ARTICLE https://doi.org/10.1038/s41467-020-15184-1 OPEN How paired PSII–LHCII supercomplexes mediate the stacking of plant thylakoid membranes unveiled by structural mass-spectrometry 1,4 2,3,4 1 2,3 Pascal Albanese , Sem Tamara , Guido Saracco , Richard A. Scheltema & Cristina Pagliano Grana are a characteristic feature of higher plants’ thylakoid membranes, consisting of stacks of appressed membranes enriched in Photosystem II (PSII) and associated light-harvesting complex II (LHCII) proteins, together forming the PSII-LHCII supercomplex. Grana stacks undergo light-dependent structural changes, mainly by reorganizing the supramolecular structure of PSII-LHCII supercomplexes. LHCII is vital for grana formation, in which also PSII- LHCII supercomplexes are involved. By combining top-down and crosslinking mass spec- trometry we uncover the spatial organization of paired PSII-LHCII supercomplexes within thylakoid membranes. The resulting model highlights a basic molecular mechanism whereby plants maintain grana stacking at changing light conditions. This mechanism relies on interactions between stroma-exposed N-terminal loops of LHCII trimers and Lhcb4 subunits facing each other in adjacent membranes. The combination of light-dependent LHCII N- terminal trimming and extensive N-terminal α-acetylation likely affects interactions between pairs of PSII-LHCII supercomplexes across the stromal gap, ultimately mediating membrane folding in grana stacks. 1 2 Applied Science and Technology Department–BioSolar Lab, Politecnico di Torino, Environment Park, Via Livorno 60, 10144 Torino, Italy. Biomolecular Mass Spectrometry and Proteomics, Bijvoet Center for Biomolecular Research and Utrecht Institute for Pharmaceutical Sciences, University of Utrecht, Padualaan 8, 3584 CH Utrecht, The Netherlands. Netherlands Proteomics Centre, Padualaan 8, 3584 CH Utrecht, The Netherlands. These authors contributed equally: Pascal Albanese, Sem Tamara. email: firstname.lastname@example.org; email@example.com NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications 1 1234567890():,; ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 iological processes are fundamentally driven by an intricate evidences suggest that: (1) Lhcb3 is present in one copy per M- network of interacting macromolecular protein complexes trimer, likely with mostly two Lhcb1 subunits as partners; (2) Bwith large and dynamic structures. These properties make it Lhcb2 is likely a component only of the S-trimer. As a result, so challenging to resolve structures at atomic detail with classical far little is known about the exact composition and localization of structural biology approaches, such as crystallography and cryo- Lhcb1 and Lhcb2 mostly within the S-trimers present in the electron microscopy (cryo-EM), especially when the complex is PSII–LHCIIsc. buried in a membrane . A multi-subunit pigment–protein com- The relative ﬂat stromal surface of PSII–LHCIIsc allows it to be plex embedded in the thylakoid membranes of all oxygenic accommodated in the tightly stacked region of the thylakoid photosynthetic organisms is Photosystem II (PSII). This is an membranes called grana, where the distance between neighbour- 10,23 enzyme crucial for life on Earth as, over the past three billion ing membranes is within 2–3.6 nm . In plants, the dynamic years, powered by solar energy, it has catalysed the oxidation of control of grana stacking is crucial for photosynthetic adaptation 24,25 water, thus creating the oxygenic atmosphere that sustains all to light cues and, under variable irradiances, depends on the 2 24 aerobic forms of life . The core complex of PSII is composed of a reversible macro-reorganization of PSII–LHCIIsc . Besides its large number of intrinsic subunits and few extrinsic polypeptides seminal importance, stacking of grana is a topic not yet fully (i.e. PsbO, PsbP, PsbQ and PsbR). Among the membrane pro- understood . It is thought to be mainly driven by adhesion of teins there are the large reaction centre D1 and D2 subunits and LHCII trimers in adjacent membranes , driven by non-speciﬁc inner antenna proteins CP43 and CP47, which are accompanied electrostatic interactions of Lhcb stroma-exposed N-terminal 20,27 by several subunits with low molecular mass (<10 kDa, e.g. PsbF, loops . Experimental evidences also suggest the involvement of PsbK, PsbH, PsbT, etc.), accounting for more than half of the the PSII–LHCIIsc in the stacking of grana membranes. Indeed, entire complex. The structural organization of the PSII catalytic contacts between PSII–LHCIIsc located in adjacent thylakoid core has been fundamentally conserved throughout the evolution membranes, mediated by the stromal surfaces of both LHCIIs and of photosynthetic organisms from cyanobacteria to higher PSII cores, have been detected in vivo . In addition, unidentiﬁed plants , and its molecular organization has been detailed in physical connections were observed in the cryo-EM map at 14 Å 4,5 previous studies . The plant PSII core is serviced by peripheral resolution of paired (C S M)×2 supercomplexes interacting on 2 2 antenna proteins forming the light-harvesting complex II their stromal side, which were tentatively assigned as mutual (LHCII). This antenna system has been prone to evolutionary interactions of two long N-terminal loops of Lhcb4 spanning the diversiﬁcation, producing a wide range of species-speciﬁc LHCII stromal gap . So far the structural determination at high- isoforms encoded by multiple genes while maintaining a strictly resolution of these stromal protein–protein interactions by clas- 6,7 conserved fold and structural organization . Different types and sical structural methods has suffered from multiple limitations: the numbers of LHCII proteins bind to the PSII core to form dynamic nature of the large paired PSII–LHCIIsc assembly (over PSII–LHCII supercomplexes (PSII–LHCIIsc), whose dynamic 2 MDa), the heterogeneity of the LHCII subunits, and the high remodelling allows plants to adapt to ever-changing environ- ﬂexibility of their stroma-exposed N-terminal loops. In addition, 8,9 mental light conditions . N-terminal processing and post-translational modiﬁcations Plant PSII–LHCIIsc are composed of a PSII dimeric core (C ) (PTMs) such as phosphorylation and acetylation, either on lysine with two strongly bound (S ) LHCII trimers, which are hetero- residues or on free termini, occur in the majority of LHCII N- 29,30 31 trimers formed by the Lhcb1 and Lhcb2 subunits, and up to two terminal loops . For instance, reversible phosphorylation and additional moderately bound (M ) LHCII trimers, containing lysine acetylation on Lhcb2 N-terminal loops are central for 1–2 10,11 also the Lhcb3 protein . The binding of LHCII trimers to the functional LHCII redistribution during state transitions from PSII, PSII core relies on three monomeric LHCII subunits, Lhcb4, located in grana stacks, to Photosystem I (PSI), conﬁned in single- Lhcb5 and Lhcb6. Lhcb5 acts as the linker exclusively for the S- layered thylakoid domains (i.e. stroma lamellae). Conversely, trimer, Lhcb6 exclusively for the M-trimer and Lhcb4 connects to permanent N-terminal α-acetylation is known to stabilize proteins 10 33 both trimers . The predominantly occurring PSII–LHCIIsc are and mediate protein–protein interactions , potentially further of type C S M ,C S M and C S , whose relative abundances in impacting stromal interactions of paired PSII–LHCIIsc and con- 2 2 2 2 2 2 2 9,13 the thylakoid membranes depend on the light intensity . Due sequently thylakoid macro-organization. to the intrinsically dynamic arrangement of the outer antenna Considering the complexity of PSII–LHCIIsc, both in terms of system it was only recently possible, with the emergence of single light-driven structural dynamics and heterogeneous composition particle cryo-EM , to resolve the structure of plant PSII–LHCIIsc of LHCII, we combined top-down mass spectrometry (TD-MS) 15–17 at near-atomic detail . From these high-resolution structures, and crosslinking mass spectrometry (XL-MS) to resolve so far Lhcb3 was clearly assigned and exclusively localized within the hidden structural details of paired PSII–LHCIIsc. In this study, we M-trimer, as the monomer in contact with the Lhcb4 and used paired supercomplexes isolated from stacked thylakoid Lhcb6 subunits . However, not everything could be resolved, as membranes of pea plants grown at three light intensities, ranging e.g. Lhcb2 could not be differentiated from Lhcb1. Also the from limiting to excessive light. This set allows detection of the atomic structures available for isolated LHCII trimers do not stable structural features common to the different conditions and 18–20 allow discrimination between these two Lhcb proteins ,as potentially involved in maintaining a basic degree of grana they show high sequence similarity and differentiate mostly at stacking. TD-MS is a method capable of identifying intact proteins 21 34 their N-terminus , a feature which is missing in the available of up to ~100 kDa with high throughput and characterizing high-resolution structures. Their discrimination is difﬁcult even proteins below ~30 kDa by uncovering mature protein sequences, by biochemical methods, although a recent mass spectrometry thus potentially disclosing unknown variants, and PTMs . This study performed on preparations of LHCII trimers with different approach is particularly suitable for this study as the genome of P. conﬁgurations revealed that the M-trimer is enriched in Lhcb1, sativum (pea) is not fully sequenced. For the XL-MS data analysis, while Lhcb2 is almost absent in this trimer compared with S- intimate knowledge of the protein sequences is required. Con- trimers . This result is in accordance with evidence obtained sidering the lack of structural features for the ﬂexible stroma- with biochemical studies on isolated PSII–LHCIIsc with different exposed portions of the PSII–LHCIIsc in the high-resolution organizations supporting that Lhcb2 is a speciﬁc component of structure available , and the high homology sequence displayed the S-trimer, whereas Lhcb1 is present in both S- and M- by the LHCII proteins , the preliminary TD-MS analysis was trimers . Taken together, all these structural and biochemical instrumental for further XL-MS dataset mining and integration. 