes, encoding the major chlorophyll-binding, light-harvesting antenna proteins in this genus4, are present in multiple copies in low-light strains but as a single copy in high-light strains5. The basis of this differentiation, however, has remained obscure. Here we show that the moderate low-light-adapted strain Prochlorococcus sp. MIT 9313 has one iron-stress-induced pcb gene encoding an antenna protein serving photosystem I (PSI)—comparable to isiA genes from cyanobacteria6, 7—and a constitutively expressed pcb gene encoding a photosystem II (PSII) antenna protein. By comparison, the very low-light-adapted strain SS120 has seven pcb genes encoding constitutive PSI and PSII antennae, plus one PSI iron-regulated pcb gene, whereas the high-light-adapted strain MED4 has only a constitutive PSII antenna. Thus, it seems that the adaptation of Prochlorococcus to low light environments has triggered a multiplication and specialization of Pcb proteins comparable to that found for Cab proteins in plants and green algae8.
In order to gain a better understanding of the origin, function and localization of the divinyl-chlorophyll a/b-binding antenna complexes of Prochlorococcus species and how these properties relate to the light niche to which different strains are adapted, we have undertaken gene expression and structural studies on the moderate low-light-adapted strain MIT 9313 for which the full genome sequence is available (http://www.jgi.doe.gov/JGI_microbial/html/). This strain contains two pcb genes: pcbA (PMT 1046) and pcbB (PMT 0496). This contrasts with the very low-light-adapted strain SS120 that contains eight pcb genes (pcbA to pcbH)—analysis of its genome sequence (http://www.sb-roscoff.fr/Phyto/ProSS120/) revealed the presence of one more pcb gene (pcbH) than previously thought5—and the high-light-adapted strain MED4, which has only a single pcb gene (pcbA)4, 5.
Recently it was shown by electron microscopy and single-particle analyses that SS120 contains a giant supercomplex consisting of the PSI reaction centre trimer surrounded by a light-harvesting antenna ring composed of 18 Pcb subunits9, similar to the 18-mer IsiA–PSI supercomplex induced in cyanobacteria when deprived of iron6, 7. However studies on MIT 9313 grown under similar conditions to SS120 did not reveal the presence of an 18-mer Pcb–PSI supercomplex but only 'naked' trimeric PSI complexes, which matched with the cyanobacterial X-ray structure10 (Fig. 1a, b). Instead, electron microscopy (Figs 1c, d and 2a) indicated that Pcb proteins associate with the dimeric reaction centre complex of PSII to form a Pcb–PSII supercomplex having dimensions of approximately 210 
290 Å. Our interpretation is that this Pcb–PSII supercomplex consists of eight Pcb subunits with four distributed on each side of the PSII dimer as shown in Fig. 1c. This is emphasized by overlaying onto the projection map the published X-ray-derived models of the PSII reaction centre dimer and CP43, a PSII antenna protein structurally similar to Pcb9, 11 (Fig. 1d). In some cases, the four Pcb subunits on one side of the dimer were missing (Fig. 1e, f ). The 'naked' PSI trimers and Pcb–PSII complexes shown in Fig. 1 were located in a chlorophyll-containing band (band 2 in Fig. 2b, insert +Fe) obtained by sucrose density centrifugation after solubilizing isolated thylakoid membranes with the detergent 
-d-dodecyl maltoside. Also contained in this band were some PSII reaction centre dimers free of Pcb proteins (Fig. 1g, h ). Analysis of all discernible particles, taken from band 2 sample micrographs, resulted in 1,192 particles assigned to PSI and PSII, and gave a PSI:PSII ratio of about 2. We assume this to be indicative of the ratio in the intact thylakoid membrane given that most PSI and PSII particles were in band 2. Amino-terminal sequencing of the Pcb protein in band 2 from +Fe conditions showed it to be the product of the pcbA gene only, a result which was also found for the free-Pcb proteins in band 1 and for Pcb protein in thylakoid membranes (Fig. 3).
