HSP90C interacts with PsbO1 and facilitates its thylakoid distribution from chloroplast stroma in Arabidopsis

Arabidopsis plastidic HSP90C is an HSP90 family molecular chaperone that is required for chloroplast development and function. To understand the mechanism of action of HSP90C within the chloroplast, we conducted a yeast two-hybrid screening and revealed it interacts directly with the photosystem II extrinsic protein PsbO1, which performs a canonical function in the thylakoid lumen. To understand the biological significance of HSP90C-PsbO1 interaction, we investigated the role of HSP90C in modulating the stromal and thylakoid distribution of PsbO1GFP fusion protein. Fusion to GFP significantly delays the PsbO1 thylakoid transport and induces a variegation phenotype. Overexpression of HSP90C promotes the thylakoid distribution of PsbO1GFP and alleviates the leaf variegation. By tracking the chloroplast maturation during photomorphogenesis, we observed PsbO1GFP tends to form distinct fluorescent clusters within the stroma with delayed thylakoid membrane biogenesis, while HSP90C overexpression corrects these adverse effects. We also demonstrated that active HSP90C function is specifically required for stable accumulation of mature PsbO1GFP in thylakoid by using specific inhibitor geldanamycin. This study therefore not only identified novel HSP90C interactors, but also reports for the first time that PsbO1 enroute from the cytoplasm to thylakoid lumen is tightly regulated by the HSP90C chaperone complex in plastid stroma; whereas the proper HSP90C homeostasis is also critical for chloroplast maturation and function.


Introduction
Chloroplast biogenesis is generally characterized by the phenotypic greening of proplastids or etioplasts and de novo formation of thylakoid membranes [1]. At the molecular level, this morphological transition requires the cytosolic synthesis of a large set of chloroplast-targeted proteins, as the majority of chloroplast proteins are nuclear-encoded. Upon successful import into the chloroplast, many proteins must undergo folding, assembly, and thylakoid transport for proper function, or immediate degradation. These processes necessitate chloroplast quality control systems that include both molecular chaperones and proteases [2,3]. Malfunction of protein quality control components have been shown to impair chloroplast function and plant development [4][5][6]. As an extreme consequence, chloroplast proteins may also undergo bulk degradation through senescence associated vacuoles (SAVs) [7], autophagy [8], or CV (chloroplast vesiculation)-containing vesicles [9], especially under adverse environmental conditions. HSP90C is an HSP90 family heat shock protein located in the plastid of higher plants and green algae. HSP90C does not show very high similarity to cytosolic homologues [10], particularly at the extreme C-terminal ends [11]. However, it forms a "foldosome" complex consisting of HSP70B, CDJ1 and CGE1 [12,13], mimicking the cytosolic HSP90 protein complexes required for substrate folding [14]. In the flowering plant Arabidopsis, HSP90C is located in the chloroplast stroma [15] and is required for protein import through the TOC/TIC complex [16]. Seedlings with reduced HSP90C expression caused by transgene-induced gene silencing manifest a variegated phenotype while HSP90C T-DNA insertion homozygous knockouts are embryonic-lethal [11,16,17]. Additionally, an early study on point mutation line cr88 indicated that HSP90C malfunction negatively impacts the expression of light-induced nuclearlocalized genes encoding chlorophyll a/b binding protein (CAB), small subunit of ribulose bisphophate carboxylase (RBCS) and NR2, resulting in delayed de-etiolation, underdeveloped plastids and yellow cotyledons [18]. These suggest HSP90C has pleiotropic effects in chloroplast maturation and physiology.
HSP90 family proteins generally aid in the late stage of protein folding [19] and its in vivo function is revealed by its client proteins and/or cochaperones, which are collectively termed HSP90 binding partners [14]. Both genetic and physical interactors of either the cytosolic or organellar HSP90 isoforms for fungi and human cells have been extensively studied by high throughput analyses [20][21][22][23][24][25]. However, known interactors of plastidic HSP90C are still limited. Aside from the interactors forming the foldosome [12], HSP90C has been shown to interact with Tic110, Tic40 [16] and VIPP1 (vesicle-inducing proteins in plastid 1) [17]. As a stromal protein involved in protein import through the TIC complex, it is expected that HSP90C may interact with many plastidic proteins. However, because of the developmentally regulated expression [11], it is difficult to study the role of HSP90C in planta.
In order to study the impact of altered HSP90C expression, we previously generated HSP90C overexpression lines and demonstrated that HSP90C overexpression increased the plant sensitivity to salt, osmotic and high calcium stresses [26]. It was also shown that the expression level of HSP90C in young leaves are tightly regulated [11]. To further understand the role of HSP90C in plant growth and development, we performed a yeast two-hybrid screening for Arabidopsis HSP90C interactors in this study. It was identified that PsbO1, a lumen-targeted subunit of photosystem II (PSII) interacts with HSP90C. Additionally, by using PsbO1GFP fusion protein, whose thylakoid transport is delayed compared to native PsbO1, we visualized how this delay negatively impacts chloroplast development in vivo and HSP90C homeostasis in chloroplast. We further showed that overexpression of HSP90C alleviates leaf variegation induced by PsbO1GFP overexpression and facilitates thylakoid development. We also analyzed chloroplast maturation during photomorphogeneis and provided evidence that HSP90C level is critical in maintaining chloroplast protein homeostasis and proposed a model of the HSP90C role in guiding PsbO1 targeting from cytoplasm to the thylakoid lumen.
domains (HSP90C-MC) as bait and a prey cDNA library constructed from inflorescence tissues [27]. After three rounds of screening, six HSP90C candidate interactors were obtained reproducibly (S1A Fig). Out of these interactors, two proteins encoded by genes At4G23050 (Per-Arnt-Sim domain-containing kinase) and At1G34770 (melanoma-associated antigen protein), were previously reported as AtHSP90C interactors from Arabidopsis Interactome Consortium [28]. The other four interactors are encoded by At5G08170 (agamatine deiminase), At4G21960 (PRXR1, Peroxidase 42), At5G66570 (PsbO1 or oxygen evolving complex 33) and At2G05100 (light harvesting complex B subunit 2, LHCB2). PsbO1 is the major isoform of the two PsbO proteins in Arabidopsis [29]. It is localized in the thylakoid lumen with no known molecular chaperone being previously reported to aid in its chloroplast targeting, thylakoid transport or assembly into photosystem II complex [30]. We therefore chose to conduct in-depth analyses of the HSP90C-PsbO1 interaction.
PsbO1 is nuclear-encoded and the pre-protein contains a chloroplast-targeting peptide and a thylakoid-targeting peptide which are both cleaved after the successful translocation ( Fig  1A). Interestingly, the identified PsbO1 clone in our screen contains only part of the chloroplast targeting sequence and bears a T200A point mutation in the middle of the mature protein (shown as PsbO1 Ã in Fig 1A and S1A Fig). To rule out the possibility that the interaction is mediated only by the chloroplast/thylakoid targeting sequence, we re-constructed the full length and the mature form of PsbO1 T200A and confirmed that HSP90C interacts with both the pre-protein and mature forms by yeast two-hybrid ( Fig 1B) and co-immunoprecipitation from yeast cell lysates (S1B Fig). We also corrected the T200A point mutation by site-directed mutagenesis and showed that the native PsbO1 mature form interacts with HSP90-MC, while not with the HSP90C N-terminal domain alone (Fig 1C). Further analyses by in vitro size exclusion chromatography with purified proteins also indicated that HSP90C shifts the elution profiles of both native and mutant PsbO1 to the high molecular weight range, with an indication that HSP90C interacts more strongly with the mature form of PsbO1 T200A (Fig 1D). Collectively, these proteomics and biochemical data demonstrated that the chloroplast stromalocalized HSP90C directly interacts with the chloroplast lumen-localized PsbO1.