2 NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 ARTICLE XL-MS, which uses small chemical crosslinkers, has demonstrated mutual interactions between N-terminal loops either of Lhcb1 considerable potential in gaining structural insights at inter- or Lhcb4.2 subunits of adjacent supercomplexes. The two 36–39 mediate resolution on large protein assemblies and even in Lhcb4.2 subunits were found to be tightly interacting in a position 40–42 complex matrixes on a proteome-wide scale . This approach is close to the stromal connecting density deﬁned as the “knot” in particularly suited for this study as the structure of paired the (C S M)×2 cryo-EM map , ﬁnally providing an identity to 2 2 PSII–LHCIIsc from plants is available only at intermediate reso- this density. Most interactions detected in vitro on the isolated lution . We applied TD-MS to proﬁle LHCII isoforms and their paired supercomplexes were furthermore supported by XL-MS proteoforms (i.e. different forms of a protein arising from a given results obtained in situ on the corresponding thylakoid mem- gene with a variety of sequence variants and PTMs). The struc- branes. These ﬁndings represent the ﬁrst biochemical evidence tures of the most abundant proteoforms, complete with N- that mutual interactions between either LHCII trimers or terminal stroma-exposed regions, were modelled and ﬁtted into Lhcb4.2 subunits occur between PSII–LHCIIsc facing each other the cryo-EM density map of the (C S M)×2 supercomplex , in adjacent thylakoid membranes, suggesting their direct invol- 2 2 which represents the most abundant PSII–LHCIIsc common to all vement in mediating grana stacking. three light conditions . To investigate their structural interactions, we treated paired PSII–LHCIIsc isolated from the three light conditions with two complementary chemical crosslinkers, tar- Results geting different residues and producing complementary sets of The structure of paired PSII–LHCIIsc. As starting material, we 43,44 distance restraints . Detected crosslinks were used to uncover a used paired PSII–LHCIIsc puriﬁed from stacked thylakoid tight and speciﬁc network of N-terminal loops interacting within membranes isolated from plants grown at three different light the stromal gap and to localize Lhcb2 within the S-trimer. We intensities (low, L; moderate used as control, C; and high, H) deﬁned speciﬁc sites of interaction between the N-terminal (Fig. 1). Depending on the light intensity, these samples were loops of either Lhcb1 or Lhcb2 with PSII core proteins (D1 and enriched in different types of PSII–LHCIIsc, among which the CP43 or D1 and PsbH, respectively), potentially acting as hubs to (C S M)×2 was the most abundant and common to all three light 2 2 control PSII–LHCIIsc structural dynamics. In addition, we found conditions (Fig. 1). To peek through the keyhole of the tight Pea plants Isolated paired PSII-LHCIIsc C S 2 2 M-trimer (Lhcb1/3) Lhcb5 C Lhcb6 Lhcb4 30 150 750 –2 –1 μmol photons m s PSII dimeric 90° C S M core (C ) 2 2 Isolation of Lhcb5 stacked Lhcb4 thylakoids C S M 2 2 2 S-trimer (Lhcb1/2) Cross-linking MS in vitro & in situ Top-down MS Lys Lys Database search LC-MS/MS a(Y6) ~11 Å MS1 27+ β(Y2) 20+ a(B3) +β a(Y8) β(Y3) β(Y6) + a m/z FK D L G E E H F K 5+ HCD MS2 Lys Asp 7+ 11+ LS Q K F P K Lys Glu Database from Protein top-down MS m/z identification Modification positioning C 2 N 1 Integrative modelling Unreported modifications Sequence adjustment Structural predictions Proteoform Truncated C-terminus N-term characterization N-term Proteoform Mature Ac profiling proteoforms 90° n12 n8 n3 Cryo-EM fitting Fig. 1 Workﬂow of the integrated MS-based approach for characterizing the light-driven modulation of paired PSII–LHCIIsc. Isolation of heterogeneous mixtures of PSII–LHCIIsc from pea plants grown at different light intensities (L, low; C, moderate used as control; H, high). PSII–LHCIIsc preparations were used for TD-MS and XL-MS in vitro, performed with DSSO and EDC crosslinkers on two independently isolated PSII–LHCIIsc for each light condition pooled together. DSSO XL-MS in situ was conducted on thylakoid membranes isolated from three independent batches of pea plants grown in moderate (C) light. NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications 3 # Mods ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 stromal gap between two facing supercomplexes, we combined N-terminus (Ser3 and Ala4), while harbouring four single amino in-depth TD-MS proﬁling of intact proteins and detection of acid substitutions (Supplementary Data 1). Among the remaining protein–protein interactions by XL-MS (Fig. 1). The latter was identiﬁed sequence variants, we found C-terminally truncated performed either in vitro on isolated PSII–LHCIIsc or in situ on forms, two of Lhcb1, cleaved at Asp153 and Asp162, and one of the starting stacked thylakoid membranes, representing a close- Lhcb2, cleaved at Asp149 (Fig. 2d, e), showing a light-dependent to-native environment. accumulation at increasing light intensities (Supplementary Data 1). Overall, the majority of LHCII proteoforms, with exception of Lhcb2, Lhcb3 and Lhcb4.3, were detected as both complete and The heterogeneity of LHCII unveiled by TD-MS. Given the truncated forms, primarily with clipping of a few amino acids from large degree of heterogeneity expected for individual LHCII the N-terminus (Fig. 2d, e). All the N-terminally truncated forms subunits, we applied TD-MS to identify proteoforms and (with exception of Lhcb6) lacked acetylation on the N-terminal estimate relative abundances. For identiﬁcation we used the domain, which, conversely, was observed for the majority of the medium-/high-resolution workﬂow previously described (see complete forms of Lhcb1, Lhcb2, Lhcb4.2, Lhcb4.3 and Lhcb5 “Methods” section for details). Of the total detected intensity, (Supplementary Fig. 3) and the PSII core proteins D1, D2, CP43, approximately 85% could be attributed to identiﬁed mass fea- PsbF and PsbT (Supplementary Data 1). Among the complete tures in any light condition (both in medium-resolution and forms of the LHCII proteins, only Lhcb3 and Lhcb6 were primarily high-resolution MS1; this does not indicate relative abundance) observed in non-acetylated state (i.e. unmodiﬁed, Supplementary (Supplementary Data 1). A total of 35 proteins were identiﬁed, Fig. 3). with 90 different proteoforms, including all the major subunits of The light-dependent modulation of the PSII–LHCIIsc archi- the PSII–LHCIIsc. The large subunits CP47 and CP43 and the tecture is mainly driven by variations in the amount of the M- small subunits PsbH, PsbK, PsbF and PsbT (also referred to as trimer subunit Lhcb3, with its speciﬁc linker Lhcb6, and of PsbTn ) of the PSII core were detected uniquely in the medium- Lhcb4.3 (Fig. 2c). Interestingly, Lhcb4.3 compared to Lhcb4.2 resolution and high-resolution measurement, respectively (Sup- lacks the ~10 amino acids at the C-terminus essential for binding plementary Fig. 1 and Supplementary Data 1). Less than 10% of the M-trimer . Indeed, the approximate 50% reduction of Lhcb3 the total intensity for all samples was assigned as contaminants, and Lhcb6 in conjunction with an over ten-fold increase of mainly PSI/LHCI subunits and related proteins (Supplementary Lhcb4.3 observed in H compared to L, indicate the detachment of Fig. 1 and Supplementary Data 1), likely arising from cross- M-trimers in high light leading to PSII–LHCIIsc with reduced 9,45 9 contamination during sample puriﬁcation . Among the antennae . Notably, Lhcb2 showed the same light-dependent detected subunits we accounted for the PSII extrinsic PsbR accumulation trend as Lhcb1 at nearly one-third of the protein, whose positioning is still debated (i.e. it is absent from abundance (Fig. 2c). This observation conﬁrms that Lhcb2 is 15–17 12 high-resolution structures of PSII–LHCIIsc ), as well as not part of the M-trimer , as otherwise in H light this subunit potentially new components of the PSII–LHCIIsc as Psb27 and would decrease in the same fashion as Lhcb3 does. It is also likely TL18.3 (here referred to as “PSII-related”) (Supplementary Fig. 1 that Lhcb2 is present in one copy per S-trimer. We base this on and Supplementary Data 1). Even though their accumulation our experimental setup, where the samples were loaded based on appears to be light-dependent (Supplementary Data 1), in the same chlorophyll content and the amount of PSII core agreement with previous ﬁndings for PsbR , these soluble proteins and Lhcb1 and Lhcb2 is rather constant in the three light proteins are either transiently bound to the PSII in speciﬁc light conditions . In fact, if more than one copy of Lhcb2 were present 46 47 conditions (i.e. Psb27 and TL18.3 ) or partially lost during per S-trimer, the Lhcb2:Lhcb1 ratio would increase in H, since in 48 9 PSII–LHCIIsc puriﬁcation (i.e. PsbR protein ). Therefore, from this light condition most of LHCII trimers are of S-type , and in a structural perspective, the study of these proteins requires a this case we would expect to see, based on MS intensity, a Lhcb2: dedicated sample preparation to conﬁdently localize their posi- Lhcb1 ratio of 2:1 instead of the experimentally observed tion in the PSII–LHCIIsc architecture. 1:2 shown in Fig. 2c. Even though the abundance comparison By considering the set of identiﬁed PSII–LHCIIsc mass features provides a rough estimate, it supports the 1:2 stoichiometry as a ﬁngerprint for each of the light conditions, it was evident that between Lhcb2 and Lhcb1, which was further conﬁrmed by the major light-dependent variability in proteoform composition absolute quantiﬁcation at peptide level (Supplementary Fig. 4 and occurred within LHCII (Fig. 2a), with distinct patterns emerging Supplementary Note 1). From the TD-MS we were additionally when the corresponding monoisotopic masses were plotted against able to determine most of the remaining PSII–LHCIIsc mature their retention times (Fig. 2b). For each of the LHCII trimer sequences (Fig. 2b, c), whose stroma-exposed N-terminal loops building blocks, Lhcb1, Lhcb2 and Lhcb3, as well as the monomeric are largely missing in the currently available high-resolution 15–17 Lhcb4 (i.e. Lhcb4.2 and Lhcb4.3), the complete primary isoform structures (Fig. 2e). These portions are the most diversiﬁed was dominant in all light conditions, with only marginal amounts of (Supplementary Fig. 2b), thus the knowledge of their sequence is truncated forms detected (Fig. 2c, d). Conversely, for monomeric determinant for the discrimination of the different LHCII Lhcb5 and Lhcb6 the truncated primary isoforms were highly isoforms. As these domains are thought to play a key role in 23,28 abundant in all light conditions, representing up to 50% of all the PSII–LHCIIsc structural and functional pairing , we contin- detected proteoforms in the L sample (Fig. 2c, d and Supplementary ued our investigation with an integrative structural biology Data 1). Lhcb1 emerged as the most abundant LHCII component, approach. with intensity levels at least two-fold higher than Lhcb2 and Lhcb3 in any light condition (Fig. 2c, d). Such