Figure 1: Characteristic top view of negatively stained particles isolated from Prochlorococcus MIT 9313.
a–j, Particles isolated from cells grown in the presence (a, c, e, g) or absence (i) of iron viewed by electron microscopy and shown after overlaying the X-ray structures of the PSI trimer (green, b, j)11, PSII core dimer (green, d, f, h) and CP43 (red/orange)12. The Pcb proteins are coloured red or orange to identify them as products of the pcbA (PMT 1046) or pcbB (PMT 0496) genes respectively (see text). Scale bar, 100 Å.
High resolution image and legend (126K)Figure 2: Electron micrographs of particles isolated from Prochlorococcus MIT 9313 separated on sucrose density gradients.
a, b, Band 2 (a) of cells grown in the presence of iron (+ Fe) containing both PSII particles (white arrow) and naked PSI trimers (black arrow), and band 3 (b) of the sucrose density gradient from cells grown in iron-deficient medium (- Fe) containing the 18-mer Pcb–PSI supercomplex (black arrow). The insert is the banding profile obtained by sucrose density gradient centrifugation of thylakoid membranes solubilized with 1% -d-dodecyl maltoside isolated from Prochlorococcus MIT 9313. Band 1 consists of free Pcb proteins. Scale bar, 500 Å.
Figure 3: SDS–PAGE (left panel) and western-blot analyses (right panel) of thylakoid membranes isolated from Prochlorococcus MIT 9313 cells grown in the presence (+ Fe) and absence (- Fe) of iron.
![Figure 3 : SDS|[ndash]|PAGE (left panel) and western-blot analyses (right panel) of thylakoid membranes isolated from Prochlorococcus MIT 9313 cells grown in the presence (|[plus]| Fe) and absence (- Fe) of iron. Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, or to obtain a text description, please contact npg@nature.com](/nature/journal/v424/n6952/images/nature01933-f3.0.jpg)
The expression of the pcbB (PMT 0496) gene under -Fe conditions gives rise to an additional band in the SDS–PAGE gel, where PcbA and PcbB proteins are marked by black and white asterisks, respectively. The identity of these proteins was confirmed by N-terminal sequencing giving MQTYGNPN for PcbA and MQTYGKTD for PcbB. The antibodies for detecting the level of PSII were raised against PsbA (D1 protein) and for PSI raised against PsaC (FA/FB protein). The blotting indicates that the PSI:PSII ratio remains about the same for +Fe and -Fe cells.
High resolution image and legend (44K)The absence of the PcbB protein and the 18-mer Pcb–PSI supercomplex in iron-replete MIT 9313 cells spurred us to investigate the expression of pcbA and pcbB genes in this strain. When the cells were grown in medium supplemented with iron (+ Fe), we found that only the pcbA gene was expressed (Table 1). However, when cells were transferred to culture medium without added iron (- Fe), expression of the pcbB gene was activated, a surprising result as pcb genes of Prochlorococcus were not known to be regulated by iron as is the iron-stress-induced isiB gene (Table 1). The latter gene encodes flavodoxin12, which substitutes for ferredoxin as an electron acceptor to PSI, and its expression indicates that the cells had acclimatized to conditions of iron depletion. On the other hand, the expression of the pcbA as well as the psaA and psbA genes, encoding PSI and PSII reaction centre proteins, respectively, were downregulated (Table 1).
