Crosstalk between chloroplast protein import and the SUMO system revealed through genetic and molecular investigation

The chloroplast proteome contains thousands of different proteins that are encoded by the nuclear genome. These proteins are imported into the chloroplast via the action of the TOC translocase and associated downstream systems. Our recent work has revealed that the stability of the TOC complex is dynamically regulated by the ubiquitin-dependent chloroplast-associated protein degradation (CHLORAD) pathway. Here, we demonstrate that the stability of the TOC complex is also regulated by the SUMO system. Arabidopsis mutants representing almost the entire SUMO conjugation pathway can partially suppress the phenotype of ppi1, a pale yellow mutant lacking the Toc33 protein. This suppression is linked to the increased stability of TOC proteins and improvements in chloroplast development. In addition, we demonstrate using molecular and biochemical experiments that the SUMO system directly targets TOC proteins. Thus, we have identified a regulatory link between the SUMO system and chloroplast protein import.


27
The chloroplast is a membrane-bound organelle that houses photosynthesis in all green plants 28 (Jarvis and Lopez-Juez, 2013). Chloroplasts have an unusual evolutionary history -they are the 29 integrated descendants of a cyanobacterial ancestor that entered the eukaryotic lineage via 30 endosymbiosis. Although chloroplasts retain small genomes, almost all of the proteins required for 31 chloroplast development and function are now encoded by the central, nuclear genome (Jarvis, 32 2008). These proteins must be imported into the organelle after synthesis in the cytosol, and this 33 import is mediated by the coordinate action of the TOC and TIC complexes (the translocons at the 34 outer and inner envelope membranes of chloroplasts) (Jarvis, 2008). 35 The TOC complex contains three major components: the Omp85 (outer membrane protein, 85 kDa)- ligase, was identified. A series of sp1 mutations were shown to partially suppress the phenotypic 46 defects of ppi1 with respect to chlorosis, chloroplast development, and chloroplast protein import. 47 In addition, SP1 function was shown to promote plastid interconversion events (for example, the 48 development of the chloroplast from its precursor organelle, the etioplast). Later work 49 demonstrated that SP1 function is also important for abiotic stress tolerance, by enabling 50 optimisation of the organellar proteome via protein import regulation (Ling and Jarvis, 2015). Thus, 51 through SP1, the ubiquitin-proteasome system promotes TOC complex degradation and 52 reconfiguration in response to developmental and/or environmental stimuli. 53 Ubiquitinated TOC proteins are extracted from the chloroplast outer envelope membrane and 54 degraded in the cytosol. Recent work identified two proteins that physically associate with SP1 and 55 promote the membrane extraction of TOC proteins (Ling et al., 2019). These are SP2, an Omp85-type 56 β-barrel channel protein that was identified in the same genetic screen as SP1, and Cdc48, a well-57 characterised cytosolic AAA+ chaperone ATPase that provides the motive force for the extraction of 58 proteins from the chloroplast outer envelope. The three proteins -SP1, SP2 and Cdc48 -together 59 define a new pathway for the ubiquitination, membrane extraction, and degradation of chloroplast 60 outer envelope proteins, which has been named chloroplast-associated protein degradation, or 61 CHLORAD. In addition to CHLORAD, there exist cytosolic ubiquitin-dependent systems that also 62 contribute to chloroplast biogenesis, by regulating the levels of unimported preproteins (Lee et al., 63 2009; Grimmer et al., 2020), and by controlling the stability of the Toc159 receptor prior to its 64 integration into the outer envelope membrane (Shanmugabalaji et al., 2018). 65 The discovery of SP1 and the CHLORAD pathway demonstrated that the TOC complex is not static 66 but, instead, can be rapidly ubiquitinated and degraded in response to developmental and 67 environmental stimuli. To complement this work, we decided to explore whether the TOC complex is 68 also regulated by the SUMO system. This work was motivated by the results of a high-throughput 69 screen for SUMO substrates in Arabidopsis (Elrouby and Coupland, 2010). This screen suggested that 70 Toc159, a key component of the TOC complex, is a SUMO substrate. SUMOylation is intricately 71 involved in plant development and stress adaptation, and so we were interested to determine 72 whether the TOC complex is targeted by the SUMO system, and whether any such SUMOylation is 73 functionally important. As crosstalk between the SUMO system and the ubiquitin-proteasome 74 system is common, we reasoned that answering these questions might provide insights into the 75 regulation of SP1 and the CHLORAD pathway. 76 To explore the relationship between chloroplast protein import and the SUMO system, we carried 77 out a comprehensive series of genetic, molecular and biochemical experiments. Mutants 78 representing most components of the Arabidopsis SUMO pathway were found to partially suppress 79 the phenotype of the chlorotic Toc33 null mutant, ppi1, with respect to leaf chlorophyll 80 accumulation, chloroplast development, and TOC protein abundance. Conversely, overexpression of 81 either SUMO1 or SUMO3 enhanced the severity of the ppi1 phenotype. Moreover, the E2 SUMO 82 conjugating enzyme, SCE1, was found to physically interact with the TOC complex in bimolecular 83 fluorescence complementation experiments; and TOC proteins were seen to physically associate 84 with SUMO proteins in immunoprecipitation assays. In combination, our data conclusively 85 demonstrate significant crosstalk between the SUMO system and chloroplast protein import, and 86 emphasise the complexity of the regulation of the TOC translocase. 87