a quantitative assessment is likely correct due to the extremely high sequence homology Interactions between paired PSII–LHCIIsc captured by XL-MS. between these proteins (Supplementary Fig. 2) and structural biases To uncover details on how the PSII–LHCIIsc structurally interact in the detected intensities are therefore not expected. Lhcb1 was across the stromal gap, we applied XL-MS using two com- present as three distinct isoforms, named according to their plementary crosslinking reagents (disuccinimidyl sulfoxide, DSSO; transcriptome entries 0081729, 0074459,and 0050874 (Supple- and 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide, EDC) to mentary Data 2). Lhcb1_0081729 accounted for over 80% of all paired PSII–LHCIIsc preparations representative of L, C and H Lhcb1 in any light condition, while the other two isoforms were plants. These preparations are heterogeneous mixtures of the three less abundant. These two isoforms lacked two amino acids at the main types of paired PSII–LHCIIsc (C S )×2, (C S M)×2 and 2 2 2 2 4 NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications lhcb1 lhcb2 lhcb3 lhcb42 lhcb43 lhcb5 lhcb6 0081729 0062077 0083085 0076852 0068262 0087274 0079196 lhcb1 lhcb2 lhcb3 lhcb42 lhcb43 lhcb5 lhcb6 0081729 0062077 0083085 0076852 0068262 0087274 0079196 lhcb1 lhcb2 lhcb3 lhcb42 lhcb43 lhcb5 lhcb6 NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 ARTICLE a b L C H PSII-LHCII LHCII PSII Corr H 0.73 0.89 1 H 0.4 0.57 1 H 0.91 0.92 1 1.0 lhcb1 0.5 lhcb2 C 0.76 1 0.89 C 0.42 1 0.57 C 0.92 1 0.92 lhcb3 0.0 lhcb42 −0.5 lhcb43 L 1 0.76 0.73 L 1 0.42 0.4 L 1 0.92 0.91 −1.0 lhcb5 lhcb6 L C HL C H L C H Retention time [min] Complete primary isoform Truncated primary isoform lhcb1_0074459 lhcb1_0050874 cd e PDB 5xnl:1/2/G/N/Y/g/n/y 15.16 13.00 n55 n5 n4 c162 7.58 6.50 c153 0.00 0.00 4.71 4.68 c149 2.36 2.34 0.00 0.00 PDB 5xnl:3/7 6.17 6.17 3.09 3.09 0.00 0.00 PDB 5xnl:R/r 2.52 2.52 n12 1.26 1.26 n4 0.00 0.00 0.30 0.30 0.15 0.15 0.00 0.00 PDB 5xnl:S/s 3.45 2.52 n60 n12 1.72 1.26 n8 n3 0.00 0.00 PDB 5xnl:4/8 2.54 1.43 n2 0.72 1.27 0.00 0.00 N-term C-term 1 45 90 135 180 225 LC H LC H Sequence position Fig. 2 TD-MS proﬁling of LHCII diversity. a Spearman correlation of mass features detected in the PSII–LHCIIsc L, C and H samples for PSII–LHCIIsc (left), LHCII (middle), and PSII (right) proteins. b Region of the LC–MS chromatogram corresponding to the elution of LHCII proteoforms displayed as assigned mass features versus retention time (see Supplementary Fig. 1 for overview of all proteins detected in TD-MS). c Average abundances of LHCII proteins and d detailed average abundances of the distinct isoforms and proteoforms for each LHCII protein, accession number of the primary isoform is reported in the box on the right (see Supplementary Fig. 3 for quantiﬁcation of distinct proteoforms detected in TD-MS). Error bars represent standard error of the mean abundance for each proteoform. e Schematic representation of sequence alignment of LHCII isoforms and proteoforms detected in TD-MS analyses and corresponding sequences resolved in the high-resolution structure of the PSII–LHCII supercomplex from pea plants (PDB: 5xnl, and chains therein; the orange box highlights the portion of protein sequence with resolved structure). (C S M )×2, whose relative abundances depend on the light dataset for C light, Supplementary Data 2 and 3). In total, we 2 2 2 intensity . Among them, we focused our structural investigations identiﬁed 260, 304 and 289 crosslinks with DSSO and 358, 479 on the (C S M)×2, representing roughly half of all paired and 495 with EDC for the L, C and H samples, respectively 2 2 PSII–LHCIIsc in any light condition . Identiﬁed crosslinks present (Supplementary Data 3). The overlap between the datasets from in at least two out of three light conditions were mapped on the the three light conditions was ~42% for DSSO and ~34% for EDC, (C S M)×2 model derived from the cryo-EM density map from with 69% and 61% of the crosslinks present in at least two out of 2 2 pea plants (Fig. 1). The crosslinking reactions, performed in three samples, respectively (Fig. 3). These two crosslinkers used in solution under mild conditions (see “Methods” section for details), tandem proved to be highly complementary, while showing also a preserved paired PSII–LHCIIsc from random aggregation while reproducible pattern of subunit linkage between the different maintaining their stromal interactions (Supplementary Fig. 5). We illumination conditions (Supplementary Fig. 6). This suggests that found that inclusion of the TD-MS-derived sequence variants into the overall structure of the paired PSII–LHCIIsc does not undergo the database search was a key step, as the number of detected major structural light-driven remodelling. A small amount of PSI/ crosslinks compared to the non-supplemented available database LHCI proteins was found in low abundance in both the TD- was almost doubled (from 161 to 304 considering the DSSO and XL-MS datasets (Supplementary Data 1 and 3), but a lack of NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications 5 Average fractional abundance [%] Average fractional abundance [%] n12 c153 n12 n3 n2 n8 n4 n4 n5 n60 n55 n3 n2 c149 c153 n60 m [kDa] c162 9.0 9.5 n3 n2 n4 c149 c153 10.0 n23 n60 c162 10.5 11.0 9.0 9.5 10.0 10.5 11.0 9.0 9.5 10.0 10.5 11.0 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 M-trimer Lhcb6 Lhcb3 Lhcb1 Lhcb5 S-trimer Lhcb1 Lhcb2 Lhcb2 Lhcb2 PSII core D1 Lhcb4.2 Lhcb5 PsbH Lhcb4 S-side M-side 90° Lhcb1-S Lhcb4.2 Lhcb5 90° D1 Lhcb1-M Lhcb1-S Lhcb1-S EDC DSSO 4% 6% Self-links 9% 5% 9% Lhcb5 Lhcb1-M 11% Lhcb1-S CP43 34% Lhcb1-S 42% 11% 13% 22% 14% CH CH 7% 13% Fig. 3 Mapping of crosslinks in the paired PSII–LHCIIsc predicted structural model. a Schematic top-view of the (C S M)×2 ﬁtted in the cryo-EM map 2 2 EMD-3825, showing the overall arrangement of the PSII–LHCIIsc subunits in the predicted models 1–3 (M-side and S-side are indicated). b Side-view with mapped crosslinks within the distance cut-off of 17 Å for EDC (blue lines) and 33 Å for DSSO (orange lines), and DSSO self-links (orange-black dashed lines). Venn diagrams showing the overlap between datasets of PSII–LHCIIsc L, C and H samples; only crosslinks present in at least two out of three samples were considered. The enlarged views highlight the crosslinks involving Lhcb2 (c, d) and Lhcb1 (e, f) at the periphery of the supercomplex (d, e) and close to the PSII core (c, f). Subunits are coloured as follows: Lhcb1 in light green; Lhcb2 in cyan; Lhcb4.2 in red; Lhcb5 in yellow; D1 in purple, PsbH in green and CP43 in grey. Subunits not involved in crosslinks are left transparent. inter-links with PSII–LHCIIsc proteins (Supplementary Fig. 6) Considering the evolutionary diversity and extensive light- supports that these rather arose from cross-contamination. The dependent conformational variation of the plant LHCII antenna new components of PSII–LHCIIsc suggested by the TD-MS results system and, conversely, the high conservation of its PSII core ,we (i.e. TL18.3 and Psb27, Supplementary Fig. 1 and Supplementary validated the DSSO and EDC datasets on the high-resolution Data 1) exhibited a low reproducibility at the XL-MS level (Sup- structure available for the pea plant PSII core . By mapping the plementary Fig. 6 and Supplementary Data 3), likely due to their crosslinks detected for this region, we were able to conﬁdently differential accumulation in the different samples (Supplementary place 66 crosslinks for DSSO and 61 for EDC, 87% and 74% of the Data 1). Uncovering their positioning in the wider PSII–LHCIIsc total PSII core detected crosslinks, respectively (Supplementary structure will require a dedicated sample preparation in future Fig. 7). The remaining set of 13% and 26% crosslinks violating the experiments. The amount of intra-protein crosslinks was ~40% strict cut-off distance (i.e. >33 Å for DSSO and >17 Å for EDC) (Supplementary Data 3), a lower percentage compared to that can be however largely validated considering an acceptable large 49 50 commonly found in this type of study. Since most of the cut-off of <35 Å for DSSO and <30 Å for EDC (Supplementary PSII–LHCIIsc subunits are densely packed transmembrane pro- Fig. 7). These over-length crosslinks involved mainly PsbO and teins, for which the membrane itself provides steric hindrance PsbP, two extrinsic subunits protruding on the lumenal side of the reducing solvent-accessible residues, this low percentage is likely PSII core, suggesting that limited structural rearrangements occur determined by the reduced access for the crosslinkers to the within these domains. Overall, the structural validity of the transmembrane helices. Indeed, no crosslinks were found on the majority of the PSII core detected crosslinks supports that under membrane buried parts of the proteins and many of the detected the conditions applied also crosslinks involving the peripheral crosslinks involved ﬂexible loops of several PSII–LHCIIsc com- LHCII can be considered signiﬁcant. ponents. Most of these (35–45% of total crosslinks in every sample) belonged to stroma-exposed LHCII N-terminal loops, Structural modelling of LHCII explains stromal interactions.A which incidentally were not detected when using the non- large number of detected crosslinks involved N-terminal regions. supplemented protein database (Supplementary Data 3). The structural details of the protein interactions they represent 6 NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 ARTICLE are impossible to resolve without access to the N-termini, regions in our structural model (Fig. 3a) and suggests a secondary role for that are missing from the high-resolution structures available so Lhcb2 with respect to Lhcb1 in PSII–LHCIIsc pairing. Further- 15–17 far (e.g. for LHCII see Fig. 2e). To overcome this limitation, more, considering the Lhcb1:Lhcb2 ratio ranging between 2:1 in a the structures of all LHCII proteins along with D1, D2, CP47 and C S and 4:1 in a C S M , as deduced from our structural model 2 2 2 2 2 PsbH of the PSII core with unknown N-terminal atomic details placing Lhcb2 uniquely in the S-trimer, and the higher ratio of were predicted by structural modelling (Supplementary Data 4) Lhcb1 over Lhcb2 observed by TD-MS in the PSII–LHCIIsc in any using the sequences of their most abundant proteoform uncov- light condition tested (Fig. 2), this ﬁnding further suggests an ered by TD-MS (Supplementary Data 1). Substitution of these important role for Lhcb1 in maintaining LHCII trimer super- predicted subunits into the pea PSII–LHCIIsc high-resolution imposition in any type of paired PSII–LHCIIsc at changing light structure can be conﬁdently performed