![Table 1 - Quantitative RT|[ndash]|PCR analyses of the expression of the indicated genes under |[plus]|Fe and -Fe conditions](/common/images/table_thumb.gif)
The question arising from these expression studies is: where is the PcbB protein targeted in cells exposed to low iron levels? As before, thylakoid membranes were isolated, solubilized and subjected to sucrose density gradient centrifugation. As shown in insert -Fe of Fig. 2b, an additional chlorophyll-containing band was observed compared with the iron-supplemented cells (band 3). Electron microscopy (Fig. 2b) revealed that this band contained the 18-mer Pcb–PSI supercomplex (Fig. 1i, j), similar to that observed in Prochlorococcus SS120 (ref. 9). SDS–polyacrylamide gel electrophoresis (PAGE) and N-terminal sequencing indicated that the Pcb protein of the PSI supercomplex is derived from the pcbB gene. This gene product was readily observed in SDS–PAGE profiles of thylakoid membranes isolated from cells grown under iron deficiency and shown in Fig. 3 (white asterisk). It ran at a slightly higher apparent molecular mass compared with the PcbA protein observed in both +Fe and -Fe cells (black asterisk), consistent with the difference in their predicted molecular masses of 38,511 Da (PcbA) and 40,737 Da (PcbB). Notably, despite the downregulation of the pcbA gene under iron-depleted conditions, a relatively high level of the PcbA protein was detected by SDS–PAGE (see Fig. 3) and Pcb–PSII structures similar to those present in band 2 of iron-supplemented (see Fig. 2, insert +Fe) cells were observed by electron microscopy and image analysis in iron-deprived cells. Intriguingly, although in cyanobacteria the induction of the 18-mer IsiA–PSI supercomplex under iron deficiency seems to compensate for the reduction in the level of PSI relative to PSII and for a decrease in the synthesis of phycobiliproteins, these reasons cannot apply to Prochlorococcus MIT 9313. Immunoblotting analyses (Fig. 3) showed that the PSI:PSII ratio, estimated to be about two, remained approximately the same in +Fe and -Fe conditions despite the lowering of the expression of psaA and psbA genes in iron-depleted cells (Table 1).
Phylogenetic analyses of the pcb/isiA/psbB/psbC gene superfamily5, 13 indicate that the pcbA and pcbB genes of MIT 9313 partition into two different clusters and that the latter is very closely related to the pcbG gene of Prochlorococcus SS120, which is the gene providing the Pcb protein of the 18-mer Pcb–PSI supercomplex of this strain under iron-replete conditions, as indicated by SDS–PAGE and N-terminal sequencing (data not shown). In the case of MED4 we have been unable to detect an 18-mer pcb–PSI supercomplex either under iron-rich or iron-depleted conditions. Indeed in this strain the expression of the single pcb gene (pcbA) was not significantly effected by iron depletion (Table 1). In fact, using electron microscopy and associated image analyses of single particles we found that the Pcb proteins of MED4 associated with PSII in a similar fashion to those of MIT 9313.
Therefore we conclude that the PcbA protein of MIT 9313, similar to the PcbA protein of MED4, is targeted to PSII where they interact with the reaction centre dimer and increase the light-harvesting capacity of this photosystem. In contrast the PcbB protein of MIT 9313, similar to the PcbG protein of SS120, is targeted to PSI where it forms an 18-mer light-harvesting antenna ring around the PSI reaction centre trimer.
Under iron depletion conditions, expression of the pcbC gene of strain SS120 strongly increased while expression of pcbG was downregulated (Table 1). It is therefore possible that PcbC replaces PcbG in Pcb–PSI supercomplexes under these conditions.
The observations reported here give support to the hypothesis that the common ancestor of Prochlorococcus and marine Synechococcus possessed an isiA-like gene, similar to that found in other cyanobacteria12, even if, surprisingly, no isiA is present in the only currently available genome of a marine Synechococcus (strain WH8102; http://www.jgi.doe.gov/JGI_microbial/html/). The difference in Pcb structure between strains can be interpreted in terms of their respective light niche. For SS120, having both constitutive PSI and PSII antennae possibly confers to this strain an adaptive advantage to grow under the very low irradiances found at the bottom of the euphotic zone3. In the high-light-adapted strain MED4, only a PSII-related pcb gene is required because in the upper, well-illuminated layer of the ocean an additional antenna for PSI may not be needed. Indeed, according to the recent X-ray structure10, the 'naked' cyanobacterial PSI reaction centre already binds almost 100 light-harvesting chlorophyll molecules. The intermediate situation found in MIT 9313 is harder to interpret, as it appears to have an antenna ring around PSI only under iron depletion. Perhaps this suggests that there is a more complex inter-relationship between light intensity and availability of iron in the oceanic environment than hitherto considered.