Expression of PsbO1GFP caused delayed cotyledon greening and variegation in vegetative leaves
PsbO1 is not well-folded in solution and it is normally rapidly processed in the stroma with only the mature form readily observable in wild-type chloroplasts [29,31,32]. PsbO is translocated to the thylakoid lumen in an unfolded state by the SEC complex located in thylakoid membrane [33] To better study the stromal interaction between HSP90C and PsbO1 in vivo, we constructed a series of PsbO1-GFP fusion proteins (Fig 2A) that have been shown to delay thylakoid transport in isolated chloroplasts [34,35]. In order to monitor the processing of PsbO1GFP in planta, we also constructed GFP fusion proteins for the full-length pre-protein form (pPsbO1GFP), the stromal intermediate form (iPsbO1GFP) and the thylakoid lumenlocalized mature form (mPsbO1GFP) and expressed them in E. coli for comparison (Fig 2A). The three processing forms of PsbO1GFP expressed in Arabidopsis plants were readily observable (Fig 2A bottom left), indicating delayed chloroplast targeting and thylakoid transport in agreement with previous studies [36]. Fractionation of purified chloroplasts also indicated that iPsbO1GFP is mainly trapped in the stromal fraction (Fig 2A bottom right). Interestingly, we observed that seedlings expressing PsbO1GFP showed delayed cotyledon greening with less chlorophyll content (S2 Fig) and the severity of delay depends on the expression level of PsbO1GFP, particularly on the enrichment of stromal iPsbO1GFP (Fig 2B, right). Nevertheless, the yellowish cotyledon phenotype gradually disappears as the seedlings develop and grow older (Fig 2C, bottom), suggesting the chloroplast maturation is simply delayed. As control lines, seedlings expressing stroma-targeted PsbO1 1-58 GFP or thylakoid-targeted PsbO1 1-85 GFP (Fig 2A) did not show any observable phenotype compared to wild type seedlings, in spite of high accumulation of GFP or PsbO1 59-85 GFP in stroma (Fig 2C). In addition, we examined a psbo1 knockout line and did not observe any severe phenotype at the seedling stage under normal growth conditions (S3 Fig) as reported previously [37]. Taken together, Control constructs of PsbO1 1-58 GFP and PsbO1 1-85 GFP are also shown. The bottom left immunoblot shows individually expressed pPsbO1GFP, iPsbO1GFP and mPsbO1GFP from E. coli and, the different processing forms in senescing plant lysate stably transformed with pPsbO1GFP (OEX11), which in this study is simply referred as to PsbO1GFP and the Wild type seedling lysate control. The bottom right shows immunoblotting detection of HSP90C, PsbO1 and PsbO1GFP (α-GFP) as well as Coomassie blue staining of stroma (Str) and thylakoid (thy) fractions after fractionation of purified chloroplasts from wild type and PsbO1GFP expressing OEX9 seedlings. Total represents the total chloroplast lysate. (B) T2 seedlings of 3 independent transgenic lines (OEX9, OEX11 and OEX14) expressing PsbO1GFP and wildtype control (Col-0), grown on ½ MS medium for 5 days after germination (5-DAG). Immunoblot analysis was performed with a total of 10 μg total proteins by using anti-GFP, anti-PsbO and anti-HSP90C antibodies. (C) T2 seedlings expressing PsbO1GFP, PsbO1 1-58 GFP and PsbO1 1-85 GFP at 4-, 7-and 12-DAG. The top immunoblot for 2μg of total proteins indicates the expression of PsbO1 1-58 GFP and PsbO1 1-85 GFP in transgenic seedlings by using anti-GFP antibody and Ponceau S staining. these results suggest that overexpression of the full-length PsbO1GFP fusion protein is capable of inducing delayed cotyledon greening.
At late developmental stages, no significant phenotype was observed for plants expressing PsbO1 1-58 GFP or PsbO1 1-85 GFP. However, plants expressing full length PsbO1GFP show a variegated phenotype in both rosette leaves ( Fig 3A) and cauline leaves ( Fig 3B). In agreement with the phenotype observed for cotyledons, the severity of the variegation in rosette and cauline leaves is strongly correlated with the expression level of PsbO1GFP, especially the intermediate iPsbO1GFP. Surprisingly, the leaf tissues with severe variegation also contain high levels of HSP90C and HSP70B, the two main chaperones of the foldosome [12], suggesting PsbO1GFP in chloroplasts might be mainly bound in the foldosome complex. Nevertheless, the endogenous mature PsbO1 level is significantly reduced in rosette leaves and there seems to be some accumulation of the stroma-localized native intermediate iPsbO1 (Fig 3A). We also analyzed mesophyll cells of the cauline leaf using fluorescence microscopy and observed an increase in GFP signal from the green tip to whitened base regions (Fig 3C), further indicating that variegation is caused by accumulation of PsbO1GFP.
To better visualize the sub-plastid localization of GFP signal, we attempted to examine individual chloroplasts in mesophyll cells of the 4-DAG cotyledon. We observed a distinct localization of GFP signal that does not overlap well with chlorophyll autofluorescence in seedlings expressing PsbO1GFP (S4A Fig top). Many plastids form stromule-like extensions, almost like a network containing GFP signal (S4A Fig middle). In some cases, GFP signal was observed in punctate clusters inside these stromules (S4A Fig bottom). We took Z-stack confocal images of the live mesophyll cells and observed that some stromules containing GFP clusters appeared to extend from one plastid and approach or connect to another plastid (S4B Fig). Taken altogether, these results indicate that overexpression of PsbO1GFP might stimulate the formation of stromules, which seem to facilitate inter-plastid transport of the fusion protein.