88
The E2 SUMO conjugating enzyme mutant sce1-4, and the E3 SUMO ligase mutants siz1-4 and siz1-89 5, partially suppress the phenotype of Toc33 mutant, ppi1 90 Two key components of the CHLORAD pathway, SP1 and SP2, were identified in a forward genetic 91 screen for suppressors of ppi1, an Arabidopsis double mutant was phenotypically characterised, and, intriguingly, it appeared greener than the 108 ppi1 single mutant ( Figure 1A; Figure 1 -Supplement 1A). This was linked to a moderate increase in 109 leaf chlorophyll concentration ( Figure 1B; Figure 1 -Supplement 1B). Next, we asked whether the 110 phenotypic suppression observed in ppi1 sce1-4 was linked to changes in the development of 111 chloroplasts. The chloroplasts of ppi1 sce1-4 were visualised via transmission electron microscopy. 112 Interestingly, the chloroplasts of the ppi1 sce1-4 double mutant appeared larger and better 113 developed than those of the ppi1 control ( Figure 1C). The transmission electron micrographs were 114 quantitatively analysed, and the ppi1 sce1-4 chloroplasts were indeed found to be significantly larger 115 than those of ppi1 ( Figure 1D) maturity. We mapped the integration sites of the T-DNA insertions in these two mutants (Figure 1 -124 Supplement 2A), and showed that both display a strong reduction in SIZ1 transcript by RT-PCR 125 analysis (Figure 1 -Supplement 2B). In addition, both mutants displayed defects in global 126 SUMOylation in response to heat shock, similar to the published alleles ( Figure 1 -Supplement 3). 127 The two new siz1 mutants were crossed with ppi1 and the resulting double mutants were 128 phenotypically characterised. Both the ppi1 siz1-4 and the ppi1 siz1-5 double mutants appeared 129 greener than the ppi1 control ( Figure 1G; Figure 1 -Supplement 1C). In addition, the double mutants 130 showed dramatic increases in leaf chlorophyll concentration relative to ppi1 ( Figure 1H; Figure 1 -131 Supplement 1D). Next, we asked whether the phenotypic suppression observed in ppi1 siz1-4 and 132 ppi1 siz1-5 was linked to changes in the abundance of TOC proteins. To this end, protein samples 133 were taken from the two double mutants and relevant control plants and resolved via 134 immunoblotting. Both double mutants displayed clear increases in the abundance of Toc159 and 135 Toc75, two core components of the TOC complex, relative to ppi1 ( Figures 1I, 1J, and 1K). 136 The suppression effects mediated by the SUMO system mutants are specific 137 As discussed in the previous section, the SUMO system is encoded by a remarkably small number of 138 genes in Arabidopsis. As a consequence, SUMO system mutants have highly pleiotropic molecular 139 and physiological phenotypes. We therefore asked whether the partial suppression of ppi1 by SUMO 140 system mutants was specific to the ppi1 background. We crossed sce1-4 with tic40-4 and hsp93-V-1, 141 two TIC-complex-associated mutants. These mutants are chlorotic, due to defects in protein import 142 across the chloroplast inner membrane, and in this respect are highly similar to ppi1 (Kovacheva et  143 al., 2005). Significantly, the resulting double mutants, tic40-4 sce1-4 and hsp93-V-I sce1-4, were 144 indistinguishable from tic40-4 and hsp93-V-1, their respective single mutant controls (Figures 2A and  145 2C). Moreover, the double mutants did not display changes in leaf chlorophyll accumulation relative 146 to the single mutant controls ( Figures 2B and 2D). We therefore concluded that the suppression 147 effects observed in ppi1 sce1 plants were background-specific and associated with the TOC complex. 148 Next, we asked whether the sce1-4 and siz1-4 single mutants display an increase in chlorophyll 149 concentration even in the wild-type background. However, neither mutant appeared greener than  Strikingly, SCE1-nYFP was found to physically associate with all tested TOC proteins -nYFP-Toc159, 166 nYFP-Toc132, nYFP-Toc34 and nYFP-Toc33 ( Figure 3). Moreover, these interactions were 167 concentrated at the periphery of the chloroplasts, placing them in an appropriate subcellular context 168 for the in situ regulation of the chloroplast protein import machinery. Conversely, SCE1-cYFP was not 169 found to physically associate with the negative control protein ΔOEP7-nYFP. First, we analysed SUMO1 and SUMO2. We obtained sum1-1 and sum2-1, two previously 183 characterised null mutants (Saracco et al., 2007), and crossed them with ppi1. To account for the 184 functional redundancy between these two genes, we also sought a ppi1 sum1-1 sum2-1 triple 185 mutant. However, as SUMO1 and SUMO2 are collectively essential, ppi1 sum1-1 sum2-1 plants that 186 were homozygous with respect to ppi1 and sum2-1, but heterozygous with respect to the sum1-1 187 mutation, were selected from a segregating population. The double and triple mutants were 188 phenotypically characterised, and all three appeared larger and greener than the ppi1 control plants 189 ( Figure 4A). Moreover, the double and triple mutants showed corresponding increases in leaf 190 chlorophyll concentration, with the triple mutant showing a larger increase than the double mutants 191 ( Figure 4B). These were synthetic effects, as the sum1-1, sum2-1, and sum1-1 sum2-1 single and 192 double mutants did not appear greener than wild-type plants, or show increases in chlorophyll 193 accumulation ( Figure 4 -Supplement 1). We therefore concluded that the sum1-1 and sum2-1 194 mutants can additively suppress the phenotype of ppi1. 195 To complement the preceding experiment, we generated transgenic plants overexpressing SUMO1 196 in the ppi1 background. The SUMO1 coding sequence was cloned into a vector carrying a strong, decreases in leaf chlorophyll concentration ( Figure 4D). 204 Next, we turned our attention to SUMO3. We obtained sum3-1, a previously characterised null 205 mutant (van den Burg et al., 2010), and crossed it with ppi1. The resulting double mutant was 206 phenotypically characterised, but it did not appear obviously different from the ppi1 control ( Figure  207 4E). Correspondingly, it did not display any clear increase in leaf chlorophyll concentration relative to 208 ppi1 ( Figure 4F). To complement this experiment, we generated transgenic plants overexpressing 209 SUMO3 in the ppi1 background, using the approach described above, and a line carrying a single, 210 homozygous insert was identified and taken forward for analysis. The overexpression of SUMO3 was 211 confirmed by RT-PCR (Figure 4 -Supplement 2B). Interestingly, the transgenic plants showed a 212 striking increase in the severity of the ppi1 phenotype: the plants were severely dwarfed and paler 213 than the ppi1 control ( Figure 4G), and displayed a significant decrease in leaf chlorophyll 214 accumulation ( Figure 4H). These findings are particularly noteworthy when considered alongside a 215 previous report which explored the consequences of overexpressing SUMO3 in wild-type plants (van 216 den Burg et al., 2010). In that study, SUMO3 overexpression was not found to alter the appearance 217 of the transgenic plants, which implies a degree of specificity in the phenotypic accentuation 218 observed here. 219