considering the conditions. restricted changes to the PSII–LHCIIsc structures in diverse plant 15–17 species and that they were used as templates. These models Lhcb4 N-terminal loops anchor the paired PSII–LHCIIsc. were subsequently ﬁtted into the cryo-EM structure of the Lhcb4 occupies a pivotal position within the PSII–LHCIIsc, serving (C S M)×2 , signiﬁcantly increasing the number of validated 2 2 as a linker for either S- or M-trimers. Owing to the proximity of crosslinks, with e.g. an increase from 143 to 232 in the DSSO two facing Lhcb4 subunits on the M-side of a paired PSII–LHCIIsc dataset for C light (Supplementary Fig. 8). Placement of Lhcb2 (Fig. 3a), these proteins were previously suspected to provide a within the PSII–LHCIIsc structure was driven by 28 unique inter- structural anchor between facing supercomplexes by tying the protein crosslinks involving Lhcb2 (excluding two inter-links “knot” connection through the mutual interaction of their long N- whose sequences overlapped with Lhcb1; Supplementary Data 3) terminal loops (Fig. 4a). We identiﬁed this subunit by TD-MS and the assumption of one copy of Lhcb2 per S-trimer as pro- (Supplementary Data 1) and, based on the complete amino acid vided by the TD-MS results. We generated nine theoretical sequence, detected numerous crosslinks (Fig. 4b–d). Although we models where Lhcb2 substituted one Lhcb1 in any possible detected two isoforms of Lhcb4 (i.e. Lhcb4.2 and Lhcb4.3, position within the S-trimer, either on the M-side or on the S-side Fig. 2b–e), we only placed Lhcb4.2 into our structural model as it is of the supercomplex (Supplementary Fig. 9 and Supplementary at least tenfold higher in abundance and, importantly, its amount is Data 4). The most probable position of Lhcb2 was determined by stable in all light conditions (Fig. 2c, d). Combined, this led to more combining DSSO and EDC inter-protein crosslinks involving this reproducible crosslinks for Lhcb4.2 than for Lhcb4.3 (Supple- subunit (Supplementary Data 5) and ranking the models con- mentary Data 3). We detected crosslinks involving the stroma- sidering the score of the search algorithm and the number of exposed long hairpin (Pro42–Phe87) of Lhcb4.2 and the PSII core crosslinks involving Lhcb2 validated within the distance threshold proteins CP47 and PsbH (i.e. Lhcb4.2′Glu85-PsbH′Lys24, Lhcb4.2′ (see “Methods” section for details). By ranking the nine theore- Asp74-CP47′Lys227 and Lhcb4.2′Asp74-CP47′Lys130) (Fig. 4b tical models, we found that the highest-ranking model, used and Supplementary Data 3). These crosslinks indicate that at least hereafter, placed the Lhcb2 within the S-trimer close to the PSII half of this hairpin runs mostly parallel to the stromal surface, as core on the M-side and peripherally on the S-side of the super- 15–17 observed in previous PSII–LHCIIsc structures .In these complex (corresponding to models 1–3 in Supplementary Fig. 9 structures, however, the N-terminal domain (Arg1–Asp27) was not and Supplementary Data 4). Two clusters of crosslinks deﬁned resolved. We detected more than 20 crosslinks involving Lhcb4.2 the interactors of Lhcb2 in the predicted (C S M)×2 (Fig. 3a, b) 2 2 N-terminus (Arg1–Asp27) in all light conditions, among which as: (1) near the PSII core on the M-side, the Lhcb2 N-terminal one was a self-link (here deﬁned as intra-link between neigh- loop interacts with the PsbH N-terminus and D1′Glu6 residue, in bouring lysine residues of Lhcb4.2, below the minimum DSSO cut- addition to Lhcb4.2 and Lhcb5 (Fig. 3c); and (2) peripherally on off distance of ~7 Å) (Supplementary Data 3 and Supplementary the S-side, Lhcb2 interacts with Lhcb1 and Lhcb5 (Fig. 3d). Fig. 10). This self-link can unambiguously be assigned to mutual interactions between two Lhcb4.2 subunits (Fig. 4c, d), which act as a structural anchor for the PSII–LHCIIsc pairing across the stromal Lhcb1 determines PSII–LHCIIsc pairing across the stromal gap. Their putative interaction site inferred from the cryo-EM gap. The predicted model for the (C S M)×2 showcased an 2 2 structure (i.e. the “knot”) is ~18 Å from the site predicted by our intricate network of subunits interacting across the stromal gap XL-MS data (Fig. 4c). Considering that no constraints were (Fig. 3b). Indeed, 104 crosslinks for DSSO and 54 for EDC were imposed for the structural prediction of Lhcb4.2, it is conceivable uniquely attributable to subunits interacting across the stromal that the exact position of the two interacting N-terminal loops is gap. We found that Lhcb1 N-terminal loops of facing super- slightly shifted, justifying some degree of mobility that allows their complexes mutually interact, forming well-deﬁned and intricate accommodation within the “knot” density. clusters. These clusters are localized either on the M-side, where six Lhcb1 proteins are peripherally superimposed (Fig. 3a, e), or on the S-side, where two facing Lhcb1 proteins interact with the In situ XL-MS validates PSII–LHCIIsc pairing in thylakoids. PSII core by forming a cluster with the N-terminal loop of D1 Whether the structures of protein complexes seized from their (Lhcb1′Lys2 and Lhcb1′Lys7 crosslink either D1′Glu5 or D1′ cellular milieu represent their native conformation is still largely Asp8; Lhcb1′Lys2 crosslinks also D1′Glu10) and CP43 (i.e. inter- debated. To verify our results in a close-to-native state, we applied link Lhcb1′Lys2-CP43′Lys457) (Fig. 3a, f). Lhcb1 mutual inter- the DSSO XL-MS workﬂow to thylakoid membranes isolated in actions were also supported by self-links, crosslinked peptide stacked conformation from plants grown in moderate light pairs involving the same lysine residue (i.e. Lhcb1′Lys2–Lhcb1′ intensity (C), prior to solubilization and PSII–LHCIIsc puriﬁca- Lys2 and Lhcb1′Lys8–Lhcb1′Lys8) in peptides with different tion. Owing to the considerable amounts of PSII–LHCIIsc missed cleavages, which can only occur if the Lhcb1 interacts with embedded in the thylakoids, we were able to detect 296, 302 and itself across the stromal gap (Fig. 3b–f, Supplementary Fig. 10). 293 crosslinks attributable to PSII–LHCIIsc subunits in each of Mutual interactions between distinct copies of Lhcb1 in the three independent replicates. Of these, only 64, 70 and 67 adjacent supercomplexes were detected in all three light conditions were unique for each replicate, resulting in an overlap of ~77% of (Supplementary Data 3); conversely, such interactions between crosslinks common to at least two replicates (Supplementary Lhcb2 subunits were not observed (Supplementary Data 3). This Data 6). Notably, the crosslinking pattern obtained in situ showed ﬁnding agrees with the distant position of the two Lhcb2 subunits similar subunit linkages compared to in vitro (Supplementary NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 a c ~18 Å EDC DSSO Self-links 90° Pro42-Phe87 d PsbH Lhcb2 Arg1-Asp27 CP47 Lhcb4.2 PsbH Fig. 4 Mapping of Lhcb4.2 crosslinks putatively responsible for the structural anchor of paired PSII–LHCIIsc. a Side-view of the (C S M)x2 ﬁtted in the 2 2 cryo-EM map EMD-3825, with the “knot” connecting density highlighted in red. b Enlarged view highlighting the inter-protein crosslinks involving the N-terminus (Arg1–Asp27) and the long hairpin (Pro42–Phe87) of Lhcb4.2. These two regions of the N-terminal loop of the Lhcb4.2 are shown in black in the inset. Side-view c and end-view d of the putative site of interaction between the ﬂexible N-termini (Arg1–Asp27) of two Lhcb4.2 subunits facing from adjacent supercomplexes are shown, and their ~18 Å displacement from the “knot” density is indicated. Crosslinks within the distance cut-off of 17 Å for EDC (blue lines) and of 33 Å for DSSO (orange lines), and DSSO self-links (orange-black dashed lines) are shown. Subunits are coloured as follows: Lhcb2 in cyan; Lhcb4.2 in red; PsbH in green and CP47 in grey. Subunits not involved in crosslinks are left transparent. Fig. 6), with most of the crosslinks detected in situ in two out of Indeed, in any light condition most of the primary non-truncated three replicates also detected in vitro in PSII–LHCIIsc in at least isoforms of Lhcb1, Lhcb2, Lhcb4.2 and Lhcb5, as well as the PSII two out of three light conditions (Fig. 5a). These results suggest core proteins D1, D2, PsbF and PsbT were found to be acetylated that the XL-MS predicted model for the paired PSII–LHCIIsc can in one of the ﬁrst 20 amino acids by TD-MS (Fig. 6, Supple- be regarded as structurally valid. The coherent positioning of mentary Fig. 3 and Supplementary Data 1). Characterization of Lhcb2 in the predicted structural model was supported by a PTMs on crosslinked peptides involving stroma-exposed reproducible network of crosslinks also found in situ between PSII–LHCIIsc terminal loops, either in vivo or in situ, pin- Lhcb1, Lhcb2, Lhcb4.2 and Lhcb5 (Fig. 5b). PSII–LHCIIsc pointed acetylation predominantly as N-α-acetylation, which structural pairing across the stromal gap was also unambiguously occurred only on proteins with a complete N-terminus (with supported by the occurrence of self-links of Lhcb1 (i.e. Lhcb1′ exception of an acetylated truncated Lhcb6) (Fig. 6 and Supple- Lys8–Lhcb1′Lys8) (Fig. 5c) and Lhcb4.2 (i.e. Lhcb4.2′ mentary Fig. 3). The lack of evidence for N-α-acetylation in Lys10–Lhcb4.2′Lys10 and Lhcb4.2′Lys8–Lhcb4.2′Lys10) (Fig. 5d) Lhcb2 crosslinked peptides (Supplementary Data 3 and 6), (for corresponding spectra see Supplementary Fig. 11). The despite the detection of this PTM at its N-terminal domain coherence between in situ and in vitro DSSO XL-MS results (Fig. 6), suggests the occurrence of a lysine acetylation in this supports that LHCII of PSII–LHCIIsc facing from adjacent region (i.e. likely on Lys5 ), whose interplay with phosphoryla- membranes of stacked thylakoids structurally interact within the tion of Lhcb2-Thr3 is required to trigger the state transitions . native environment through mutual interactions of Lhcb4.2 and Notably, our data suggest that stable N-α-acetylation and rever- Lhcb1 N-terminal loops. sible phosphorylation at the N-terminal domain might play concerted roles. This is based on the potential occurrence of both PTMs in D1 and D2 on the ﬁrst Thr (Fig. 6), as previously Extensive N-terminal acetylation occurs in the stromal gap. reported , and on Lhcb1 and Lhcb4.2 (i.e. the main Lhcbs Both thylakoid stacking and PSII–LHCIIsc pairing depend on the 28,51 involved in PSII–LHCIIsc pairing), where N-α-acetylation occurs presence of cations . Conversely, thylakoid unstacking is on the ﬁrst Arg while the putative phospho-sites are localized triggered by reversible phosphorylation of stroma-exposed N- elsewhere (Fig. 6). terminal loops, introducing negative charges on membrane sur- faces . Here, we found that irrespective of the light condition, acetylation is a widespread PTM on the stroma-exposed N- Discussion terminal loops of many of the PSII–LHCIIsc proteins in stacked The peculiar bipartite structure of plant thylakoid mem- thylakoid membranes, which can be reproducibly detected by branes, consisting of grana stacks and helically wound stroma both TD-MS and XL-MS (Fig. 6 and Supplementary Data 1). lamellae, undergoes light-dependent dynamic structural changes 8 NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 ARTICLE a b PSII-LHCIIsc Lhcb2 Lhcb4.2 Thylakoids Thylakoid PSII-LHCIIsc Self-links membranes 17% 45% 38% Lhcb1-S Lhcb1-M Lhcb5 Lhcb1-S Lhcb4.2 90° 90° Lhcb1-M Lhcb5 Lhcb1-S Lhcb4.2 Fig. 5 Detection of paired PSII–LHCIIsc in stacked thylakoid membranes by in situ XL-MS with DSSO. a Side-view of the (C S M)×2 with mapped 2 2 crosslinks found in at least two out of three thylakoid samples and two out of three PSII–LHCIIsc samples, together with the overlap of the two datasets. Enlarged views matching Figs. 3d, e and 4d are shown to highlight the positioning of Lhcb2 (b) and the occurrence of Lhcb1 and Lhcb4.2 mutual interactions (c, d) in the thylakoid membranes. Crosslinks within the distance cut-off of 33 Å for DSSO either for isolated PSII–LHCIIsc (black lines) or thylakoid membranes (yellow dashed lines), and DSSO self-links (red dashed lines) are shown. Subunits not involved in crosslinks are left transparent. predominantly by remodelling the supramolecular organization determined in this work. The Lhcb1 was the most abundant Lhcb of PSII–LHCIIsc within the grana. Although PSII–LHCIIsc can detected (Fig. 2c), and mutual interactions of Lhcb1 N-terminal be isolated in paired conformation from stacked thylakoid loops across the stromal gap were found to bridge facing 16,28 membranes , the available cryo-EM structures suffer from PSII–LHCIIsc either in vitro (Fig. 3) or in situ (Fig. 5). Similar extensive ensemble averaging, resulting in loss of structural mutual interactions were detected between N-terminal loops of details of sub-stoichiometric subunits and proteoforms with Lhcb4.2 subunits (Figs. 4 and 5). These results provide clear ﬂexible domains, such as the heterogeneous LHCIIs, which are biochemical evidence for the role of Lhcb1 N-terminal loops to 20,23,27 thought to play a role in grana stacking . enforce thylakoid stacking supporting the so-called “Velcro 20,23 In this work, we demonstrated how the integration of in-depth effect” . This effect was hypothesized to be driven by inter- TD-MS proﬁling of intact proteoforms is beneﬁcial for achieving actions of positively charged amino acids at the N-terminus with comprehensive XL-MS analyses performed concomitantly in vitro the negatively charged stromal surface of LHCII, whose mutual on isolated supercomplexes and in situ on thylakoid membranes. interactions across the stromal gap are mediated by cations . Demonstrating the feasibility of such an integrated approach, this Intriguingly, both Lhcb1 and Lhcb4.2 showed extensive N- work paves the way for future disentanglement of the structural terminal truncation in the L sample, with removal of the ﬁrst dynamics of plant PSII–LHCIIsc in response to light cues. This positively charged amino acids (Figs. 2d, e and 6). Similarly, a task might be achieved by performing an accurate quantitative XL- positively charged tail made of the Lys3, Lys4 and Lys5 residues MS analysis (e.g. through TMT-labeling) once cryo-EM structures was missing in the n8 and n12 truncated Lhcb5, which was found at intermediate resolution become available for PSII–LHCIIsc most abundant in the L sample. Conversely, this positively isolated from plants grown under different irradiances. In addi- charged tail was present in the n3 truncated proteoform, which tion, also the potentially novel interactors of the PSII–LHCIIsc was found enriched in the C and H samples (Fig. 2d, e and 6). uncovered in this study (i.e. the lumenal TL18.3 and Psb27 pro- The presence in low light of truncated forms of Lhcb1, Lhcb4.2 teins, Supplementary Fig. 1 and Supplementary Data 1) would and Lhcb5 lacking either positive N-terminal tails or N-α- beneﬁt from such an integrated structural study. acetylation (Fig. 6 and Supplementary Data 1), the latter known Combining results from TD-MS, XL-MS, and integrative to stabilize proteins , might determine an overall imbalance of modelling, we were able to construct and validate a structural surface charges. Furthermore, the accumulation of destabilizing model of (C S M)×2 with one copy of Lhcb2 per S-trimer at a truncated Lhcb1 forms in low light indicates their preferential 2 2 ratio 1:2 with Lhcb1 (Fig. 3). So far cryo-EM produced high- localization within M-trimers, whose selective undocking from resolution structures of plant PSII–LHCIIsc that, due to high PSII–LHCIIsc is advantageous for plants to cope with excessive sequence homology between Lhcb1 and Lhcb2, were unable to irradiation. Accordingly, we found limited amounts of N-α- reveal the exact composition of LHCII trimers in terms of Lhcb1 acetylation on Lhcb6 and absence of this PTM on Lhcb3 (Fig. 6 and Lhcb2 stoichiometries as well as the precise localization of and Supplementary Fig. 3), leaving unshielded the N-termini of 15–17 these two isoforms therein ; our hybrid MS approach did these two M-trimer-related proteins, which are abundant in low succeed in this challenging task. So far the Lhcb2 structure has light (Fig. 2c). This charge-modulated interaction between spe- only been disclosed in the mobile LHCII trimer involved in ciﬁc LHCII N-terminal loops may serve to ﬁnely tune M-trimers binding PSI during state transitions . Intriguingly, this trimer docking to the PSII core and PSII–LHCIIsc pairing across the showed the same subunit composition of the S-trimer stromal gap. This is further exacerbated when negative charges NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications 9 Stromal gap ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 M-trimer Lhcb6 Lhcb3 Lhcb1 Lhcb5 S-trimer Lhcb1 Lhcb2 Lhcb4 90° 25 D1 D2 CP43 4.2 1 2 n4 n5 n4 n12 n2 n3 n8 n12 CP43 15 3 6 D2 D1 12 N-terminal acetylation N-terminal peptide quantified by TD-MS Trimming positions (n) > 98 % > 90 % Acetylation of primary isoform: PSII-LHCIIsc by DSSO-XL-MS N-α-acetylation PSII-LHCIIsc by EDC-XL-MS ac ( ) Thylakoids by DSSO-XL-MS T Putative phosphorylation site Fig. 6 Map of acetylated N-terminal domains of PSII–LHCIIsc subunits spanning the stromal gap. Schematic top- and side-view of the (C S M)×2, 2 2 highlighting the acetylated N-terminal domains of the proteins spanning across the stromal gap detected and quantiﬁed by TD-MS (red box indicates acetylation rate of the complete primary isoform in any light condition, above 98% solid line; above 90% dashed line). Trimming position(s) and putative phosphorylation sites (based on homologous phosphosites previously detected in other plants ) are indicated. N-α-acetylation (ac) detected in crosslinked peptides is shown. N-α-acetylation detected by XL-MS in vitro on isolated PSII–LHCIIsc (treated with DSSO, pentagon; treated with EDC, square) and in situ on the thylakoid membranes (treated with DSSO; triangle) is shown. are introduced through extensive Thr phosphorylation of and Lhcb2 interact closely with D1 and PsbH on the M-side stroma-exposed residues , which are excluded from N-terminal (Fig. 3c) and (2) Lhcb1 forms a tight network with D1 and CP43 truncations (Fig. 6). Light-dependent phosphorylation of LHCII on the S-side (Fig. 3f). As D1 is buried in the bulk of the PSII and PSII components appears to initiate most of the regulatory core and its phosphorylation is crucial for the regulation of the mechanisms that lead to thylakoid structural changes in response whole photosynthetic process , these hubs might play a key role 26,55 to environmental light variations . Since we used dark- in triggering the structural reorganization of PSII–LHCIIsc in adapted samples, extensive phosphorylation was neither expec- response to light cues. ted nor observed. However, the N-terminal tail networks of The results obtained highlight the occurrence of a basic mole- Lhcb1 and Lhcb4.2 uncovered in this study, locking the super- cular mechanism based on mutual interactions of LHCII trimers complex in place (Fig. 3), might facilitate the access through the and Lhcb4.2 N-terminal loops bridging facing PSII–LHCIIsc 10,23 narrow stromal gap for the STN7 and STN8 kinases, which across the stromal gap, that together with a widespread light- are responsible for the light-dependent phosphorylation of the independent N-α-acetylation of stroma-exposed N-terminal loops stroma-exposed N-terminal loops of LHCII and PSII core pro- ultimately strengthen grana stacking at any light condition. In 56–58 teins, respectively (Fig. 6). Indeed, two functional “hubs” conclusion, the power of TD-MS in disentangling LHCII hetero- appear in our (C S M)×2 structural model, where (1) Lhcb4.2 geneity was indispensable for XL-MS and integrative structural 2 2 10 NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications PsbF PsbT PsbF PsbT NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 ARTICLE modelling to unravel the structural details of paired PSII–LHCIIsc the reduced database, ProSight parameters were set as follows: precursor mass tolerance—500 Da; fragment mass tolerance—20 ppm. For High–High searches, in vitro and in situ, providing evidence of their involvement in only Xtract was used with the same parameters for deconvolution of spectra in mediating the grana stacking in plants. both full MS and MS/MS scans. For validation of unreported PTMs and sequence-processing events, custom scripts were used to combine replicate MS/MS scans for each proteoform with Methods distinct precursor masses prior to assigning fragments. Intensities of assigned Isolation of thylakoids and PSII–LHCIIsc puriﬁcation. P. sativum L. plants were masses were z-scored, i.e. the intensity divided by the standard deviation after grown inside the growth chamber SANYO MLR-351H at 20 °C and 60% humidity 60 subtracting the mean. The same approach was employed to characterize abundant for 3 weeks under 8 h daylight at three different light intensities, 30 (low, L), 150 −2 −1 peaks not identiﬁed by the automated searches. Data visualization was done in R (moderate used as control, C) and 750 (high, H) µmol photons m s . The L with the ggplot2 package . To generate proteoform abundance plots and and C conditions were provided by turning on 3 and 15 ﬂuorescent lamps (FL40SS Supplementary Data 1, monoisotopic or average masses of proteoforms were taken W/37) in the growth chamber, respectively; H condition was supplied