HSP90C overexpression alleviates PsbO1GFP overexpression-induced variegation and reduces chloroplast extensions
The delayed greening process for PsbO1GFP plants resembles the chloroplast molecular chaperone HSP90C point mutation line cr88 [15]. However, this phenotype is distinct from that of the HSP90C cosuppression lines for which the variegated phenotype does not disappear over time [11], suggesting the HSP90C function in PsbO1GFP-expressing lines might be only mildly affected. In order to understand the mechanisms by which the greening process is delayed, we transformed the PsbO1GFP construct into Arabidopsis that has been previously transformed with a FLAG-tagged HSP90C [11], designated as HSP90C FLAG lines in this research. Interestingly, analyses of many independent transgenic lines in multiple FLAGtagged HSP90C transgenic backgrounds did not identify any transgenic plants that have significantly delayed greening process (S1 Table, and Fig 4A). This suggests that overexpression of HSP90C alleviates the PsbO1GFP-induced variegation. We also analyzed guard cells of PsbO1GFP-expressing seedlings. Comparing variegated 4-DAG to less-variegated 7-DAG cotyledons respectively (Fig 4B), we observed that the plastids at 4-DAG have less chlorophyll ( Fig 4B) and are swollen, resembling "ballooned" chloroplasts previously observed in vipp1 mutant lines [38]. Analyses of guard cells of seedlings co-expressing PsbO1GFP and FLAGtagged HSP90C indicated that they possessed smaller chloroplasts than PsbO1GFP alone and more closely resembled the thylakoid-targeted control line expressing PsbO1 1-85 GFP ( Fig 4B  and S5 Fig). To further understand how PsbO1GFP-induced variegation develops over time, we analyzed the expression levels of PsbO1GFP in seedlings aged from 5-to 13-DAG. We observed a substantial decrease of the relative amount of iPsbO1GFP as seedlings became older and co-expression with HSP90C-FLAG greatly accelerated the drop in iPsbO1GFP compared to PsbO1GFP expression alone (Fig 4C and 4D).
It should be noted that the quantitative data in Fig 4C and 4D are from ten independent PsbO1GFP-expressing lines and four independent PsbO1GFP and HSP90CFLAG coexpression lines, respectively. To confirm that the effect of HSP90C overexpression on PsbO1GFP-induced variegation was not due to the differential expression of PsbO1GFP in analyzed lines, we crossed two independent PsbO1GFP lines with a line that overexpresses HSP90C. The F3 generation plants were obtained and it was shown that the presence of an extra HSP90C allele indeed alleviates variegation induced by PsbO1GFP at all developmental stages after germination ( Fig 5A and S6 Fig). Interestingly, compared to wild type and HSP90C FLAG co-expressing plants, PsbO1GFP plants developed slightly smaller vegetative leaves. However, their overall development (e.g. transitioning from juvenile to adult and then flowering), was not significantly affected. Particularly, the rosette leaves of PsbO1GFP plants appeared very similar to WT and the HSP90C FLAG co-expression lines after six weeks ( Fig 5B). Immunoblot analysis of these rosette leaves indicated that stromal iPsbO1GFP levels remain relatively constant with or without HSP90C expression. However, mature mPsbO1GFP accumulated to a greater degree in the presence of FLAG-tagged HSP90C ( Fig  5B). Taken together, these results suggest that HSP90C helps to reduce the PsbO1GFPinduced variegation and facilitates the thylakoid targeting of mature PsbO1GFP, at least in the mature adult leaves.