Biochemical analysis reveals SUMOylation of TOC proteins in vivo 220
The genetic and molecular experiments described thus far strongly suggested that TOC proteins are 221 SUMOylated. However, to our knowledge, conclusive evidence that chloroplast-resident proteins are 222 SUMOylated is currently lacking. To investigate whether chloroplast proteins may be SUMOylated, 223 we isolated chloroplasts from seedlings by cell fractionation and analysed them by anti-SUMO 224 immunoblotting. For this analysis, we employed a proven commercial antibody against SUMO1, 225 which is one of the most abundant SUMO isoforms in Arabidopsis making it more tractable for 226 analysis, and which furthermore is known to accumulate in response to heat and other stresses 227 ( immunoblotting. The YFP-HA and SCE1-YFP fusion proteins both showed robust expression and 241 strong recovery in the IP elutions ( Figure 5A). Remarkbly, the SCE1-YFP fusion protein was found to 242 be associated with native Toc159 and Toc132, but not with the negative control proteins Tic110 or 243 Tic40 ( Figure 5A). Conversely, YFP-HA did not associate with any of the tested proteins. Given that 244 SCE1 is a promiscuous enzyme that associates with thousands of proteins (Elrouby and Coupland, 245 2010), and that these interactions are likely to be transient, it is remarkable that TOC co-elution was 246 detectable in this experiment. 247 In the second experiment, we cloned the SUMO1, SUMO2 and SUMO3 coding sequences into a 248 vector that appends an N-terminal YFP tag to its insert ( in protoplasts alongside the YFP-HA negative control construct. As in the previous experiment, the 253 protoplasts were subjected to YFP-Trap immunoprecipitation, and the samples were subsequently 254 analysed by immunoblotting. Remarkably, all three YFP-SUMO proteins were found to physically 255 associate with Toc159, although YFP-SUMO3 clearly bound Toc159 with the greatest affinity ( Figure  256 5B; Figure 5 -Supplement 3). Moreover, inspection of an extended exposure of the anti-YFP blot 257 revealed a number of higher molecular weight bands that we interpret to be SUMO adducts and 258 indicative of the functionality of the fusions ( Figure 5 -Supplement 4). In contrast with the SUMO 259 fusions, the YFP-HA negative control did not associate with Toc159, and none of the four YFP fusion 260 proteins physically associated with Tic40, a negative control protein ( Figure 5B). 261 The immunoprecipitation experiment described above identified SUMO3 as having the highest 262 affinity for Toc159. To extend our analysis of SUMO3 to include another TOC proteins, and to 263 provide more direct evidence for TOC protein SUMOylation, the experiment was repeated with 264 modifications, as follows. Protoplasts were co-transfected with YFP-SUMO3 and Toc33-HA, or YFP-265 HA and Toc33-HA; in each case, Toc33 was transiently overexpressed to aid detection of this 266 component and its adducts. Upon co-expression of these construct pairs, the protoplast samples 267 were subjected to YFP-Trap immunoprecipitation analysis, as described earlier. In accordance with 268 the Toc159 result ( Figure 5B), YFP-SUMO3, but not YFP-HA, was found to physically associate with 269 Toc33-HA ( Figure 5C). Moreover, bands of the exact expected molecular weight for Toc33-HA 270 bearing one or two YFP-SUMO3 moieties (75 and 114 kDa) were also detected. These bands were 271 accompanied by a high molecular weight smear at the top of the immunoblot, which is indicative of 272 complex, multisite or chain SUMOylation. 273