The apparent requirement of at least some Pcb proteins to associate with PSII in Prochlorococcus is in line with the relatively low level of about 30 light-harvesting chlorophylls bound to PSII (ref. 12). Therefore in all types of photosynthetic organisms the light-harvesting capacity of PSII usually needs to be boosted by additional antenna systems: chlorophyll a/b-binding Cab proteins of plants and green algae8 and phycobiliproteins of cyanobacteria and red algae14. In the case of Prochlorococcus, each Pcb subunit adds an additional 13 chlorophylls assuming that they bind the same level of this pigment as CP43 (ref. 12). Consequently it takes the binding of just two Pcb subunits to approximately double the light-harvesting capacity of PSII.
Methods
Isolation and biochemical characterization
Prochlorococcus sp. MIT 9313 (ref. 2) was grown in PCR-S11 medium1. Iron-stressed cultures were obtained by two consecutive transfers of the cells into fresh medium containing only 1% of the standard FeCl3 concentration of the PCR-SII medium but having the normal Na2-EDTA concentration. Cells in the exponential phase of growth were collected and thylakoid membranes isolated according to ref. 15. The thylakoids (1 mg chlorophyll ml-1) were then solubilized with 1% -d-dodecyl maltoside at 4 °C for 10 min. The solubilized membranes were centrifuged at 100,000g using a Beckman Ti70 rotor and the solubilized fraction subjected to sucrose density centrifugation6. The resulting bands were independently removed for biochemical and structural characterization. SDS–PAGE and western blotting were performed according to ref. 16. Antibodies were a gift from P. Nixon (anti-D1) and J. Golbeck (anti-PsaC). N-terminal sequencing of proteins was performed by J. Keen.
RNA isolation and RT–PCR analysis
RNA was isolated from frozen cell pellets as previously described17. Quantitative RT–PCR reactions were performed on an ABI Prism 5700-sequence detection system (PE Applied Biosystems) as detailed elsewhere18 using specific primers of selected photosynthetic genes (see Table 1), defined according to their sequences in MIT 9313, MED4 and SS120 genome data banks. For each quantitative RT–PCR experiment, measurements were made in duplicate and experiments were repeated on two distinct cultures.
Electron microscopy
Isolated preparations were placed onto carbon-evaporated, glow-discharged, 300-mesh copper grids and negatively stained with 2% uranyl acetate and imaged using a Philips CM 100 electron microscope at 80 kV. The magnification was calibrated to be 50,850. Five electron micrographs were taken for each preparation where the first minima of the contrast transfer function was calculated as being in the range 21–22 Å. Micrographs were digitized using a Leafscan 45 densitometer at a step size of 10 
m. All image processing was performed within the IMAGIC-5 software environment19 at a sampling frequency of 3.92 Å per pixel on the specimen scale. Single-particle data sets of approximately 1,200 (Pcb–PSI supercomplex; -Fe, band 3), 5,900 (+ Fe, band 2) and 1,600 (- Fe, band 2) were obtained by selecting all possible single particles from the micrographs. Reference-free alignment and multi-variate statistical classification20 were used to identify the initial class averages used for iterative refinement that resulted in the improved class averages shown.
Image processing
Coordinate data sets were obtained from the RCSB Data Bank (http://www.rcsb.org) under the entry codes for 1JB0 (PSI 2.5 Å structure10) and 1IZL (PSII 3.7 Å structure11). These structural models were visualized using the program ViewerLite (Accelrys Inc.) and overlaid onto the calculated electron-microscopy-derived two-dimensional projection maps at the same scale. The coordinates of CP43 were extracted from the 1IZL.pdb file and modified by the removal of the large loop joining helices V and VI so as to better represent a typical Pcb protein. These coordinates were modelled into the centre of mass assigned to each Pcb subunit.