HSP90C expression affects the spatial targeting of PsbO1GFP in chloroplasts
HSP90C has been well demonstrated to play a role in protein import through the TOC-TIC complex [16]. However, it is not clear whether HSP90C is directly involved in thylakoid protein transport. To confirm immunoblotting results that show the positive role of HSP90C in affecting PsbO1GFP intermediate and mature form processing (Fig 5B), we analyzed threedimensional sub-plastid localization of PsbO1GFP when HSP90C is overexpressed using fluorescence microscopy. Chlorophyll autofluorescence, the thylakoid indicator, does not overlap with signal from RbcS 1-79 GFP, a stroma-targeted GFP control ( Fig 6A). Conversely, PsbO1 1-85 GFP signal partially overlapped with chlorophyll autofluorescence (Fig 6B) confirming that a portion of GFP protein was successfully targeted to the thylakoid membrane as demonstrated previously by others [34,35,39]. GFP signal from full length PsbO1GFP also shows minimal overlap with chlorophyll autofluorescence (Fig 6C), suggesting PsbO1GFP is primarily localized to the stroma. Interestingly, GFP signal from seedlings co-expressing PsbO1GFP and FLAG-tagged HSP90C is mostly overlapped with chlorophyll autofluorescence (Fig 6D), strongly suggesting that overexpression of HSP90C facilitates the attachment or targeting of PsbO1GFP to the thylakoid. cells from PsbO1GFP line, PsbO1 1-85 GFP line and the line co-expressing PsbO1GFP and HSP90C FLAG . Samples are from 4-and 7-DAG seedlings. (C) Representative immunoblot analyses of seedlings expressing PsbO1GFP (left), or co-expressing PsbO1GFP and HSP90C FLAG (right) at 5, 7 and 9-DAG. Anti-FLAG was used to indicate the expression of exogenous HSP90C. (D) Quantitative analysis of relative iPsbO1GFP signals. Four independent transgenic lines were analyzed for those expressing PsbO1GFP and those coexpressing PsbO1GFP and HSP90C FLAG , respectively. The error bars represent standard deviation.
https://doi.org/10.1371/journal.pone.0190168.g004 The effect of HSP90C on etioplast-to-chloroplast development Light is the key factor inducing chloroplast and thylakoid differentiation. When grown in the dark, chloroplast maturation arrests at the etioplast stage as characterized by prolamellar bodies (PLB). Upon exposure to light, the PLB gradually loses its crystallinity and undergoes metamorphosis into primary thylakoids and ultimately thylakoid grana [40,41]. To understand the chloroplast maturation process and the role of HSP90C on PsbO1 processing during the morphological transition, we germinated different plant lines in the dark and then switched them to constant light, so as to synchronize chloroplasts maturation. It was observed that only PsbO1GFP lines experience dramatically delayed greening, while the other transgenic lines that express HSP90C FLAG , PsbO1 1-58 GFP, PsbO1 1-85 GFP, or both PsbO1GFP and HSP90C FLAG green very similarly to the wild type seedlings (Fig 7A).
By tracking the chlorophyll fluorescence, we observed three distinct patterns in wild type and HSP90C FLAG lines over a period of 72 hours after switching to constant light, the "dot," "filamentous" and "granular" stages ( Fig 7B, top) which correspond to prolamellar body, pre-/ pro-thylakoid lamellae and grana formation, respectively, as reported in literature [40][41][42]. The PsbO1GFP line showed a prolonged "filamentous" stage and accumulated less chlorophyll (S7 Fig) within swollen, rod-like structures as opposed to the more uniform dispersal pattern seen in the wild type line. Interestingly, at 8h, distinct punctate of GFP signal appeared within the stromal regions of PsbO1GFP line chloroplasts. However, when HSP90C FLAG was coexpressed, these punctate were scarcely observed ( Fig 7B). GFP punctate were also absent from PsbO1 1-58 GFP and PsbO1 1-85 GFP lines, which showed wildtype-like plastidic development overall (Fig 7B).
HSP90 and the other classes of molecular chaperones all have general chaperone activities in preventing client proteins from aggregation under stress conditions. We speculate that some GFP punctate might represent insoluble aggregates form of iPsbO1GFP at the pre-/prothylakoid stage. We therefore performed a differential centrifugation combined with high concentration of detergent for seedling samples grown 8 hours after switching to light. Indeed, iPsbO1GFP was observed in the insoluble fraction of PsbO1GFP line seedlings, with little in membrane fraction which represent membrane bound or vesicle associated PsbO1GFP ( Fig  7C). Co-expression of HSP90C caused a striking reduction of insoluble PsbO1GFP protein overall. Nevertheless, it seems that majority of PsbO1GFP is still in soluble form and little or no HSP90C is trapped with insoluble PsbO1GFP (Fig 7C).