by four LEDs 61 from a list of identiﬁed precursor masses and matched against deconvoluted mass (LXR7-SW50) mounted inside the growth chamber . Stacked thylakoid mem- features from the full MS-only LC–MS experiments with a mass tolerance window branes were isolated at the end of the daily dark phase, using buffers supplemented 2+ 62 63 of ±2 Da. Then, all the identiﬁed mass features were binned in 3 Da mass windows with divalent cations (Mg ) to mimic the native chloroplast ionic conditions and allowing to ﬁlter out the proteoforms present in <4 technically replicate runs (out preserve the stacked morphology of the grana membranes. After mild solubilization of 6 total runs). Consequently, protein abundances were calculated as the sum of of stacked thylakoid membranes with 50 mM n-dodecyl-α-D-maltoside, puriﬁcation proteoform fractional abundances (i.e. summed intensity of all charge states of paired PSII–LHCIIsc was performed by sucrose gradient ultracentrifugation in the 2+ normalized on total ion intensity of the LC–MS run). dark with a buffer containing divalent cations (i.e. 5 mM Mg ) at mild acidic pH 16,28 (5.7) to preserve their macro-organization and functionality . Optimization of crosslinking conditions. Crosslink reaction conditions were optimized within the range of 0.5–5 mM for DSSO and 1–50 mM for EDC on Top-down sample preparation and LC–MS/MS. A total of 100 µg of PSII–LHCIIsc puriﬁed from plants grown in C light. DSSO was considered as a PSII–LHCIIsc for each light condition (L, C and H) was buffer exchanged into 10% “long-range” crosslinker, with a spacer arm of ~11.3 Å and reactive groups tar- Formic acid by using 5000 MWCO VIVASPIN centrifuge ﬁlters (Vivaproducts geting primary amines (lysine and amino termini of proteins). This reagent pro- Inc., Littleton, USA). The ﬁnal mixture was then diluted to a ﬁnal concentration of vides distance constraints between ~7 and ~31 Å considering the ﬂexibility of 1 µg/µL. To avoid light-induced degradation, samples were prepared under dim lysine side-chains (7 Å + 7 Å) and the α-carbon backbone (6 Å). EDC was con- green light and kept in amber glass thread vials during all consequent steps. sidered as a “short-range” crosslinker, lacking a spacer arm and with reactive Chromatographic separation was performed on a Thermo Scientiﬁc Vanquish Flex groups targeting carboxylic acids and primary amines (aspartic and glutamic acids UHPLC instrument coupled on-line with a MAbPac reversed-phase analytical to lysine and amino termini of proteins). This reagent provides distance constraints column (2.1 mm × 50 mm) heated to 80 °C to a Q Exactive HF-X instrument between 0 and 17 Å considering the ﬂexibility of the side-chains (7 and 5 Å for (Thermo Fisher Scientiﬁc, Bremen, Germany) . A total of 2–3 µg of material was lysine and carboxylic acids, respectively) and the α-carbon backbone (6 Å). The loaded onto the analytical column and separated over 36 min at a ﬂow rate of 250 distance cut-off for DSSO crosslinks was previously increased up to 35 Å con- µL/min. Gradient elution was performed using mobile phases A (H O/0.1% sidering overall protein ﬂexibility , and an average cut-off distance of 33–35 Å was CH O ) and B (C H N/0.1% CH O ): 25–60% B ramp-up in 34 min. Each sample 2 2 2 3 2 2 experimentally determined considering the solvent accessible surface distance .In run was followed by two cleaning cycles with increasing mobile phase B from 10% this work, even considering the high degree of intrinsic PSII–LHCIIsc ﬂexibility, we to 100% and column equilibration for 10 min with 90% buffer A. The high amount decided to consider a strict 33 Å cut-off distance for the long-range of chlorophylls and carotenoids carried by LHCII proteins can signiﬁcantly hamper crosslinker DSSO. protein detection inside the mass spectrometer regardless of ionization conditions. PSII–LHCIIsc, at a ﬁnal protein concentration of 1 mg/mL in a buffer made of However, during the reversed-phase liquid chromatography (RP-LC) step, the 20 mM MES pH 5.7, 0.65 M Sucrose, 10 mM NaCl, 5 mM MgCl were crosslinked strong binding of these molecules to the analytical column led to elution of proteins 2, for 30 and 120 min with DSSO and EDC, respectively. Each reaction was stripped of all pigments simplifying the mass analysis. performed at 4 °C in the dark and without stirring to minimize induced LC–MS(/MS) data were collected with the mass spectrometer set to the Intact conformational changes and unspeciﬁc aggregation. The reaction was then Protein Mode and trapping gas pressure set to 0.2. During analysis, two methods quenched with 100 mM Tris–HCl pH 8. Unspeciﬁc aggregation was investigated were used with complementary resolutions in full MS mode, either medium- 69 60 either by denaturing SDS–PAGE , or by non-denaturing lpBN–PAGE (see resolution of 7500 at 200 Th or high-resolution of 120,000 at 200 Th . Although Supplementary Fig. 5). The occurrence of one clear band in the region the medium-resolution allows for improved detection of ions with masses above corresponding to the mass of paired PSII–LHCIIsc above the 2 MDa MW marker ~30 kDa, at this resolution the instrument lacks sensitivity for detecting low-mass in the lpBN–PAGE further supports the absence of unspeciﬁc aggregation and that ions. In contrast, the high-resolution provides accurate mass detection for ions with EDC, and partially DSSO, are able to maintain the paired conformation during the masses below ~30 kDa but not above. Full MS scans were acquired for the range of electrophoretic run. Once the optimal protein:crosslinker ratio was determined, 400–2400 Th with AGC target set to 3e6. The maximum injection time was set to 125 µg of two independently isolated PSII–LHCIIsc from each condition (L, C and 16 ms with 1 µscan recorded for the medium-resolution and 250 ms with 5 µscans H) were pooled and crosslinked either with DSSO or EDC at their optimal ﬁnal for the high-resolution scans. All MS/MS scans were recorded with a resolution of concentration of 1 and 35 mM, respectively. To remove pigments and sucrose, 120,000, a maximum injection time of 250 ms, an AGC target of 3e6 and 5 µscans compounds potentially interfering with LC–MS analysis, samples were precipitated for the three most intense proteoforms in each cycle as determined by the advanced by stepwise addition of twice the volume of ice-cold acetone every 30 s while precursor determination algorithm . The ions of interest were mass selected by shaking to reach a ﬁnal protein:acetone ratio of 1:8 (v/v) and incubated overnight quadrupole in a 2 Th isolation window and collected to an AGC Target of 3e6 ions at −20 °C. Samples were centrifuged at 15,000 × g for 15 min at 4 °C and the prior to fragmentation at NCE = 30. Only the single most intense charge state acetone completely poured of prior to digestion. Tryptic digestion was conducted was selected for isolation/fragmentation in dd-MS/MS per deconvoluted peak by a two-step workﬂow where the ﬁrst step was performed at 1:50 ratio (protein: array with other charge states excluded from the candidate list for an exclusion protease) at twice the initial volume (i.e. protein concentration of 0.5 mg/mL in time of 6 s. 250 µL) directly on the protein pellet for 4 h at 37 °C constantly shaking to loosen the protein pellet. Pre-digested proteins were denatured by addition of Urea and Top-down data analysis. A custom protein database was derived from the tran- Thiourea to a ﬁnal concentration of 4 and 1 M, respectively. After resuspension of scriptome of P. sativum (p.sativum_csﬂ_reftransV1 downloaded from https://www. the pellet under gentle shaking (~2 h), reduction with 10 mM Dithiothreitol for 1 h coolseasonfoodlegume.org/organism/Pisum/sativum/reftrans/v1) and supple- at 37 °C, and alkylation with 20 mM Iodoacetamide for 30 min at room mented with homologous sequences available for P. sativum in UniProtKB/ temperature in the dark was performed. The second tryptic digestion was TrEMBL (version 17.01.2019 containing 1803 sequences). Isotopically resolved or performed at 1:25 ratio (protein:protease) for 16 h at 37 °C, after diluting the Urea unresolved spectra were deconvoluted, respectively, with either Xtract or ReSpect down to 1 M, and then quenched with 10% Triﬂuoroacetic acid. The ﬁnal peptide (Thermo Fisher Scientiﬁc, Bremen, Germany). Automated searches against our mixtures were desalted with C Sep-Pak cartridges (Waters). The crosslinking database were performed in Thermo Proteome Discoverer (version 220.127.116.112) reaction on thylakoid membranes was performed on three independent samples of extended with the ProSightPD nodes for Medium–High (medium-resolution in full 500 µg of total protein isolated from pea plants grown in moderate light (C). MS) and High–High (high-resolution in full MS) experimental workﬂows. Para- Thylakoids were crosslinked for 30 min with 1 mM DSSO at a ﬁnal protein meters for the Medium–High method were set as follows. ReSpect: precursor m/z concentration of 0.5 mg/mL in a buffer composed of 25 mM MES pH 6.0, 10 mM tolerance—0.2 Th; relative abundance threshold—0%; precursor mass range— NaCl, 5 mM MgCl . After the crosslinking reaction, the samples were further 3–100 kDa; precursor mass tolerance—30 ppm; charge state range—3–100. Xtract: processed with the same procedure used for PSII–LHCIIsc. signal/noise threshold—2; m/z range—400–2400 Th. Initially, a large precursor tolerance window of 10 kDa was set to identify proteins with unknown sequence processing/PTMs followed by cycles of database ﬁltering and manual sequence Fractionation of crosslinked peptides and LC–MS/MS. Fractionation of 250 µg adjustment to produce a reduced database with mature protein sequences (i.e. ﬁnal of desalted crosslinked peptides was performed by HPLC-SCX (strong cation base amino acid sequences resulting from RNA transcript processing and enzy- exchange) chromatography on an Agilent 1200 HPLC system with a C trap matic cleavage of terminal amino acids). For the ﬁnal absolute mass search against column (Opti-Lynx TRAP column C /49 µm, 5 mm) connected to an analytical NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 PolyLC column (PolySULFOETHYL A/3 µm, 50 × 1.0 mm). Samples were recon- Data availability stituted in 10% (v/v) formic acid in water and separated on the analytical column The mass spectrometry raw data and associated databases used in this study have been with a 65 min linear gradient from Buffer A (20% (v/v) ACN and 0.05% (v/v) deposited to the ProteomeXchange Consortium (http://proteomecentral. formic acid in water) to 90% Buffer B (20% (v/v) ACN and 0.05% (v/v) formic acid proteomexchange.org) via the PRIDE partner repository with the dataset identiﬁer in a 0.5 M