274
This work has revealed a genetic and molecular link between the SUMO system and chloroplast 275 protein import. The genetic experiments demonstrated that SUMO system mutations can suppress 276 the phenotype of the Toc33 mutant, ppi1, while the molecular and biochemical experiments 277 indicated that TOC proteins associate with key SUMO system proteins and are SUMOylated. Visible 278 suppression effects observed in the ppi1 / SUMO system double mutants were linked to 279 improvements in chloroplast development and enhanced accumulation of key TOC proteins. Thus, 280 our results suggest that SUMOylation acts to destabilise the TOC complex, and that when such 281 SUMOylation is perturbed the TOC proteins are stabilised. We interpret that TOC complexes 282 containing Toc34, Toc75 and Toc159 accumulate at higher levels in ppi1 / SUMO system double 283 mutants, and that this synthetically improves the double mutant phenotypes relative to the ppi1 284 control. Importantly, each core TOC protein, including all of those analysed in this study, was 285 predicted with high probability to have one or more SUMOylation sites (  Table 2  344 for primer sequences) and phenotypic analysis (including the double and triple mutants). The 345 positions of the T-DNA insertions were mapped via PCR. The following primer pairs were used to 346 generate diagnostic amplicons from gDNA: LB1 and Siz1-Seq-1R (for mapping siz1-4), and LBb1 and 347 Siz1-Seq-3R (for mapping siz1-5) (see Table 2D

Chloroplast isolation and protein extraction 363
Chloroplasts were isolated from 14-day-old, plate-grown seedlings as described previously (Flores-364 Pérez and Jarvis, 2017). Some of the seedlings were heat-shocked immediately prior to chloroplast 365 isolation. To do this, the plates containing the seedlings were wrapped in clingfilm and placed into a 366 water bath (42°C for one hour, followed by a one hour recovery period at 22°C). Protein samples 367 were prepared from the isolated chloroplasts by extraction using SDS-PAGE sample buffer, as well as 368 from whole 14-day-old seedlings as previously described (Kovacheva et al., 2005). In some cases, the 369 samples were treated with 10 mM N-ethylmaleimide (Hilgarth and Sarge, 2005); this was added 370 directly to the protein extraction buffer (whole seedling samples), or to the chloroplast isolation 371 buffer following polytron homogenization and all subsequent buffers (chloroplast samples). 372

Plasmid constructs 373
The constructs used in the BiFC experiments were generated as follows. The coding sequences of 374 SCE1, SIZ1, TOC159, TOC132, TOC34 and TOC33 were PCR amplified from wild-type cDNA (see Table  375 2 for primer sequences). In the case of ΔOEP7, the first 105 base pairs of the OEP7 coding sequence 376 were amplified; this encodes a truncated sequence which is sufficient to efficiently target the full- The constructs used in the immunoprecipitation experiments were generated as follows. The coding 382 sequences of SCE1, SUMO1, SUMO2 and SUMO3 were PCR amplified from wild-type cDNA using 383 primers bearing 5' attB1 and attB2 adaptor sequences (see Table 2 for primer sequences). The 384 amplicons were then cloned into pDONR221 (Invitrogen), a Gateway entry vector. The inserts from 385 the resulting entry clones were then transferred to one of two destination vectors: p2GWY7 (SCE1) 386 or p2YGW7 (SUMO1, SUMO2, SUMO3); the former appends a C-terminal YFP tag to its insert, and 387 the latter appends an N-terminal YFP tag to its insert ( Toc33-HA and YFP-HA constructs have been described previously (Ling et al., 2019). 389 The constructs used to generate transgenic plants were generated as follows. The coding sequences 390 of SUMO1 and SUMO3 were PCR amplified from wild-type cDNA using primers bearing 5' attB1 and 391 attB2 adaptor sequences (see Table 2 for primer sequences  Table 2 for primer sequences). 408

Transmission electron microscopy 409
Transmission electron micrographs were recorded using mature rosette leaves as previously co-expressed with a construct encoding Toc33-HA. The protoplasts were solubilised using IP buffer 448 containing 1% Triton X-100, and the resulting lysates were incubated with GFP-Trap beads 449 (Chromotek). After four washes in IP buffer, the protein samples were eluted by boiling in SDS-PAGE  450 loading buffer, and then analysed by immunoblotting. 451

Statistical analysis 452
The data from each experiment were analysed in R. In most cases, two-tailed T-tests were 453 performed. However, in one case, a one-way ANOVA was performed in conjunction with a Tukey 454 HSD test (as indicated in the figure legend   The transmission electron microscopy dataset was quantified. There were significant differences 640 between the ppi1 and ppi1 sce1-4 plants (Two-tailed t-test, unpaired samples, T = 4.65, p = 641 0.009674). (E, F) Thylakoid membrane development was increased in ppi1 sce1-4 relative to ppi1. 642

References and notes
The number of stacked thylakoidal lamellae per granum (E), and the number of stromal thylakoidal 643 lamellae emanating from each granum (granal interconnections) (F), was analysed using the 644 transmission electron microscopy dataset. There were significant differences between the ppi1 and 645  The siz1-4 and siz1-5 mutants display reduced global SUMOylation in response to heat shock. 708 14-day-old seedlings of the indicated genotypes were subjected to heat shock (42°C for 1 hour, 709 followed by a 1-hour recovery period at 22°C). Protein samples were taken from whole seedlings and 710 analysed by immunoblotting. The siz1-4 and siz1-5 mutants displayed a moderate reduction of global 711 SUMOylation, and this reduction was similar in magnitude to the reduction observed in the 712 previously characterised siz1-2 and siz1-3 mutants. OEP7, which is sufficient to direct targeting to the chloroplast outer envelope), which served as a 759 negative control. Scale bars = 10 µm.

Figure 4 762
Genetic interactions between ppi1 and the genes encoding three SUMO isoforms. 763 ppi1 sum1-1, ppi1 sum2-1, and ppi1 sum1-1 † sum2-1   analysis. The YFP-HA construct served as a negative control. In both cases, three samples were 829 analysed: The 'Total lysate' sample (total protein extract from solubilised protoplasts); the 'Flow 830 through' sample (the total protein sample after incubation with anti-YFP beads); and the 'IP' sample 831 (the eluted fraction of the total protein sample that bound to the anti-YFP beads). The samples were 832 analysed by immunoblotting, revealing that SCE1-YFP, but not the YFP-HA control, was associated 833 with native Toc159 and Toc132 (indicated by the two arrows). Neither SCE1-YFP nor YFP-HA was 834 associated with native Tic110 or Tic40, which were included as negative control proteins. two or several YFP-SUMO3 motifs were detected on the membrane (indicated by the upper arrows). 848 The predicted molecular weight of YFP-SUMO3 is approximately 38.9 kDa. Neither YFP-SUMO3 nor 849 YFP-HA was associated with Tic40, which served as a negative control protein. The expression of the SCE1-YFP construct (A), and of the three YFP-SUMO constructs (B), was 872 analysed and confirmed by imaging transfected Arabidopsis protoplasts. Chlorophyll 873 autofluorescence images were employed to orientate the YFP signals in relation to the chloroplasts. 874 Representative confocal micrographs show a typical protoplast in each case. Scale bars = 10 µm. 875 an additional negative control. The samples were run on a 10% acrylamide gel for 1 hour, as 882 opposed to an 8% acrylamide gel for four hours (as in Figure 5B)  The membrane shown is the same as the one presented in Figure 5B (YFP panel). In this case, the 893 membrane was visualised using a long exposure to aid the detection of weakly abundant protein 894 bands. As well as the YFP-SUMO monomers, additional high molecular bands and smears were seen