HSP90C interacts with both PsbO1 mature protein and the thylakoid targeting peptide in vivo and is required for PsbO1GFP thylakoid targeting
We have demonstrated that HSP90C interacts with the mature form mPsbO1 using yeast twohybrid, in vitro pulldown and size exclusion chromatography (Fig 1 and S1 Fig). To confirm the interaction in plant cells and to investigate whether HSP90C interacts with the PsbO1 thylakoid targeting peptide alone, we performed co-immunoprecipitation from plant lysates expressing PsbO1GFP and PsbO1 1-85 GFP. Immunoblots showed that HSP90C and HSP70B were co-immunoprecipitated with PsbO1GFP and PsbO1 1-85 GFP, whereas no band was detected with stroma-localized GFP, derived either from RbcS 1-79 YFP or PsbO 1-58 GFP (Fig 8).
Additionally, PsbO1GFP was able to co-purify more HSP90C than PsbO1 1-85 GFP, suggesting that HSP90C binds the stromal intermediate form of PsbO1 more tightly than the thylakoid targeting peptide alone. Mass spectrometry analysis of proteins bound to PsbO1GFP confirmed the presence of both HSP90C and HSP70B, together with PsbC/CP43 and PsbS/CP22, which are characterized as a PSII reaction center protein and PSII-associated photoprotection protein, respectively (S2 Table).
To further understand the role of HSP90C in PsbO1 thylakoid targeting, we treated 5-DAG seedlings with cycloheximide to block protein synthesis in combination with geldanamycin to inhibit HSP90 family protein ATPase activity [43]. It was observed that inhibition of protein translation only caused a reduction of GFP signal as revealed by confocal microscopy (Fig 9A) and rapid degradation of both iPsbO1GFP and mPsbO1GFP by immunoblotting (Fig 9B). However, the overall distribution of GFP signal was not significantly changed (Fig 9A). Interestingly, co-application of geldanamycin significantly reduced the co-colocalization of GFP signal with chlorophyll autofluorescence (Fig 9A left bottom), suggesting active HSP90C activity is required for thylakoid tethering of PsbO1GFP, or for the active transport of PsbO1GFP into thylakoid lumen where the mPsbO1GFP is rapidly degraded (Fig 9B left). Since HSP90C interacts with PsbO1 thylakoid targeting sequence (Fig 8), we also tested geldanamycin on seedlings expressing PsbO1 1-85 GFP (Fig 9A right). Similarly, co-localization of GFP signal  with chlorophyll autofluorescence was dramatically reduced as well (Fig 9A right) and the presumably lumen localized GFP protein was degraded rapidly, while the stroma localized intermediate form was relatively stable (Fig 9B, right), thus in agreement with the observation for PsbO1GFP line and suggesting HSP90C activity is required for thylakoid tethering or transport of GFP using the bipartite PsbO1 signal peptide only.

Discussions
Chloroplasts, being the hallmark of plants, have their own intricate protein homeostasis network that shares very high similarity to its cytosolic counterpart. The sequence and domain structure of plastidic HSP90C is highly conserved [10]. Despite being an essential chaperone for higher plant development and growth, the role and mechanism of action of HSP90C still remains elusive. In this study, we attempted to investigate the Arabidopsis HSP90C interacting partners by yeast two-hybrid screen and obtained a small set of candidate interactors (S1A Fig). This small number may indicate that weak interactions have not been well identified. It should be also noted that a large portion of the proteins within chloroplasts are membraneassociated and the classical yeast two-hybrid screen does not favour the identification of membrane proteins [44]. Screening membrane proteins using the split ubiquitin system [45], or using the recently developed BioID technique [46] could be attempted in the future. Nevertheless, we identified PsbO1 as an HSP90C interactor and confirmed the interaction by a set of in vivo and in vitro biochemical assays (Figs 1 and 8). In particular, we demonstrated that HSP90C alleviates the variegated phenotype caused by accumulation of PsbO1GFP within stroma (Figs 4 and 5) and facilitates the PsbO1GFP fusion protein tethering/transport to thylakoid (Figs 6 and 9A). This study therefore not only demonstrates a novel role for HSP90C in thylakoid protein transport, but also exemplifies one critical role of HSP90C in maintaining the chloroplast stroma protein homeostasis, which involves protein import from TOC/TIC, transport into thylakoid and degradation in stroma. An integrated model was shown in Fig 9C fractions were collected after high-speed centrifugation and washed with 2% NP-40. TL, total lysates, Sol, soluble proteins, Mem, membrane bound proteins, Insol, insoluble proteins. The different fractions were loaded equivalently. Anti-HSP90C and anti-GFP were used.