NaCl) . The fractions containing crosslinked peptides, from min 14 to PXD017382. Detailed reference to all the data shown in the manuscript is reported in the 30, were pooled two by two from min 14–21 and 29–30, while the others were kept Source data ﬁle. separate. This resulted in 11 fractions for each sample, which were desalted with C 96-well elution plates (Oasis HLB), lyophilized and resuspended in 10% Formic Received: 24 June 2019; Accepted: 18 February 2020; acid prior to injection. For measurement, the fractions were injected and separated on a 50 cm × 75 µm C analytical column, packed in-house (Poroshell 120 EC-C / 18 18 2.7 µm) connected to Agilent 1290 LC system and an Orbitrap Fusion Tribrid Mass Spectrometer (Thermo Fisher Scientiﬁc, Bremen, Germany). Acquisition settings for MS analysis of DSSO crosslinked samples were as previously described with minor modiﬁcations. Brieﬂy, for DSSO crosslinked samples, a survey MS1 scan at high-resolution (60,000) was followed by a Top-N of 10 MS2-CID (collision- References induced dissociation) scans on selected high-charged precursors (z = 3–8) at 1. Allen, J. P. Recent innovations in membrane-protein structural biology. resolution of 30,000, producing signature peaks for subsequent MS3 fragmentation F1000Research 8, 211 (2019). of potential crosslinked peptides by higher collisional dissociation (HCD) frag- 2. Barber, J. Photosystem II: the water-splitting enzyme of photosynthesis. Cold mentation at NCE = 30. For samples crosslinked with EDC, a survey MS1 scan at Spring Harb. Symp. Quant. Biol. 77, 295–307 (2012). high-resolution (60,000) was followed by a Top-N of 10 MS2-HCD scans on 3. Pagliano, C., Saracco, G. & Barber, J. Structural, functional and auxiliary selected high-charged precursors (z = 3–8) at a resolution of 30,000. proteins of photosystem II. Photosynth. Res. 116, 167–188 (2013). Raw data were processed with the XlinkX nodes incorporated in Proteome 4. Hankamer, B., Morris, E. P. & Barber, J. Revealing the structure of the oxygen- Discoverer 18.104.22.1682 (Thermo Fisher Scientiﬁc Bremen, Germany). For peptide/ evolving core dimer of photosystem II by cryoelectron crystallography. Nat. protein identiﬁcation Mascot was used to search ﬁltered spectra against two Struct. Biol. 6, 560–564 (1999). databases: (1) the transcriptome-derived database merged with sequences of 5. Umena, Y., Kawakami, K., Shen, J.-R. & Kamiya, N. Crystal structure of oxygen- P. sativum available on UniProtKB/TrEMBL repository and, (2) the TD-MS- evolving photosystem II at a resolution of 1.9 Å. Nature 473,55–60 (2011). derived database containing all proteoforms for which a sequence has been 6. Koziol, A. G. et al. Tracing the evolution of the light-harvesting antennae in conﬁdently determined (Supplementary Data 1). The XlinkX node for the analysis chlorophyll a/b-containing organisms. Plant Physiol. 143, 1802–1816 (2007). of the DSSO crosslinks was set as follows: enzyme name—trypsin (full); maximum 7. Ballottari, M., Girardon, J., Dall’Osto, L. & Bassi, R. Evolution and functional number of missed cleavages—2; minimum peptide length—5 amino acids; properties of Photosystem II light harvesting complexes in eukaryotes. minimum and maximum peptide mass—300 and 7000 Da, respectively. Precursor Biochim. Biophys. Acta—Bioenerg. 1817, 143–157 (2012). mass tolerance was set to 10 ppm, FTMS fragment mass to 20 ppm. Modiﬁcations 8. Betterle, N. et al. Light-induced dissociation of an antenna hetero-oligomer is allowed were carbamidomethylation of cysteines, as a ﬁxed modiﬁcation, and needed for non-photochemical quenching induction. J. Biol. Chem. 284, oxidation of methionines and protein N-terminal acetylation, as variable 15255–15266 (2009). modiﬁcations. For the analysis of EDC crosslinked samples XlinkX was set to focus 9. Albanese, P. et al. Dynamic reorganization of photosystem II supercomplexes on K-DE linkage, using the same peptide and crosslinks search parameters as for DSSO. For both crosslinking reagents, the FDR was controlled at 1% by Percolator. in response to variations in light intensities. Biochim. Biophys. Acta-Bioenerg. 1857, 1651–1660 (2016). 10. Dekker, J. P. & Boekema, E. J. Supramolecular organization of thylakoid membrane proteins in green plants. Biochim. Biophys. Acta—Bioenerg. 1706, Structural modelling and crosslink mapping. Structural predictions with I- 12–39 (2005). TASSER were performed using the amino acid sequences determined by TD-MS 11. Pagliano, C. et al. Proteomic characterization and three-dimensional electron for the most abundant sequence variant (Supplementary Data 1) of every protein microscopy study of PSII–LHCII supercomplexes from higher plants. for which the N-terminus was not resolved in the available high-resolution Biochim. Biophys. Acta—Bioenerg. 1837, 1454–1462 (2014). structure of pea PSII–LHCIIsc (PDB: 5xnl). This includes all LHCII subunits and 12. Caffarri, S., Kouřil, R., Kereïche, S., Boekema, E. J. & Croce, R. Functional PSII proteins D1, D2, CP47 and PsbH (Supplementary Data 4). Of the ﬁve models architecture of higher plant photosystem II supercomplexes. EMBO J. 28, generated by I-TASSER for each protein, only the one with the highest score was 3052–3063 (2009). further considered, despite the very high score of the other models due to the 13. Kouřil, R., Wientjes, E., Bultema, J. B., Croce, R. & Boekema, E. J. High-light template-based prediction. vs. low-light: effect of light acclimation on photosystem II composition and The starting model of the paired supercomplex was derived from the cryo-EM organization in Arabidopsis thaliana. Biochim. Biophys. Acta—Bioenerg. 1827, map of the (C S M)×2 at 14 Å resolution (EMD-3825), ﬁtted with the high- 2 2 411–419 (2013). resolution structure of the C S M (PDB: 5xnl) devoid of one M-trimer and its 2 2 2 14. Nogales, E. & Scheres, S. H. W. Cryo-EM: a unique tool for the visualization of speciﬁc linker Lhcb6 and duplicated to ﬁt the other moiety of the (C S M)×2 cryo- 2 2 macromolecular complexity. Mol. Cell 58, 677–689 (2015). EM map. Final placement of the predicted structures was done by alignment and 15. Wei, X. et al. Structure of spinach photosystem II–LHCII supercomplex at 3.2 substitution in the resulting (C S M)×2 model with matchmaker in Chimera 2 2 Å resolution. Nature 534,69–74 (2016). v1.12 . The root mean square deviation for each structural alignment, considering 16. Su, X. et al. Structure and assembly mechanism of plant C2S2M2-type pruned atom pairs of the backbone that account for over 90% of the total, was PSII–LHCII supercomplex. Science 357, 815–820 (2017). below 1 Å. For Lhcb2, not distinguishable from Lhcb1 because of limited resolution 17. van Bezouwen, L. S. et al. Subunit and chlorophyll organization of the plant of the LHCII trimers and thus currently not localizable with certainty in any known 15–17 high-resolution PSII–LHCIIsc structure , we assumed one copy per S-trimer photosystem II supercomplex. Nat. Plants 3, 17080 (2017). 18. Kühlbrandt, W., Wang, D. N. & Fujiyoshi, Y. Atomic model of plant light- based on our TD-MS and absolute quantiﬁcation results. This generated nine models in total, one for each putative position of one Lhcb2 in both S-trimers harvesting complex by electron crystallography. Nature 367, 614–621 (1994). present in the C S M (Supplementary Fig. 9, Supplementary Data 4), further 19. Liu, Z. et al. Crystal structure of spinach major light-harvesting complex at 2 2 doubled to occupy the other moiety of the structure. All the theoretical models 2.72 Å resolution. Nature 428, 287–292 (2004). were evaluated by scoring as follows. Detected crosslinks were grouped with an in- 20. Standfuss, J., Terwisscha van Scheltinga, A. C., Lamborghini, M. & house script in R generating all possible combinations between repeated subunits in Kühlbrandt, W. Mechanisms of photoprotection and nonphotochemical each structural model considered (i.e. in the (C S M)×2 Lhcb1 is present in 12 2 2 quenching in pea light-harvesting complex at 2.5 A resolution. EMBO J. 24, copies, Lhcb2 in 4 copies, Lhcb3 and Lhcb6 in 2 copies, Lhcb4, Lhcb5 and all PSII 919–928 (2005). core subunits in 4 copies). Crosslinks were then visualized and processed with 21. Crepin, A. & Caffarri, S. Functions and evolution of Lhcb isoforms composing ChimeraX . The scoring of each of the nine models was based on XlinkX score for LHCII, the major light harvesting complex of Photosystem II of green each Lhcb2 unique interlink validated within the distance restraint in the model eukaryotic organisms. Curr. Protein Pept. Sci. 19, 699–713 (2018). considered. To avoid overestimation of crosslinks occurring several times within 22. Galka, W. P. et al. Functional analyses of the plant Photosystem I–Light- the same model (e.g. for the multiple copies of Lhcb1), the sum of XlinkX scores Harvesting Complex II supercomplex reveal that Light-Harvesting Complex II was further weighted by the number of crosslinks ﬁtted in each model loosely bound to Photosystem II is a very efﬁcient antenna for Photosystem I (Supplementary Data 5). Circos-XL plots provided in Supplementary Fig. 6 were in state II. Plant Cell 24, 2963–2978 (2012). produced in R with “circlize” package . 23. Daum, B., Nicastro, D., Austin, J., McIntosh, J. R. & Kühlbrandt, W. Arrangement of Photosystem II and ATP Synthase in chloroplast membranes of spinach and pea. Plant Cell 22, 1299–1312 (2010). 24. Anderson, J. M., Horton, P., Kim, E. H. & Chow, W. S. Towards elucidation of Reporting summary. Further information on research design is available in dynamic structural changes of plant thylakoid architecture. Philos. Trans. R. the Nature Research Reporting Summary linked to this article. Soc. B 367, 3515–3524 (2012). 12 NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 ARTICLE 25. Wood, W. H. J. et al. Dynamic thylakoid stacking regulates the balance 54. Ree, R., Varland, S. & Arnesen, T. Spotlight on protein N-terminal acetylation. between linear and cyclic photosynthetic electron transfer. Nat. Plants 4, Exp. Mol. Med. 50, 90 (2018). 116–127 (2018). 55. Wood, W. H., Barnett, S. F. H., Flannery, S., Hunter, C. N. & Johnson, M. P. 26. Pribil, M., Labs, M. & Leister, D. Structure and dynamics of thylakoids in land Dynamic thylakoid stacking is regulated by LHCII phosphorylation but not its plants. J. Exp. Bot. 65, 1955–1972 (2014). interaction with photosystem I. Plant Physiol. 180, 2152–2166 (2019). 27. Wan, T. et al. Crystal structure of a multilayer packed major light-harvesting 56. Bellaﬁore, S., Barneche, F., Peltier, G. & Rochaix, J.-D. State transitions and complex: implications for grana stacking in higher plants. Mol. Plant 7, light adaptation require chloroplast thylakoid protein kinase STN7. Nature 916–919 (2014). 