Active HSP90C protein level is tightly regulated in chloroplasts
Cytosolic molecular chaperones such as HSP90 are generally very abundant under physiological conditions and their expression levels are even higher after heat shock [47]. Arabidopsis plastidic HSP90C, however, is developmentally regulated and not very responsive to heat shock [15]. Reduced HSP90C expression by transgene-induced gene silencing caused a variegation phenotype [11,17]. This suggests HSP90C level within chloroplasts is well-controlled and there is not much extra HSP90C in the reserve state. In this study, we observed a variegated phenotype for plants that expressed and accumulated high level of intermediate PsbO1GFP (Figs 2 and 3). One scenario could be that PsbO1GFP intermediate forms retained in stroma bind an excess of HSP90C and limit the availability of HSP90C for the other processes, thus mimicking the transgene-induced HSP90C co-suppression plants [11]. However, the variegated phenotype shown in PsbO1GFP plants is different from HSP90 co-suppression lines as the former gradually recovers over time (Figs 2C and S6), while the latter remains variegated. This is reasonable because the steady state level of intermediate PsbO1GFP diminishes over time (Fig 4C). Consequentially, the need for HSP90C is also decreasing and the chaperone can be released from iPsbO1GFP to participate in other processes as the plant continues to develop. In agreement with this hypothesis, we noticed high accumulation of HSP70B in chloroplasts as well ( Fig 3A) and both HSP90C and HSP70B are associated with PsbO1GFP (Fig 8,  S2 Table). This clearly indicates that iPsbO1GFP is associated with the HSP90C chaperone complex within the stroma and, non-physiological presence of PsbO1GFP significantly impairs the balance of chaperone network (Fig 9C).
Nevertheless, it should be noted that variegation in green leaves is normally a complicated phenotype and any factor affecting the normal chloroplast biogenesis and maturation may trigger the variegated phenotype [48]. It is technically difficult to identify the cause that primarily triggers the variegation (i.e. by HSP90C cosuppression or by PsbO1GFP expression). Similar variegated phenotypes have been observed for plants with defective function of the other chaperone family members. For example, Arabidopsis T-DNA insertion mutants affecting DnaJ-like proteins wco [49], sco1 [50] and cyo1 [51] also display albino cotyledons as we observed for PsbO1GFP expressing seedlings. Additionally, to understand the molecular mechanism of the variegated phenotype induced by PsbO1GFP expression, re-introducing the different fusion constructs into Arabidopsis with endogenous PsbO knocked out as shown in S3 Fig and analyzing the distribution of the fusion proteins without one of or both the endogenous PsbO proteins may be performed in the future, so as to study the sole effect of exogenous PsbO1 fusion proteins.

Vital role of HSP90C in thylakoid biogenesis
During embryogenesis, cotyledon cells develop functional green chloroplasts which are later converted to etioplasts at the end of seed maturation [41,42]. When seeds germinate, etioplasts transported into the thylakoid lumen through the SEC translocon, and integrated into the PSII complex or subjected to degradation immediately. The formation of PsbO1-HSP90C complex in stroma promotes the import, thylakoid transport and likely degradation. Exogenous expression of HSP90C to higher level within stroma promotes the PsbO1-HSP90C complex formation, while the complex formation negatively regulates the availability of HSP90C for other client proteins. The dashed arrows indicate the interconversion of different forms of PsbO1 and the solid arrows indicate the positive role of HSP90C in the processes.
https://doi.org/10.1371/journal.pone.0190168.g009 in cotyledons rapidly differentiate into chloroplasts upon light exposure, while in emerging vegetative leaves chloroplasts are differentiated primarily from proplastids, though mature chloroplasts can also undergo division in certain cells [52]. Thylakoids are characteristic of photosynthetic chloroplasts and therefore chloroplast biogenesis and maturation is marked by the de novo thylakoid membrane formation. However, even with extensive studies [53,54], the mechanism that underlines the thylakoid membrane biogenesis is far from being clearly understood. VIPP1 is a plastid vesicle-inducing protein and required for thylakoid biogenesis [55]. VIPP1 is also required for the thylakoid membrane insertion of PSII complex [56][57][58]. Interestingly, HSP90C also interacts with VIPP1 [17]. Electron microscopy analysis indicated that the thylakoid membrane is significantly degenerated in both HSP90C co-suppression cells [11] and the HSP90C point mutation, cr88, cells [59]. Our yeast two-hybrid analysis also identified that HSP90C interacts with LHCB2, the light harvesting complex protein (S1A Fig). All these studies suggest that HSP90C is a key component in directing thylakoid membrane biogenesis.
Additionally, we examined the photomorphogenesis of seedlings after 3 days of growth in the dark. In agreement with the variegated phenotype observed for seedlings grown under regular light cycles, seedlings expressing PsbO1GFP displayed a significant delay in the greening process that was marked by delayed appearance of chlorophyll (Fig 8). In particular, there are large amounts of green fluorescent clusters in the chloroplasts of cells that express PsbO1GFP only. Co-expression of extra HSP90C in these cells reduces the clusters and enables chloroplast differentiation to proceed as observed in wild type seedlings. It is likely that PsbO1GFP, either in soluble forms or in aggregated clusters, binds and sequesters too much HSP90C, thus causing an imbalance of the chloroplast chaperone homeostasis (Fig 9C).

Molecular chaperones are required for PsbO1 enroute to the thylakoid lumen
PsbO1 is a thylakoid lumen-localized extrinsic subunit protein of PSII and plays a crucial role in the oxygen-evolving complex by stabilizing the catalytic manganese cluster [60]. The soluble state of PsbO1 has been characterized as a flexible molten globule [61]. It is generally believed that PsbO1 does not require molecular chaperones for folding [62]; rather it folds into native state upon assembly into PSII complex [63][64][65]. However, we cannot exclude the possibility that PsbO1 does require chaperones during its long journey en route to the thylakoid lumen after synthesis in the cytosol. In this study, we provided multiple lines of evidence showing the direct interaction between PsbO1 and the HSP90C foldosome (Figs 1 and 8) and analyzed the role of HSP90C in reducing the PsbO1GFP toxicity (Figs 4, 5 and 7). This clearly indicates that HSP90C may act as a vital regulator for PsbO1 in stroma. Since PsbO1 is not well-folded in solution and is prone to protease attack, it would be plausible to assume the HSP90C foldsome protects PsbO1 from degradation in the stroma and prevents it from aggregating, as our pilot assay revealed an accumulation of insoluble PsbO1GFP proteins at 8 hours after switching to photomorphogenesis (Fig 7). However, we cannot rule out the possibility that association with HSP90C might also facilitate protein degradation, at least for PsbO1GFP, since the steady state level of PsbO1GFP diminishes much more rapidly in the presence of extra HSP90C (Fig 4C). The dual role for cytosolic HSP90 in determining the fate of substrates has been well studied with CHIP E3 ligase [66,67]. In particular, HSP90C has been previously shown to interact with HSP93, an unfoldase of the chloroplast clpP protease system, in the co-immunoprecipitation experiment [16]. However, we did not identify HSP93 when analyzing the PsbO1GFP complex by tandem mass spectrometry (S2 Table). PsbO1GFP homeostasis and the kinetics of PsbO1GFP degradation within stroma needs to be further investigated.
We initially identified the interaction between HSP90C and a PsbO1 T200A mutant ( Fig  1A). It should be noted that T200 is correspondent to T115 with regards to mature PsbO1. This threonine is highly conserved in all PsbO isoforms and we speculate that the point mutation might have arisen during the cDNA library construction and does not exist in vivo. The switch from polar amino acid to non-polar alanine seems to slightly increase the strength of interaction with HSP90C (Fig 1D). This is in agreement with the general role of molecular chaperones in binding hydrophobic batches of substrates during folding. It is likely that PsbO1 vigorously interacts with and needs protection from molecular chaperones before it is transported into thylakoid lumen; however, the interaction might be too weak and transient to be visualized by classical biochemical assays. The involvement of stroma chaperonin for thylakoid transport of plastidic type I signal peptidase has been recently reported [68]. By comparing two transgenic lines that bear the same PsbO1GFP allele, we showed that extra HSP90C facilitates the accumulation of PsbO1GFP mature form ( Fig 5B) and more thylakoid association of GFP signal (Fig 6). Additionally, HSP90C also interacts with the PsbO1 thylakoid targeting peptide (Fig 8). PsbO1 is a native SEC translocon substrate [69]. It would be interesting to investigate in the future whether HSP90C interacts with the SEC complex in thylakoid membrane thus playing a direct role in SEC protein transport.
In conclusion, through yeast two-hybrid assay, we identified photosystem II component PsbO1 as an HSP90C interacting partner and showed evidence that proper homeostasis of HSP90C is critical for chloroplast differentiation and maturation and particularly required for thylakoid transport of PsbO1. We analyzed the growth and development of plants overexpressing PsbO1GFP that has been widely used as reporter protein for chloroplast protein import/ transport research, albeit limited study in planta particularly at the late development stage.
Our study therefore provides a new view of the complicated protein homeostasis network within the unique organelle in plants.

Plant materials and growth conditions
The Arabidopsis thaliana ecotype Columbia (Col-0) was used as wild type. To grow seedlings in vitro, seeds were surface sterilized and sown on ½ strength Murashige and Skoog (MS) medium containing 1% sucrose and 0.7% agar with or without supplementation by specific antibiotics. After stratification in the dark at 4˚C for 3-4 days, seeds were cultured within a plant growth incubator set at 120 μmol.m -2 .sec -1 , 16/8 hour light/dark cycle at 22˚C. To grow plants to late stage development, seedlings were transferred to soil 10 days after germination for growth within a plant growth chamber set at 110 μmol.m -2 .sec -1 , 16/8 hour light/dark cycle at 22˚C. RbcS 1-79 GFP, RbcS 1-79 YFP were generated previously [70] and requested from the ABRC.

Yeast two-hybrid screening
The HSP90C middle and C-terminal domains (aa.320-780), designated as HSP90C-MC was cloned into pEG202 vector for yeast two-hybrid screening as the middle and C-terminal domains bind client proteins [25]. The Arabidopsis cDNA library constructed into prey vector pJG4-5 was a kind gift of Dr. Gazzarrini (University of Toronto, Canada), and the screening of positive interaction using LEU2 and lacZ marker genes was performed in S. cerevisiae EGY48 cells as previously described [27]. Positive hits were sequenced and retransformed back to yeast cells to test interactions with HSP90C-MC and the N-terminal domain HSP90C-N (aa.61-319).

Construction of PsbO1GFP fusion genes for expression in E. coli and plants
The originally identified PsbO1 gene in pJG4-5 prey vector lacks the coding sequence for the first 22 amino acids and bears a point mutation resulting in Alanine 200 instead of Threonine 200 , designated as T200A. Site directed mutagenesis was performed to correct the point mutation. An oligo encoding the 22 amino acids was synthesized and used as a primer to construct the full length native PsbO1 gene, which was later cloned into both pJG4-5 and E. coli expression vector pProEXHTa using EcoRI and XhoI sites. To construct the preprotein, intermediate or mature forms of PsbO1GFP (Fig 1), the coding sequences were amplified by PCR and cloned into the AgeI site of pEGAD vector [71], generating in-frame fusion proteins. A partial digestion for the pre-PsbO1 protein gene was applied when constructing the binary vector. To make chloroplast stroma, or thylakoid targeted GFP constructs using the PsbO1 transit peptide, the coding region for either PsbO1 1-58 or PsbO1 1-85 was cloned into the pEGAD AgeI site, creating translational fusion proteins without changing the chloroplast transit peptide or thylakoid targeting peptide cleavage site.

Arabidopsis transformation and screening of transgenic plants
A. tumefaciens GV3101 carrying binary plasmid was used to transform Arabidopsis Col-0 by floral dip [72]. Selection of transgenic plants was performed on ½ strength MS medium with 1% sucrose supplemented with 25 μg/mL of glufosinate (Crescent Chemicals), or 25μg/ml kanamycin when needed, and then confirmed by either PCR genotyping or immunoblotting with specific antibodies.

Protein expression and purification in E. coli
The constructs for His 6 -tagged mature form HSP90C (aa. 61-780), mPsbO1 and mPsbO1 T200A were made in pProEXhtb plasmid. p/i/mPsbO1GFP fusion protein constructs were made by cloning the corresponding coding regions from pEGAD into pET22b vector in NdeI site. Constructs were then introduced into E. coli BL21 (DE3)-pRIL (Stratagene) and protein expression was induced by 1 mM IPTG. His 6 -tagged proteins were purified using Ni-NTA resin (QIA-GEN), and dialyzed overnight. His 6 -tag was cleaved with tobacco etch virus (TEV) protease, removed by Ni-NTA resin and further purified by size exclusion chromatography with a Superdex 200 column (GE Healthcare) in buffer containing 25 mM Tris-HCl, pH 7.5, 150 mM KCl, 10% glycerol, and 0.5 mM DTT.

Chloroplast isolation and fractionation
The protocol for chloroplast isolation from Arabidopsis leaves was adapted from [73]. Briefly, approximately 4 g of 2-week-old Arabidopsis seedlings were collected and crushed in 1x homogenization buffer (330 mM sorbitol, 50 mM HEPES-KOH, pH 7.5, 5 mM EDTA, 5 mM EGTA, 1 mM MgCl 2 , 10 mM NaHCO 3 , and 0.5 mM DTT), and filtered through two layers of Miracloth (Calbiochem, CAT# 475855). Intact chloroplasts were collected from the interface after a 30% and 80% percoll gradient centrifugation at 2,000 × g for 10 minutes. Chloroplasts were washed twice with homogenization buffer and saved for total chloroplasts, or further suspended in osmotic lysis HM buffer (20 mM HEPES-KOH pH 7.5, 10 mM NaHCO 3 , 2 mM MgCl 2 , 2.5 mM EDTA, 2.5 mM EGTA, 1 tablet of Roche protease inhibitor in 20 ml) on ice for 15 min. Supernatant (stroma) and pellet (thylakoid) fractions were separated with centrifugation at 2,600 xg for 5min. Pellet fraction was washed three times in HM buffer. Stroma proteins were concentrated with acetone precipitation at 80% final concentration and centrifugation at 10,000xg for 30min. Both stroma and thylakoid proteins were solubilized in 2xSDS buffer with 5 M urea and then subjected to immunoblotting analysis. In experiments that required normalization or estimation of the total chlorophyll a/b contents, a formula described in [74] was used after measuring the absorbance at 663 and 646 nm in acetone.

Immunoprecipitation
To purify GFP or GFP fusion protein complexes from seedlings, seven-day-old seedlings were harvested and ground in liquid N 2 and then incubated at 4˚C for 1h with lysis buffer 25 mM Tris-HCl, pH7.5, 150 mM NaCl, 0.5% NP40, containing protease inhibitor complex (Roche). Co-immunoprecipitation was performed using 25μl GFP-trap resin (ChromoTek) and incubated with cell lysate at 4˚C for 1h. The beads were recovered by centrifugation at 1,000xg for 2 min and washed 4 times with 500μl wash buffer 25 mM Tris-HCl, pH7.5, 150 mM NaCl, 0.1% NP40. Bound proteins were released by incubation with 0.1 M glycine, pH2.5 or analyzed using 2xSDS buffer. Eluents were analyzed by SDS-PAGE and immunoblotting or by LC-MS/ MS analysis (SPARC BioCentre, Sick Kids, Toronto).

Fractionation of soluble and insoluble proteins
Possible insoluble protein aggregates were isolated from seedling materials as described [75] with slight modification. Briefly, the seedlings were ground by micro pestle in grinding buffer containing 50 mM MES-NaOH, pH 6.5, 1% glycerol, 0.1% NP40, and then centrifuged at 2,000xg for 5 min followed by 3000xg for 5 min. The supernatant was then centrifuged at 20,000g for 60 min and the supernatant was saved as soluble fraction. The pellet was suspended and washed twice in grinding buffer but with 2% NP40 and supernatant from the first wash was saved as membrane fraction. The final pellet after 20,000 xg for 60 min centrifugation was saved as insoluble fraction.

Antibodies
Polyclonal rabbit anti-HSP90C antibody was described in [11]. Anti-PsbO1 antibody was generated by Signalway AntiBody (College Park, USA) with purified PsbO1 protein. Other primary antibodies used in this study include anti-FLAG monoclonal antibody (Sigma, F3165) and anti-chloroplast HSP70B (Agrisera). GFP-TRAP resins were purchased from ChromoTek respectively.