433, 892–895 (2005). 28. Albanese, P. et al. Pea PSII–LHCII supercomplexes form pairs by making 57. Bonardi, V. et al. Photosystem II core phosphorylation and photosynthetic connections across the stromal gap. Sci. Rep. 7, 10067 (2017). acclimation require two different protein kinases. Nature 437, 1179–1182 29. Pesaresi, P., Pribil, M., Wunder, T. & Leister, D. Dynamics of reversible (2005). protein phosphorylation in thylakoids of ﬂowering plants: the roles of 58. Fristedt, R. & Vener, A. V. High light induced disassembly of Photosystem II STN7, STN8 and TAP38. Biochim. Biophys. Acta—Bioenerg. 1807, 887–896 supercomplexes in Arabidopsis requires STN7-dependent phosphorylation of (2011). CP29. PLoS One 6, e24565 (2011). 30. Wu, X. et al. Lysine acetylation is a widespread protein modiﬁcation for 59. Aro, E.-M., Virgin, I. & Andersson, B. Photoinhibition of Photosystem II. diverse proteins in Arabidopsis. Plant Physiol. 155, 1769–1778 (2011). Inactivation, protein damage and turnover. Biochim. Biophys. Acta—Bioenerg. 31. Longoni, P., Douchi, D., Cariti, F., Fucile, G. & Goldschmidt-Clermont, M. 1143, 113–134 (1993). Phosphorylation of the Light-Harvesting Complex II isoform Lhcb2 is central 60. Albanese, P. et al. Isolation of novel PSII–LHCII megacomplexes from pea to state transitions. Plant Physiol. 169, 2874–2883 (2015). plants characterized by a combination of proteomics and electron microscopy. 32. Koskela, M. M. et al. Chloroplast acetyltransferase NSI is required for state Photosynth. Res. 130,19–31 (2016). transitions in Arabidopsis thaliana. Plant Cell 30, 1695–1709 (2018). 61. Albanese, P. et al. Thylakoid proteome modulation in pea plants grown at 33. Rowland, E., Kim, J., Bhuiyan, N. H. & van Wijk, K. J. The Arabidopsis different irradiances: quantitative proteomic proﬁling in a non-model organism chloroplast stromal N-terminome; complexities of N-terminal protein aided by transcriptomic data integration. Plant J. 96,786–800 (2018). maturation and stability. Plant Physiol. 169, 1181–1196 (2015). 62. Pagliano, C., Barera, S., Chimirri, F., Saracco, G. & Barber, J. Comparison of 34. Tran, J. C. et al. Mapping intact protein isoforms in discovery mode using top- the α and β isomeric forms of the detergent n-dodecyl-D-maltoside for down proteomics. Nature 480, 254–258 (2011). solubilizing photosynthetic complexes from pea thylakoid membranes. 35. van de Waterbeemd, M. et al. Dissecting ribosomal particles throughout the Biochim. Biophys. Acta—Bioenerg. 1817, 1506–1515 (2012). kingdoms of life using advanced hybrid mass spectrometry methods. Nat. 63. Schröppel-Meier, G. & Kaiser, W. M. Ion homeostasis in chloroplasts under Commun. 9, 2493 (2018). salinity and mineral deﬁciency: II. Solute distribution between chloroplasts 36. Herzog, F. et al. Structural probing of a protein Phosphatase 2A network and extrachloroplastic space under excess or deﬁciency of sulfate, phosphate, by chemical cross-linking and mass spectrometry. Science 337, 1348–1352 or magnesium. Plant Physiol. 87, 828–832 (1988). (2012). 64. Kelstrup, C. D. et al. Performance evaluation of the Q exactive HF-X for 37. Greber, B. J. et al. The complete structure of the large subunit of the shotgun proteomics. J. Proteome Res. 17, 727–738 (2017). mammalian mitochondrial ribosome. Nature 515, 283–286 (2014). 65. Fornelli, L. et al. Advancing top-down analysis of the human proteome using a 38. Benda, C. et al. Structural model of a CRISPR RNA-silencing complex reveals benchtop quadrupole-orbitrap mass spectrometer. J. Proteome Res. 16, the RNA-target cleavage activity in Cmr4. Mol. Cell 56,43–54 (2014). 609–618 (2017). 39. Kim, S. J. et al. Integrative structure and functional anatomy of a nuclear pore 66. Zabrouskov, V., Senko, M. W., Du, Y., Leduc, R. D. & Kelleher, N. L. New and complex. Nature 555, 475–482 (2018). automated MSn approaches for top-down identiﬁcation of modiﬁed proteins. 40. Schweppe, D. K. et al. Mitochondrial protein interactome elucidated by J. Am. Soc. Mass Spectrom. 16, 2027–2038 (2005). chemical cross-linking mass spectrometry. Proc. Natl Acad. Sci. USA 114, 67. Wickham, H. ggplot2: Elegant Graphics for Data Analysis. (Springer Science & 1732–1737 (2017). Business, New York, 2009). 41. Fasci, D., van Ingen, H., Scheltema, R. A. & Heck, A. J. R. Histone interaction 68. Leitner, A. et al. The molecular architecture of the eukaryotic chaperonin landscapes visualized by crosslinking mass spectrometry in intact cell nuclei. TRiC/CCT. Structure 20, 814–825 (2012). Mol. Cell. Proteom. 17, 2018–2033 (2018). 69. Laemmli, U. K. Cleavage of structural proteins during the assembly of the 42. Chavez, J. D. et al. Chemical crosslinking mass spectrometry analysis of head of bacteriophage T4. Nature 227, 680–685 (1970). protein conformations and supercomplexes in heart tissue. Cell Syst. 6, 70. Klykov, O. et al. Efﬁcient and robust proteome-wide approaches for cross- 136–141.e5 (2018). linking mass spectrometry. Nat. Protoc. 13, 2964–2990 (2018). 43. Leitner, A. et al. Chemical cross-linking/mass spectrometry targeting acidic 71. Roy, A. et al. The I-TASSER Suite: protein structure and function prediction. residues in proteins and protein complexes. Proc. Natl Acad. Sci. USA 111, Nat. Methods 12,7–8 (2014). 9455–9460 (2014). 72. Pettersen, E. F. et al. UCSF Chimera—a visualization system for exploratory 44. Shi, Y. et al. Structural characterization by cross-linking reveals the detailed research and analysis. J. Comput. Chem. 25, 1605–1612 (2004). architecture of a coatomer-related heptameric module from the nuclear pore 73. Goddard, T. D. et al. UCSF ChimeraX: meeting modern challenges in complex. Mol. Cell. Proteom. 13, 2927–2943 (2014). visualization and analysis. Protein Sci. 27,14–25 (2018). 45. Albanese, P., Manfredi, M., Marengo, E., Saracco, G. & Pagliano, C. Structural 74. Gu, Z., Gu, L., Eils, R., Schlesner, M. & Brors, B. Circlize implements and functional differentiation of the light‐harvesting protein Lhcb4 during and enhances circular visualization in R. Bioinformatics 30, 2811–2812 land plant diversiﬁcation. Physiol. Plant. 166, 336–350 (2019). (2014). 46. Chen, H. et al. A Psb27 homologue in Arabidopsis thaliana is required for 75. Vizca¡no, J. A. et al. The Proteomics Identiﬁcations (PRIDE) database efﬁcient repair of photodamaged photosystem II. Plant Mol. Biol. 61, 567–575 and associated tools: status in 2013. Nucleic Acids Res. 41, D1063–D1069 (2006). (2013). 47. Sirpiö, S. et al. TLP18.3, a novel thylakoid lumen protein regulating photosystem II repair cycle. Biochem. J. 406, 415–425 (2007). 48. Barera, S., Pagliano, C., Pape, T., Saracco, G. & Barber, J. Characterization of PSII–LHCII supercomplexes isolated from pea thylakoid membrane by one- Acknowledgements step treatment with α- and β-dodecyl-D-maltoside. Philos. Trans. R. Soc. Lond. We thank Dominique Hagemans for technical assistance with HPLC-SCX and Barbara B 367, 3389–3399 (2012). Steigenberger for help with the XL-MS experiments, fruitful discussions and for syn- 49. Bullock, J. M. A., Schwab, J., Thalassinos, K. & Topf, M. The importance of thesizing the DSSO used in this study. P.A. acknowledges Instruct-ERIC for supporting non-accessible crosslinks and solvent accessible surface distance in modeling the Internship at Utrecht University (APPID: 368). R.A.S. acknowledges The Netherlands proteins with restraints from crosslinking mass spectrometry. Mol. Cell. Organization for Scientiﬁc Research (NWO) for supporting this research through Proteom. 15, 2491–2500 (2016). funding of the large-scale proteomics facility Proteins@Work (project 184.032.201) 50. Fernandez-Martinez, J. et al. Structure and function of the nuclear pore embedded in the Netherlands Proteomics Centre. Furthermore, R.A.S. acknowledges that complex cytoplasmic mRNA export platform. Cell 167, 1215–1228.e25 (2016). this work is part of the research programme NWO TA with project number 741.018.201, 51. Barber, J. Inﬂuence of surface charges on thylakoid structure and function. which is partly ﬁnanced by the Dutch Research Council (NWO); additional support Annu. Rev. Plant Physiol. 33, 261–295 (1982). came through the European Union Horizon 2020 programme INFRAIA project Epic-XS 52. Sharma, J. et al. Primary structure characterization of the photosystem II D1 (Project 823839). This work was supported by the Italian Ministry of Education, Uni- and D2 subunits. J. Biol. Chem. 272, 33158–33166 (1997). versity and Research, “Futuro in Ricerca 2013” programme RBFR1334SB to C.P. The 53. Pan, X. et al. Structure of the maize photosystem I supercomplex with light- authors P.A. and C.P. thank Prof. Tomas Morosinotto and Prof. Roberto Barbato for harvesting complexes I and II. Science 360, 1109–1113 (2018). their critical reading of the manuscript. NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-15184-1 Author contributions Reprints and permission information is available at http://www.nature.com/reprints P.A. and C.P. initially conceived the project. All authors planned and designed the experiments. P.A. prepared samples, performed crosslinking MS experiments, analysed Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in data and performed structural modelling. S.T. performed top-down MS experiments and published maps and institutional afﬁliations. analysed data. C.P. and R.A.S. supervised the project. G.S. and R.A.S. contributed experimental tools. P.A. and C.P. wrote the manuscript with substantial input from R.A.S. and S.T. All authors critically revised the paper and approved its ﬁnal version. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give Competing interests appropriate credit to the original author(s) and the source, provide a link to the Creative The authors declare no competing interests. Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless Additional information indicated otherwise in a credit line to the material. If material is not included in the Supplementary information is available for this paper at https://doi.org/10.1038/s41467- article’s Creative Commons license and your intended use is not permitted by statutory 020-15184-1. regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Correspondence and requests for materials should be addressed to R.A.S. or C.P. licenses/by/4.0/. Peer review information Nature Communications thanks Michael Hippler, Yi Shi and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. © The Author(s) 2020 Peer reviewer reports are available. 14 NATURE COMMUNICATIONS | (2020) 11:1361 | https://doi.org/10.1038/s41467-020-15184-1 | www.nature.com/naturecommunications
Nature Communications – Springer Journals
Published: Mar 13, 2020
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera