Complex Evolution of Light-Dependent Protochlorophyllide Oxidoreductases in Aerobic Anoxygenic Phototrophs: Origin, Phylogeny, and Function

Abstract Light-dependent protochlorophyllide oxidoreductase (LPOR) and dark-operative protochlorophyllide oxidoreductase are evolutionary and structurally distinct enzymes that are essential for the synthesis of (bacterio)chlorophyll, the primary pigment needed for both anoxygenic and oxygenic photosynthesis. In contrast to the long-held hypothesis that LPORs are only present in oxygenic phototrophs, we recently identified a functional LPOR in the aerobic anoxygenic phototrophic bacterium (AAPB) Dinoroseobacter shibae and attributed its presence to a single horizontal gene transfer event from cyanobacteria. Here, we provide evidence for the more widespread presence of genuine LPOR enzymes in AAPBs. An exhaustive bioinformatics search identified 36 putative LPORs outside of oxygenic phototrophic bacteria (cyanobacteria) with the majority being AAPBs. Using in vitro and in vivo assays, we show that the large majority of the tested AAPB enzymes are genuine LPORs. Solution structural analyses, performed for two of the AAPB LPORs, revealed a globally conserved structure when compared with a well-characterized cyanobacterial LPOR. Phylogenetic analyses suggest that LPORs were transferred not only from cyanobacteria but also subsequently between proteobacteria and from proteobacteria to Gemmatimonadetes. Our study thus provides another interesting example for the complex evolutionary processes that govern the evolution of bacteria, involving multiple horizontal gene transfer events that likely occurred at different time points and involved different donors.


Introduction
It is widely accepted that photosynthesis, the process by which photosynthetic organisms convert light energy into chemical energy, has evolved early on in Earth's history (Blankenship 2010;Hohmann-Marriott and Blankenship 2011). There are two types of photosynthesis: oxygenic and anoxygenic. In oxygenic photosynthesis, performed by cyanobacteria and plants, light energy is used for the oxidation of water, thereby releasing oxygen, electrons, and protons. In contrast, in anoxygenic photosynthesis, performed by anoxygenic phototrophic bacteria (APBs), for example, hydrogen sulfide, hydrogen or other organic substrates are used as electron donors, while the process does not generate oxygen (Hanada 2016). Anoxygenic photosynthesis hereby likely predates oxygenic photosynthesis (Hohmann-Marriott and Blankenship 2011), with the latter being assumed to have first evolved in an ancestor of cyanobacteria (Blankenship 2010;Hohmann-Marriott and Blankenship 2011;Cardona et al. 2015). Although the timing and mechanism by which oxygenic photosynthesis arose is still debated (Soo et al. 2017;Martin et al. 2018;Cardona 2019), its emergence undoubtedly led to the oxygenation of the primordial atmosphere, thereby laying the foundations for life on Earth as we know it today.
At the foundation of photosynthesis, light-absorbing pigments such as chlorophylls (Chls) and carotenoids, as part of photosynthetic reaction centers (RCs), enable the harvesting of light energy (Blankenship 2010). Chls hereby represent the main class of light-harvesting pigments essential for all phototrophic organisms of the bacterial and eukaryotic domains. It is therefore not surprising that the evolution of Chl biosynthesis, which is a branch in the synthesis pathway of modified tetrapyrroles (Bryant et al. 2020), is linked to the evolution of photosynthesis.
A key step in the complex biosynthesis pathway of Chls and bacteriochlorophylls (Bchls) is the reduction of the C17¼C18 double bond of the protochlorophyllide (Pchlide) D-ring to yield chlorophyllide (Chlide) (Beale 1999;Blankenship 2010). During evolution of (bacterio)chlorophyll biosynthesis two structurally distinct enzyme systems have emerged capable of catalyzing this reaction, namely darkoperative protochlorophyllide oxidoreductases (DPORs) and light-dependent protochlorophyllide oxidoreductases (LPORs) (Suzuki and Bauer 1995;Schoefs and Franck 2003;Yang and Cheng 2004;Reinbothe et al. 2010). For a long time, it was widely accepted that APBs contain only DPORs (Suzuki and Bauer 1995;Schoefs and Franck 2003;Yang and Cheng 2004), whereas plants, with the exception of gymnosperms, contain only LPORs (Yang and Cheng 2004;Sousa et al. 2013). The majority of cyanobacteria, ferns, mosses, gymnosperms, and algae have both enzyme systems (Fujita 1996). DPORs are multisubunit protein complexes consisting of three subunits (called BchN, BchB, BchL in APBs and ChlN, ChlB, ChlL in oxygenic phototrophs), which contain iron-sulfur clusters, and are therefore sensitive to oxygen ). They convert Pchlide in an ATP-dependent process independent of light (Bröcker et al. 2010). In contrast, LPORs are oxygen-insensitive, single-component NADPH-dependent enzymes of the short-chain dehydrogenase (SDR) family of enzymes (Townley et al. 2001), which convert Pchlide in a strictly light-dependent process (Schoefs and Franck 2003). DPORs, as the evolutionary older enzymes, likely evolved in the anoxygenic environment of the early Earth and share a common ancestor with nitrogenase-like enzymes Hu and Ribbe 2015). In contrast, LPORs are considered as evolutionary younger enzymes, which were speculated to have evolved in cyanobacteria at about the time of the great oxygenation event (Yamazaki et al. 2006). This hypothesis is primarily based on their oxygen insensitivity and their presumed absence in APBs (Yamazaki et al. 2006).
The recent discovery of a functional LPOR in the aerobic anoxygenic phototrophic a-proteobacterium Dinoroseobacter shibae DFL12 T (DsLPOR; Kaschner et al. 2014) challenged this hypothesis. Aerobic anoxygenic phototrophic bacteria (AAPBs) are a ubiquitous group of marine microbes, related to facultative anaerobic purple non-sulfur bacteria (Biebl et al. 2005). In contrast to classical APBs like Rhodobacter sp., AAPBs can perform anoxygenic photosynthesis in the presence of atmospheric oxygen (Yurkov and Beatty 1998). It seems therefore a reasonable adaption for AAPBs to possess an LPOR enzyme system to enhance Bchl synthesis under aerobic conditions. In another recent study (Kasalicky et al. 2017), additional six LPORs in nonoxygenic photosynthetic bacteria were reported. However, their functionality was not investigated.
Here, we show that LPORs are more common among AAPBs than originally assumed. We identify 36 LPORs outside of oxygenic phototrophic genera, verify activity for 10 out of 11 tested AAPB LPORs, provide biochemical and solution structural data of six and two AAPB LPORs, respectively, and discuss evolutionary processes that have led to their wide distribution outside of oxygenic phototrophic bacteria.

Results and Discussion
Identification of LPORs Outside of Oxygenic Phototrophs Prompted by our recent identification of a genuine LPOR in the AAPB D. shibae (DsLPOR) (Kaschner et al. 2014), we comprehensively analyzed 116,919 bacterial genomes covering all bacterial phyla (available in GenBank; Clark et al. 2016 To determine whether bacteria carrying putative LPORs are AAPBs, we scanned the corresponding genomes for the presence/absence of various genes (using HMMER as described above for LPOR; supplementary table S2, Supplementary Material online) typically associated with aerobic anoxygenic photosynthesis in the genomes, being well aware that the results of the bioinformatics analysis depend on the quality of the genomes provided. We scanned for the presence of three DPOR subunits BchL, BchN, and BchB proteins as markers of anoxygenic photosynthesis, whereas the absence of the large chain subunit of the ribulose-1,5bisphosphat-carboxylase/-oxygenase (RuBisCO) and phosphoribulokinase (PRK) was used as a sign for AAPBs (Brinkmann et al. 2018). We also checked the presence of the oxygen-independent (BchE) and oxygen-dependent (AcsF) magnesium-protoporphyrin IX monomethyl ester oxidative cyclase enzymes as markers for aerobic and semiaerobic chlorophototrophs (Boldareva-Nuianzina et al. 2013;Zeng et al. 2014 figure 1A and supplementary table S1, Supplementary Material online. We call a bacterium (putative) AAPB, if it possesses the three DPOR subunits, RC type II and AcsF (and/or BchE), and lacks RuBisCO (and/or PRK). For 19 out of 36 bacteria carrying a putative LPOR, it was experimentally shown that they are AAPBs ( fig. 1A) (Wakao et al. 1993;Hanada et al. 1997;Labrenz et al. 2000;Van Trappen et al. 2004;Biebl et al. 2005;Labrenz et al. 2005;Gich and Overmann 2006;Yoon et al. 2006;Wang et al. 2014;Zeng et al. 2014Zeng et al. , 2015Kasalicky et al. 2017;Cai et al. 2018;Hahn et al. 2018;Hoetzinger et al. 2019). Sixteen of 19 experimentally verified AAPBs fulfill the above rule ( fig. 1A), which means that the rule is conservative in the sense that we may miss some AAPBs. In summary, the majority of bacteria (31 out of 36) were assigned to AAPBs (see also supplementary section 1.  (Townley et al. 2001;Buhr et al. 2008;Menon et al. 2009) are conserved between known and putative LPORs from plants, cyanobacteria, and AAPBs.

AAPB LPORs Show Light-Dependent Activity
For functional characterization, we selected 11 putative AAPB LPORs from different taxonomic families ( fig. 1 and supplementary table S1, Supplementary Material online; identified by LPOR ID). We hereby selected putative AAPB LPORs that did not derive from metagenomes and were included in the RefSeq database (hence having a low probability of representing an artifact). Additionally, we included: the AAPB LPOR from D. shibae (DsLPOR) (Kaschner et al. 2014); two cyanobacterial LPORs-Synechocystis sp. PCC6803 (SsLPOR) and Thermosynechococcus elongatus BP-1 (TeLPOR); as well as two plant LPORs-Hordeum vulgare (POR A; HvLPORA) and Arabidopsis thaliana (POR C; AtLPORC) (Heyes et al. 2000;Pattanayak and Tripathy 2002;McFarlane et al. 2005;Buhr et al. 2008) (supplementary table S1, Supplementary Material online). An actinobacterial SDR from Saccharopolyspora erythraea (SeSDR), which shares 33.4% identical positions with TeLPOR (7% gaps), was analyzed as negative control. LPOR function was investigated by in vitro assays using either crude cell extracts or purified protein and by an in vivo complementation assay using a Rhodobacter capsulatus strain which carried a deletion in the bchB gene encoding for one of the three DPOR subunits (Kaschner et al. 2014) (see also supplementary section 1.2, Supplementary Material online). We consider an LPOR as functional, if it was active in at least one of the assays. The results of exemplary in vitro and in vivo activity tests are shown in figure 2 (for all results, see supplementary figs. S2 and S3 [in vitro], figs. S4 and S5 [in vivo], and table S4, Supplementary Material online). The negative control, SeSDR was inactive in both assays (in vivo and in vitro). As reported previously (Kaschner et al. 2014), DsLPOR was active in both assays. As positive controls, TeLPOR and AtLPORC showed high and intermediate activity in both assays, whereas for SsLPOR and HvLPOR, low, unsteady or no activity was observed (for details about semiquantitative ranking, see supplementary table S4, Supplementary Material online). Intriguingly, 10 out of 11 tested putative AAPB LPORs showed activity in at least one of the two assays. Seven AAPB LPORs (EbLPOR, ElLPOR, SpLPOR, PdLPOR, LfLPOR, YvLPOR, and AnLPOR) were active in both assays. SaLPOR showed high activity in vivo but was inactive in vitro. SlLPOR and GpLPOR were active in vitro but inactive in vivo. SgLPOR was the only putative AAPB LPOR without signs of activity. This observation can most likely be attributed to aggregation of the improperly folded enzyme in both Escherichia coli and R. capsulatus, as functionally important residues are conserved between SgLPOR and the other AAPB LPORs (supplementary fig. S3, Supplementary Material online). In summary, LPOR functionality could be verified for 10 out of 11 tested AAPB LPORs, thus clearly demonstrating that functional LPORs are prevalent among AAPBs. Future work should include functional tests for b-proteobacterial LPORs of the order Burkholderiales ( fig. 1A), which due to their later identification could not be tested as part of this study.

Biochemical Properties of AAPB, Cyanobacteria, and Plant LPORs
We next biochemically characterized six putative AAPB LPORs (ElLPOR, EbLPOR, PdLPOR, GpLPOR, SpLPOR, and LfLPOR), DsLPOR, as well as the cyanobacterial TeLPOR and plant AtLPORC. Unfortunately, we were unable to heterologously produce and purify other plant and cyanobacterial LPORs (HvLPORA and SsLPOR). All LPORs, which could be purified in sufficient quantity and purity, were characterized with regard to pH-and temperature-optima, temperature stability (temperature-dependent unfolding), dissociation constant K d of Pchlide binding to the ternary LPOR/ NADPH/Pchlide complex (see Materials and Methods for details), the influence of the reducing agent dithiotreitol (DTT) on the K d , and the preference for either monovinyl (MV)-and divinyl (DV)-Pchlide substrates. The latter aspect of Rhodobacter capsulatus wild-type strain and the DbchB (DPOR deficient) strain complemented with SeSDR as a negative control as well as known (TeLPOR, DsLPOR) and exemplary putative LPORs (ElLPOR, SpLPOR). (B-D) Growth curves (black lines, OD 660nm ) and BChl a accumulation, measured as normalized in vivo absorption at 860 nm (green bars, OD 860nm /OD 660nm ) over time of cultivation, as well as absorption spectra of cellular BChl a (normalized to the cell density, 240 h cultivation time). Assays were carried out for the same set of LPOR enzymes as shown in panel (A). (E) Exemplary absorption changes at 672 nm with illumination time, indicating Pchlide to Chlide turnover by known LPORs (TeLPOR, DsLPOR) and selected putative AAPB LPORs (SpLPOR, ElLPOR). As the different enzymes showed variably high activities, the data are shown on Complex Evolution of LPORs in Aerobic Anoxygenic Phototrophs . doi:10.1093/molbev/msaa234 MBE was tested because reactive thiol groups of cysteines have been implicated in either Pchlide binding or catalysis (Heyes et al. 2000). An exemplary overview of the respective data for the TeLPOR and DsLPOR enzymes is shown in figure 3 (for all results see supplementary section 1.3; supplementary figs. S6-S10 and S13, Supplementary Material online). The complete set of characteristics is summarized in table 1.
The plant and cyanobacterial enzymes showed activities between 0.02 U mg À1 (AtLPORC) and 0.65 U mg À1 (TeLPOR), whereas the activity of the AAPB LPORs similarly varied between 0.09 U mg À1 (DsLPOR) and 0.69 U mg À1 (EbLPOR). The same holds for their pH-activity optimum, with all LPORs showing similar optima between pH ¼ 7.0 and 9.0, and their thermostability (see supplementary sections 1.3.1 and 1.3.2, Supplementary Material online), where all LPORs, with the exception of the thermophilic TeLPOR enzyme, possessed melting temperatures between $35 and $55 C.
For the pH-and temperature-optimum range, AAPB LPORs, with a few exceptions, showed a broader 80% optimum range compared with the plant and cyanobacterial LPORs ( With regard to their substrate preference for MV-or DV-Pchlide, the plant and cyanobacterial LPOR enzymes showed no preference in terms of the specific activity determined for MV-or DV-Pchlide as substrate. We consider a variant to show a preference if the fold-difference between the respective activities is >1.5. In contrast, all AAPB LPORs, with the exception of GpLPOR, displayed clear preferences, with ElLPOR and PdLPOR favoring MV-over DV-Pchlide, whereas EbLPOR, SpLPOR, LfLPOR, and DsLPOR seem to favor DV-Pchlide (table 1, supplementary table S9 and  In summary, AAPB LPORs are similar to cyanobacterial and plant LPORs with respect to catalytic activity, pHactivity optimum, and thermostability. However, there are notable differences in pH-and temperature-optimum ranges (broader ranges for AAPB LPORs), K d of the ternary NADPH/ Pchlide/LPOR complex (higher values for AAPB LPORs) and variation of K d in the presence of DTT, compared with no apparent differences for cyanobacterial and plant LPORs. Moreover, in contrast to AtLPORC and TeLPOR, AAPB LPORs showed a substrate preference for MV-or DV-Pchlide. Interestingly, some of the analyzed LPORs showed higher (ElLPOR, PdLPOR, SpLPOR) or lower (AtLPOR, LfLPOR) specific activities, when their activity was measured using a mixture of MV-/DV-Pchlide as compared with the same measurement performed with the two separate substrates (see table 1; compare specific activity and activity MV and activity DV). This might hint at activating or inhibitory effects caused by the MV/DV substrate mixture, but exploring these intriguing features in more detail is outside the scope of the current work and merit further study. Analyzing more cyanobacterial and plant LPORs will provide more information about the variation in their biochemical properties across different species and could solidify the findings that the observed variation in AAPB LPORs is indeed intrinsic to these enzymes.

AAPB LPORs and Cyanobacterial LPORs Possess a Conserved Structure
To identify whether AABP LPORs show a globally conserved structure when compared with cyanobacterial LPORs, we elucidated the solution structure of the apo form (protein without NADPH and Pchlide) of two AAPB LPORs (DsLPOR and ElLPOR) in comparison to the cyanobacterial TeLPOR (Schneidewind et al. 2019) enzyme by using small-angle Xray scattering (SAXS). SAXS allows the structural characterization of biomacromolecules such as proteins in solution. In contrast to X-ray diffraction experiments performed on protein crystals, SAXS does not provide information on atomic coordinates. It is hence described as a low-resolution technique that is capable of providing high-precision information with respect to size and shape of the studied molecule (Neylon 2008;Jacques and Trewhella 2010). SAXS hereby also provides information about the oligomeric state of the studied protein and allows the computational reconstruction of their low-resolution shape as ab initio models.
All studied LPOR samples contained an N-terminal, 20 amino acids long, His-tag. To rule out that the flexible Nterminal His-tag contributes to the obtained ab initio models, we also included an ElLPOR sample, which possessed an eight-amino acid-long His-tag at the C-terminus instead of the N-terminus (ElLPOR-cHis). The corresponding molecular masses estimated from the SAXS data (table 2) are in good agreement with the theoretical molecular masses, indicating that all studied apo LPORs are monomeric, as also previously shown for TeLPOR (Schneidewind et al. 2019). To gain a better understanding of their solution structure, we compared our SAXS data with different DsLPOR, ElLPOR, and In addition, the obtained homology models were superimposed to the corresponding ab initio models, which represent the basic shape of the molecule directly computed from the SAXS data ( fig. 4, lower panels; see Materials and Methods for details). As previously shown for TeLPOR (Schneidewind et al. 2019), the overall shape of the molecule resembles a bowling-pin appearing to consist of a larger and smaller subdomain ( fig. 4, lower panels) that well accommodate the corresponding LPOR homology models. Taken together, the presented SAXS analyses provide a glimpse on the structural organization of AAPB LPORs and suggest that the newly discovered AAPB LPOR enzymes and prototypical cyanobacterial LPORs possess a conserved global structure.

Phylogeny and Evolution of the LPOR System
Phylogenetic studies to elucidate the evolution of LPOR among plants and cyanobacteria suggested that plant FIG. 3. Exemplary biochemical characterization of DsLPOR and TeLPOR, with regard to pH optima (A), temperature optima (B), temperaturedependent unfolding (C), K d of the ternary LPOR/NADPH/Pchlide complex (D, E), and MV-and DV-Pchlide substrate acceptance (F). (A) Three suitable buffer systems covering different pH values were used: sodium phosphate buffer, red data points; Tris buffer, blue data points; glycine buffer, green data points. (B) Temperature optima were determined in Tris buffers whose pH was adjusted at the target temperature. The gray line marks the arbitrary 80% activity threshold that was set to derive the pH optimum range and the T optimum range for the respective enzyme (see table 1). (C) Temperature-dependent unfolding was monitored by DSF. Three independent measurements per enzyme are shown, and the obtained melting temperature (T M ) is given. (D) The K d of the ternary LPOR/NADPH/Pchlide holoprotein complex was determined by quantifying (lower panels in D and E) the red-shift of the Pchlide Q y band due to complex formation (upper panels in D and E; free Pchlide in black; PchlideþLPOR concentration series, shades of red; Pchlideþhighest LPOR concentration, blue). (F) MV/DV-Pchlide preference determined as specific activity with 3.5 6 0.15 mM MV-(black bars) and DV-Pchlide (red bars) as substrate. Error bars correspond to the standard deviation of the mean of three independent measurements. Statistical significance (two-tailed, paired t-test, P 0.05).
Complex Evolution of LPORs in Aerobic Anoxygenic Phototrophs . doi:10.1093/molbev/msaa234 MBE LPORs were obtained by endosymbiotic gene transfer from cyanobacteria (Yang and Cheng 2004;Sousa et al. 2013). Based on the previous assumption that LPORs are absent in anoxygenic phototrophs, along with the observed oxygen-sensitivity of the DPOR system, Yamazaki et al. (2006) argued that LPORs first evolved in cyanobacteria at around the time of the great oxygenation event, that is, as a consequence of increased atmospheric oxygen levels. The authors reasoned that the altered environmental conditions would compromise DPOR function and hence provide the selective pressure for the development of the oxygeninsensitive LPOR enzyme system (Yamazaki et al. 2006). Previously, we attributed the presence of LPOR in D. shibae to a single horizontal gene transfer (HGT) event from cyanobacteria. The here presented widespread distribution of LPORs outside of cyanobacteria challenges this hypothesis, resulting in the need to reconsider the emergence and evolution of LPORs.
For phylogenetic analysis, we used 33 different AAPB LPORs plus 203 cyanobacterial LPORs (see supplementary section 1.4.1 and table S10, Supplementary Material online). Phylogenetic trees were reconstructed by performing 50 independent runs with RAxML (Stamatakis 2014) and IQ-TREE (Nguyen et al. 2015), resulting in 100 topologically different trees. According to the approximately unbiased tree test (Shimodaira 2002), none of the 100 trees was significantly worse than the others.
The majority rule consensus tree (     Thus, we consider LPOR as an example of HGT in the other direction from picocyanobacteria to AAPB proteobacteria. Based on the monophyly of the AAPB-clade within cyanobacteria, we assume that there was one HGT from cyanobacteria to AAPBs.
We now want to trace the sequence of events that led to the distribution of LPORs within AAPBs. The tree in figure 5 suggests three HGT scenarios: (H1) acquisition of LPOR before the split of aand bproteobacteria; (H2) HGT from cyanobacteria to a-proteobacteria and subsequently from ato b-proteobacteria; and (H3) HGT from cyanobacteria to b-proteobacteria and subsequently from bto a-proteobacteria.
To decide which scenario is likely we considered splitting times provided by timetree.org database (Kumar et al. 2017) as well as additional divergence times from four recent studies (Shih et al. 2017;Betts et al. 2018;Magnabosco et  According to the position of the AAPB-clade within picocyanobacteria, the presumable HGT from cyanobacteria to AAPBs happened between the split of Synechococcus elongatus from other cyanobacteria (1,484 Ma; 1 Ma ¼ 1 million years ago) and the divergence of other picocyanobacteria (801 Ma) (the corresponding range with 95% confidence interval is marked in fig. 6B).
The mean of estimates for the divergence of a-/b-/c-proteobacteria suggests that their split is much older than the earliest emergence of the potential cyanobacterial donor ( fig. 6B, red arrow), which rules out the hypothesis H1 of HGT to the ancestor of a-/b-proteobacteria. The time divergence estimates for a-proteobacteria with LPORs (orders Rhodobacterales and Sphingomonadales) and b-proteobacteria (Burkholderiales order: genera Limnohabitans and Polynucleobacter) allow for HGT from cyanobacteria to the respective ancestral lineages ( fig. 6B, green and blue arrows). Therefore, according to the timeline hypotheses H2 and H3 are possible.
Nevertheless, note, that the split of b-proteobacteria genera is much younger than the divergence of a-proteobacteria orders with LPORs. Therefore, if LPORs were transferred from cyanobacteria to b-proteobacteria and then from b to a we would expect more LPORs in b-proteobacteria. However, the opposite is observed, thus, favoring the hypothesis H2 of HGT The transfer of LPOR from ato b-proteobacteria also corroborates evolutionary studies on other photosynthetic genes, suggesting that b-proteobacteria obtained their photosynthetic apparatus from a-proteobacteria (Igarashi et al. 2001;Nagashima S and Nagashima KVP 2013;Imhoff et al. 2017Imhoff et al. , 2019. For example, in the study by Imhoff et al. (2019) on proteins of RC type II (PufHLM) and key enzymes of Bchl biosynthesis (BchXYZ) b-proteobacteria (Burkholderiales and Rhodocyclales) cluster within a-proteobacteria (Rhodospirillales and Rhizobiales). In the LPOR tree ( fig. 5), two Rhodospirillales from Acidiphilium genus cluster within Burkholderiales (b). It is possible that if additional LPOR sequences are identified in Rhodospirillales and/or in Rhizobiales the position of Acidiphilium species might change to resemble that of the PufHLM and BchXYZ proteins. Taking the findings from photosynthetic genes into account, it seems more plausible that LPOR in b-proteobacteria (Burkholderiales) was obtained via HGT from a-proteobacteria; probably together with other photosynthetic genes via a single HGT.
Finally, we point out that LPOR phylogeny for a-proteobacteria resembles the PufHLM-BchXYZ phylogeny and 16S rRNA tree (Imhoff et al. 2019). Therefore, the HGT from cyanobacteria to a-proteobacteria should have occurred before the split of Rhodobacterales and Sphingomonadales.
The LPOR of Gemmatimonadetes bacterium clusters within b-proteobacterial LPORs as a sister of Limnohabitans. At the time of revision of the manuscript, we identified an LPOR in Gemmatimonas sp. TET16 FIG. 5. A subtree of the majority rule consensus LPOR tree displaying the position of the AAPB-derived sequences within cyanobacteria. The branches are colored according to the number of occurrences of the corresponding clade in 100 inferred trees. The activity of tested LPORs was mapped onto the subtree. The activity of LPORs is marked with filled/empty squares, that is, active/not active in vivo (in red) and in vitro (in blue).
Chernomor et al. . doi:10.1093/molbev/msaa234 MBE (GenBank accession number WP_171227737.1), which clustered with Gemmatimonadetes bacterium within b-proteobacteria (data not shown). This suggests an HGT from bproteobacteria to Gemmatimonadetes. The basal placement of G. phototrophica in the AAPB-clade (outside of aand b-proteobacteria) seems to complicate the explanation. However, Zeng et al. (2014) showed that the phylogenetic position of G. phototrophica is inconsistent. For photosynthetic genes encoding AcsF and BchIDH enzymes, the same basal placement as in the LPOR tree was observed. On the A B FIG. 6. (A) The dated tree for the considered taxonomic units. The ranges in blue mark differences between minimum and maximum estimates for the corresponding nodes, when more than one estimate is available. The ranges in pink correspond to 95% confidence intervals as reported in the original article (see supplementary table S11, Supplementary Material online). Taxonomic units in gray do not possess a known LPOR. We mapped the position of AAPB-clade in the LPOR tree onto dated tree (indicated by the red circle). (B) Zoomed view of the time frame and lineages relevant for the discussion of HGT from cyanobacteria to AAPB-proteobacteria. The estimated time frame for cyanobacterial donor is between the divergence estimates of Synechococcus elongatus from other cyanobacteria and the divergence of other picocyanobacteria (light grey range). The 95% confidence intervals of divergence estimates are marked with the dark gray ranges. The colors of the diamonds correspond to the species in (A). Three alternative hypotheses are marked by the red, green, and blue arrows. The earliest emergence of cyanobacterial donor is more recent than the mean divergence estimate of a/b-proteobacteria, ruling out hypothesis of HGT from cyanobacteria to the stem of a/b-proteobacteria (H1, marked with red dashed arrow). According to the time frame, HGT from cyanobacteria to a and subsequently to b (H2, green arrows) is feasible. Also an alternative explanation: HGT from cyanobacteria to b and subsequently to a (H3, blue arrows) is feasible. Complex Evolution of LPORs in Aerobic Anoxygenic Phototrophs . doi:10.1093/molbev/msaa234 MBE other hand, BchLNB from G. phototrophica clusters within proteobacteria. Moreover, also in the PufHLM-BchXYZ phylogeny (Imhoff et al. 2019) G. phototrophica clusters basal to Burkholderiales (b) within Rhizobiales (a). It was also shown that photosynthetic gene cluster of G. phototrophica resembles that of b-proteobacteria Rubrivivax gelatinosus IL144 (Zeng et al. 2014). Summarizing, LPORs of Gemmatimonadetes species show high similarity to b-proteobacterial LPORs. Gemmatimonadetes bacterium and the newly identified LPOR of Gemmatimonas sp. TET16 clustered within b-proteobacteria. The positioning of G. phototrophica remains unclear, until more data become available. The position of Gemmatimonadetes species in LPOR tree ( fig. 5) suggests b-proteobacteria as donor for the HGT and, in principle, is in concordance with the published phylogenies on other photosynthetic genes.
Overall, our analysis suggests that AAPBs originally acquired LPORs via HGT from cyanobacteria to proteobacteria with the donor likely being an ancestral picocyanobacterium. Given multiple evidence of HGT between picocyanobacteria and proteobacteria (Badger and Price 2003;Beiko et al. 2005;Dvornyk 2006;Zhaxybayeva et al. 2006;Marin et al. 2007;Sousa et al. 2013), this transfer event seems very likely. Taking into account evolutionary studies on other photosynthetic genes and also the evolutionary timeline, we suggest a cascade of HGT: cyanobacteria transferred LPOR genes to aproteobacteria; then a-proteobacteria transferred it to b-proteobacteria; finally, b-proteobacteria transferred LPOR genes to Gemmatimonadetes. This scenario is in accordance with the suggested transfer of photosynthetic apparatus from ato b-proteobacteria and to Gemmatimonadetes. Identifying more LPORs in nonoxygenic phototrophic bacteria is needed to solidify our findings and to support the idea of HGT cascade.
Concluding Remarks: Potential Origin, Evolution, and Function of AAPB LPORs From our phylogenetic analysis, we conclude that AAPB LPORs were most likely originally transferred from ancestral picocyanobacteria to a-proteobacteria. Picocyanobacteria are the smallest cyanobacteria, ubiquitous in freshwater, brackish, and marine environments. In freshwater, mainly Synechococcus, Cyanobium, and Synechocystis genera are found, whereas marine habitats are dominated by Synechococcus and Prochlorococcus (Jasser and Callieri 2016). Numerous examples of HGT between picocyanobacteria and proteobacteria have been previously reported (Badger and Price 2003;Beiko et al. 2005;Dvornyk 2006;Zhaxybayeva et al. 2006;Marin et al. 2007;Sousa et al. 2013), providing conclusive evidence, that HGT between these bacterial groups is feasible and frequent.
Further, for six AAPBs with biochemically characterized LPORs (ElLPOR, PdLPOR, GpLPOR, SpLPOR, LfLPOR, and DsLPOR) we related LPOR biochemical properties with the growth characteristics of the host and their evolution. Habitat, site of isolation, and growth characteristics with regard to pH and temperature (if known) are summarized in supplementary table S16, Supplementary Material online. All AAPBs were isolated from aquatic habitats such as freshwater lakes and marine habitats. Some of the organism were isolated from cyanobacterial/microbial mats (E. litoralis, Loktanella fryxellensis) indicating close association with cyanobacteria. The LPOR-containing AAPBs thus seem to dwell in an environment that is commonly rich in picocyanobacteria. It is hence tempting to speculate that aquatic environments, in which environmental conditions such as light availability, oxygen levels, temperature, salinity and pH affect the growth, survival and productivity of the corresponding organisms, have contributed to the transfer of LPOR. The presence of an LPOR would hereby provide its AAPB host with a selective advantage as it would enable the organism to enhance Bchl synthesis under aerobic conditions. This adaptability is also reflected in the broad pH-and temperatureoptima ranges observed for AAPB LPORs (table 1, fig. 3

Identification of LPORs
To search for putative LPORs, we analyzed all publicly available bacterial genomes in GenBank (Clark et al. 2016) (as of July 17, 2018) using the HMMER software (http://hmmer.org/, last accessed September 16, 2020, version 3.1b2) together with an LPOR HMModel from TIGRFAMs (Haft et al. 2003) (all HMModels are listed in supplementary table S2, Supplementary Material online). Additionally, LPORs, whose biochemical properties were previously characterized, were included in the analysis for a comparison. Supplementary table S1, Supplementary Material online, displays the taxonomic information, presence/absence analysis for marker genes as well as accession numbers for all putative LPORs in AAPBs and for the known reference LPORs. Accession numbers for all bacterial LPORs used for phylogenetic tree inference are provided in supplementary table S10, Supplementary Material online.

Bacterial Strains and Culture Conditions
All strains used in this study are listed in supplementary table S3, Supplementary Material online. Escherichia coli strains DH5a (Invitrogen) and S17-1 (Simon et al. 1983) used for cloning and conjugation were grown in lysogeny broth (LB) medium (Carl Roth, Arlesheim, Switzerland) at 37 C, under constant agitation (130 rpm). Heterologous expression of all LPOR encoding genes was performed using E. coli BL21(DE3) employing different media (see below). Antibiotics were added to E. coli culture medium in the following final concentrations (mg ml À1 ): 100 (Ap, ampicillin), 50 (Km, kanamycin). Plasmid DNA was introduced into E. coli using heatshock transformation (Swords 2003 (Klipp et al. 1988) containing 2% (w/v) Select Agar (Thermo Fisher Scientific) or in RCV liquid medium (Weaver et al. 1975) supplemented with 15 mM ammonia at 30 C. Chemoorganotrophic cultivation was carried out in Erlenmeyer flasks filled with 50-ml RCV medium under permanent shaking (130 rpm) in the dark. For photoheterotrophic cultivation, capped air-tight reaction tubes were filled with 15-ml RCV cultivation medium to create an oxygen-free atmosphere. The media were supplemented with 200 mg ml À1 streptomycin for the R. capsulatus wild-type strain B10S, with 200 mg ml À1 streptomycin and 10 mg ml À1 spectinomycin for the DPOR-deficient R. capsulatus DbchB strain (Kaschner et al. 2014) and or with 200 mg ml À1 streptomycin and 25 mg ml À1 kanamycin for the DPOR-deficient R. capsulatus DbchB strains containing the LPOR expression plasmids (see below). The respective cultures were constantly illuminated with bulb light (6 Â 60 W). Plasmids were introduced into the DPOR-deficient R. capsulatus DbchB by conjugational transfer using E. coli S17-1 as donor strain as described before by Klipp et al. (1988). Rhodobacter capsulatus ZY5 (Yang and Bauer 1990), which was used for production and purification of the Pchlide substrates, was cultivated in VN-Medium (10 g/l yeast extract, 5.7 mM K 2 HPO 4 , 2 mM MgSO 4 pH ¼ 7.0) supplemented with 25 mg ml À1 Rifampicin in the dark at 30 C under constant agitation (130 rpm). Production cultures were inoculated to an OD 660nm of 0.01 and grown under microaerobic conditions (culture volume 50% of the flask volume; nonbaffled Erlenmeyer flasks).
After heterologous expression, cells were harvested by centrifugation (30 min, 6,750 Â g, 4 C) and resuspended in buffer (20 mM Tris/HCl, 500 mM NaCl, 20% [w/v] glycerol, pH ¼ 7.5). The corresponding cell suspensions (10% [w/v] wet cells) were disrupted by passing the cell suspension five times through an EmulsiFlex-C5 high-pressure homogenizer (AVESTIN, Ottawa, ON, Canada) at a pressure of 1,000 bar. All LPOR proteins were purified by immobilized metal ion affinity chromatography (IMAC) as described previously (Kaschner et al. 2014). After IMAC, samples were desalted by size exclusion chromatography (SEC) using a Sephadex G25 column (560 ml column volume [CV], XK50/30, GE Healthcare Life Science, VWR International GmbH, Langenfeld, Germany). Protein samples were concentrated to a concentration of at least 1 mg ml À1 using Nanosep Centrifugal Device concentrator (molecular weight cutoff 10,000 Da) (Pall, Germany) and further purified by preparative size-exclusion chromatography using a Superdex 200 (XK16/60, GE Healthcare Life Science, VWR International GmbH) column with 20 mM Tris/HCl buffer pH ¼ 7.5, supplemented with 500 mM NaCl and 20% glycerol, es eluent. Final samples were concentrated, flash frozen in liquid nitrogen, and stored in the same buffer at À20 C until further use.

Pchlide Production and Purification
Pchlide was produced using R. capsulatus ZY5 (Yang and Bauer 1990) in which the bchL gene encoding for one subunit of the DPOR was deleted. This strain therefore accumulates Pchlide, as a mixture of MV-and DV-Pchlide (Heyes et al. 2006). The strain was grown as described above, and the secreted Pchlide was adsorbed to hydrophobic polyurethane cubes (edge length 1 cm), which were added to the cultures during cultivation. After 24-36 h, the cubes were removed and cells washed off with tricine buffer (10 mM tricine pH ¼ 7.5). Subsequently, Pchlide was extracted from the cubes with 100% methanol and filtrated (glass fiber filter and cellulose acetate filter, pore size 0.8 and 0.45 mm). Pchlide was purified by column chromatography using an € AKTAbasic FPLC system (GE Healthcare, Solingen, Germany) using C-18 solidphase extraction (SPE) material (Sep-Pak, Waters, Milford, MA) filled into an ECO PLUS SR TAC15/500LGO-SR-2 column (75 ml CV) (YMC Europe GmbH, Dinslaken, Germany). To facilitate binding, the filtrated Pchlide extract was diluted to a final concentration of 40% (v/v) methanol with tricine buffer. The SPE column was equilibrated with methanol:tricine buffer (40:60-Vol%), the Pchlide extract was loaded, and the column was washed using the same methanol:tricine buffer mixture. To separate carotenoids and other nonwanted pigments from Pchlide, the methanol concentration was increased stepwise to 50% (after two CV) and 60% (after 25 CV). Finally, Pchlide was eluted using a methanol:tricine buffer ratio of 75:25-Vol%. The obtained purified Pchlide eluate was diluted with tricine buffer to a final methanol concentration of $25%. Subsequently, Pchlide was extracted by liquid-liquid extraction using diethyl ether. The resulting Pchlide extract (in diethyl ether) was dried with MgSO 4 , the ether was evaporated using a rotary evaporator Büchi,Flawil,Switzerland), and the dried sample was stored under argon atmosphere at À20 C in the dark.

HPLC-Photodiode Array Detector and MS Analysis of R. capsulatus Purified Pchlide
The Pchlide preparation purified from R. capsulatus ZY5, which consists of a mixture of MV-and DV-Pchlide (Heyes et al. 2006), was analyzed by high-performance liquid chromatography (HPLC) in cooperation with Dr Klaus Bollig at the Shimadzu Laborwelt (Duisburg, Germany). The dried Pchlide preparation was dissolved in 100% methanol. MVand DV-Pchlide were first separated liquid chromatography (LC-10Ai series; Shimadzu Deutschland GmbH, Duisburg, Germany, equipped with a SPDM10Avp photodiode array detector). Chromatographic separation was performed with an analytical C30 column (ISAspher 200-5 C30-CXT, 4.6 mm Â 250 mm, ISERA GmbH, Düren, Germany) using a binary mobile phase (A: 5 mM ammonium acetate buffer pH ¼ 6, 30% methanol, B: 100% methanol) in a gradient program (0-2 min: 5% A, 95% B; 25-45 min: 0% A, 100% B; 45-50 min: 5% A, 95% B). A constant flow rate of 1 ml min À1 was used. Elution was monitored at 440 nm and absorption spectra were recorded for each elution peak. Identification of the eluting Pchlide species was achieved by mass spectrometry (MS) at the Shimadzu Laborwelt (Duisburg, Germany) using an HPLC-coupled hybrid ion trap-time of flight mass spectrometer (LCMS-IT-TOF, Shimadzu, Duisburg, Germany). Electrospray ionization was used and the resulting ions were further fragmented in the ion trap using Argon as collision gas. The ion accumulation time in the octopole was 20 ms (MS 1 mode) and 40 ms (MS 2 mode). Mass spectra were acquired in positive ionization mode in the range of 150-1,000 m/z.

Purification of MV-and DV-Pchlide
Pchlide was prepared from R. capsulatus ZY5 by solid-state extraction as described above. MV-and DV-Pchlide were separated using a preparative C30 HPLC column (ISAspher 200-5 C30-CXT, 20 mm Â 250 mm, ISERA GmbH). An € AKTAbasic FPLC system (GE Healthcare Life Science, Freiburg, Germany) was adapted for the use of organic solvents by employing polyether ether ketone fittings and tubes. The system was equipped with a column oven (Gynkotek STH 585,Gynkotek,Dionex,Thermo Fisher Scientific GmbH). which was set to 35 C. Per run 8 ml of the R. capuslatus ZYderived Pchlide, dissolved in 100% methanol, was loaded onto the column. Separation of MV-and DV-Pchlide was achieved using a binary mobile phase (A: 5 mM ammonium acetate buffer pH ¼ 6, 30% methanol, B: 100% methanol) by employing a gradient program (0-2 min: 5% A, 95% B; 25-45 min: 0% A, 100% B; 45-50 min: 5% A, 95% B) at a constant flow rate of 15 ml min À1 . Elution was monitored by continuously measuring the absorbance of the soret band of MV-and DV-Pchlide at 450 nm. The MV-and DV-Pchlide-containing fractions were pooled, diluted to a final methanol concentration of <25% with 10 mM tricine buffer (pH ¼ 7.5), and extracted by liquid-liquid extraction into dry diethyl ether. The resulting MV-and DV-Pchlide extracts (in diethyl ether) were dried with MgSO 4 . The diethyl ether was subsequently evaporated using a rotary evaporator (Rotavapor R-100, Büchi), and the dried MV-and DV-Pchlide preparations were stored in the dark at À20 C under argon atmosphere until further use.

In Vitro LPOR Activity Assays
General Assay Setup All light-dependent activity measurements were performed as previously described (Kaschner et al. 2014). In brief, protein samples (0.17 mM) or cell free lysates were diluted with assay buffer (20 mM Tris/HCl buffer [pH ¼ 7.5] supplemented with 500 mM NaCl and 20% [v/v] glycerol, 160 mM NADPH, 70 mM DTT, and 0.03% [v/v] Triton X-100) in half-micro disposable cuvettes (1 cm light path). The purified and dried Pchlide substrate was dissolved in 100% methanol and added to the protein (lysate) buffer mixture to a final concentration of 3.5 mM (5% [v/v] methanol in the assay). Subsequently, the assay mixture was equilibrated for 5 min at 25 C. A blue-light Chernomor et al. . doi:10.1093/molbev/msaa234 MBE emitting LED (450 nm; 2.6 mW cm À2 ) was mounted on top of the cuvette, and light-dependent Pchlide turnover was achieved by illuminating the assay mixture employing cycles of 1-s blue-light illumination followed by 11 s in the dark during which an absorption spectrum from 620 to 720 nm was recorded. Weakly active samples such as those with AtLPORC were illuminated by cycles of 6 s of illumination followed by 12 s of darkness. Pulsed illumination was achieved by using a microcontroller-controlled LED driver (Arduino UNO, Smart Projects, Italy). Data were analyzed using a home-written shell script, which filters and removes spectra that contain illumination events. LPOR activity was quantified by linear regression on the initial linear rise in absorbance of the Chlide product (672 nm) that corresponds to the initial reaction velocity. Chlide formation was quantified using molar extinction coefficient of e 672nm ¼ 69,950 M À1 cm À1 (Klement et al. 1999;Heyes et al. 2000). One unit (U) of LPOR activity was defined as the amount of enzyme which reduces 1 mmol Pchlide to Chlide per minute under the given reaction conditions. pH and Temperature Optima All assays were carried and analyzed out as described for the general assay setup. Three different buffer systems (pH ¼ 5.5-7.5: 200 mM sodium phosphate buffer; pH ¼ 7.5-9.0: 200 mM Tris/HCl; pH ¼ 9.0-10.0: 200 mM glycine buffer; all three buffers were supplemented with 500 mM NaCl and 20% [v/v] glycerol) were used to cover the pH range between pH ¼ 5.5 and 10. LPOR temperature optima were determined for the temperature range from 10 to 55 C in 5 C increments. To avoid temperature-dependent pH changes during the measurement at elevated temperatures, the pH of the assay was adjusted at the respective temperature. Samples were equilibrated at the desired temperature in disposable halfmicro cuvettes (1 cm light path) using a cuvette-holder equipped with a Peltier-based thermostat. The accuracy of the temperature provided by the thermostat was determined by recording the actual temperature from a buffer solution.

MV-/DV-Pchlide Substrate Preference
MV-and DV-Pchlide were prepared as described in section "Purification of MV-and DV-Pchlide," and the dried powder was dissolved in 100% methanol to yield a stock solution containing at least 140 mM of the respective Pchlide species. All measurements were performed using the general assay setup as described above by using purified MV-and DV-Pchlide at a fixed concentration of 3.5 (60.15) mM. The data were analyzed as described above.
In Vivo LPOR Activity Assay The in vivo LPOR activity assay was performed as described previously (Kaschner et al. 2014). To comparatively analyze the activity of chosen LPORs in vivo, the R. capsulatus LPOR expression strain (i.e., DPOR-deficient DbchB strain carrying the respective LPOR-encoding pRhokHi-2 derivatives, supplementary table S3, Supplementary Material online) was used to characterize its ability 1) to grow photoheterotrophically and 2) to synthesize BChl a (bacteriochlorophyll a). Increasing cell densities and BChl a-mediated light absorption were used as a measure for LPOR in vivo activity. To ensure that all expression cultures were in the same growth phase for the LPOR in vivo assay, R. capsulatus expression strains were precultivated twice under aerobic chemoheterotrophic conditions in RCV liquid medium (second culture starting with OD 660nm ¼ 0.1). Cells from second precultures were used after 72 h of cultivation to inoculate the corresponding test cultures in RCV starting with an initial cell density of OD 660nm ¼ 0.1. The test cultures were grown under photoheterotrophic conditions (i.e., in the absence of oxygen and under constant illumination). To monitor LPOR activities over time, increase of cell density was analyzed for 10 days at 660 nm and BChl a formation was investigated by recording whole-cell absorption spectra of 100-ml samples between 300 and 900 nm (using a TECAN Infinite plate reader) that were normalized to a cell density of OD 660 nm ¼ 1. In addition, the color phenotypes of all test cultures were photodocumented after 240 h of cultivation. Images were taken under the same light conditions with identical camera settings including white balance.

Determination of the Dissociation Constant of the LPOR/NADPH/Pchlide Ternary Holoprotein Complex
The dissociation constant, K d , of Pchlide binding to the ternary LPOR/NADPH/Pchlide complex was determined using a spectral method similarly to previous studies on cryptochromes (Kutta et al. 2017) and CarH photoreceptors (Kutta et al. 2015). Samples were prepared containing different concentrations of LPOR (0-200 mM) apo-protein, a constant concentration of NADPH (160 mM) and a constant concentration of Pchlide (2 lM) in buffers with and without 70 mM DTT containing 20 mM Tris/HCl, 500 mM NaCl, 20 Vol% Glycerol, and Triton X-100 (0.03 Vol%). An UV-Vis absorption spectrum over 1 cm pathlength was acquired for each concentration of apo-protein used giving the spectral evolution from the unbound Pchlide spectrum into the almost fully bound Pchlide spectrum with the typical redshift of the Q y absorption band when bound to the LPOR protein pocket. This data matrix can be decomposed into its principle species spectra and corresponding mole fraction profiles. The mole fractions were determined by fitting the pure spectra of unbound and bound Pchlide to each single mixed spectrum. As the bound Pchlide spectrum is a priori unknown, it is initially estimated by the last spectrum (highest concentration of apo-protein) in the spectral sequence. Each mixed spectrum also contains an absorptive contribution due to scattered light from the sample (especially at high concentrations of apo-protein), which was accounted for by a general scatter function of the following form: where k is the wavelength, y 0 accounts for the offset, a scales the curve, and n determines the curvature of the scatter function. n was fixed to 2 representing a reasonable curvature Complex Evolution of LPORs in Aerobic Anoxygenic Phototrophs . doi:10.1093/molbev/msaa234 MBE of the scatter contribution. The mole fraction profiles show a typical binding curve of the following chemical equilibrium: where it is assumed NADPH binding to LPOR-apo has a much higher affinity so that the complex LPOR/NADPH can be seen as one molecule. The dissociation constant is defined as: The amount of complex AB, x AB , is dependent on ½A and ½B, the concentration of the two binding partners, and is given by: A ½ is the independent variable representing the concentration of the apo-protein, ½B corresponds to the 2 lM fixed Pchlide concentration, and S 0 and S max represent the minimal and maximal limits of the binding curve, respectively. The mole fractions for unbound and in LPOR bound Pchlide were globally least squared fitted to a common K d using the Levenberg-Marquardt algorithm. The first global fit was used to determine the contamination of the unbound Pchlide spectrum in the a priori assumed bound Pchlide spectrum. Then, the initial guess spectrum and the mole fractions were corrected correspondingly yielding the pure in LPOR spectrum as well as mole fractions ranging from 0 to 1 as expected.

Temperature-dependent Unfolding and Thermostability
The temperature-dependent unfolding and the thermostability of LPORs were determined using differential scanning fluorimetry (DSF), which detects the unfolding of proteins by monitoring the temperature-dependent changes in the fluorescence of the aromatic amino acids of proteins. For Nano-DSF measurements, a Prometheus NT.Flex (NanoTemper Technologies GmbH, Munich, Germany) instrument was used. Purified LPOR samples (10 ml) with a concentration of 0.6 mg ml À1 were subjected to a linear unfolding ramp (0.5 C min À1 , from 15 to 85 C). The intrinsic tryptophan fluorescence of the protein was monitored continuously (18 data points per minute) at 350 and 330 nm. Unfolding transition midpoints were determined from the first derivative of the fluorescence ratio (F350/F330) by using the RT.ThermControl Software (NanoTemper Technologies GmbH). Thermostabilities were measured at different temperatures for 30 h by using the RT.TimeControl Software (NanoTemper Technologies GmbH).

SAXS Measurements and Data Analysis
SAXS experiments were performed at beamline BM29 (Pernot et al. 2013) at the European Synchrotron Radiation Facility (ESRF, Grenoble, France) using 12.5 keV X-ray radiation with a wavelength of 0.9919 Å. All measurements were carried out at 10 C. For each LPOR protein, four samples with concentrations of $0.5, 1, 3, and 5 mg ml À1 were measured in 20 mM Tris/HCl buffer pH ¼ 7.5 supplemented with 500 mM NaCl and 20% (v/v) glycerol. The exact protein concentrations of the measured LPOR samples are listed in supplementary table S14, Supplementary Material online. The samples were continuously purged through a 1-mm quartz capillary at a flow rate of 2.3 ml s À1 . The buffer reference was measured before and after each protein sample. For each sample/reference ten frames with an exposure time of 3 s each were recorded. Frames without radiation damage were merged. The data were scaled by the protein concentration and extrapolated to infinite dilution. Scattering data were analyzed employing the ATSAS software package (Petoukhov et al. 2012). Data obtained for low and high concentration samples were merged. Lower concentration data were used for the smaller q-range, whereas the data at higher concentration were used for the high q-range. SAXS data were inspected visually for the presence of aggregation based on the Guinier-Plot. For each sample, a low and high concentration data set was merged (see supplementary table S14, Supplementary Material online, for details). The Porod volume, calculated with the program DATAPOROD, was used to estimate the molecular mass of the scattering particle, by using a division factor of 1.7 (Petoukhov et al. 2012). The distance distribution function P(r) was determined using the program DATGNOM. Ab initio models were built employing the programs GASBORP (Svergun et al. 2001), DAMMIF (Franke and Svergun 2009), and DAMMIN (Svergun 1999). For each LPOR data set 20 ab initio models were generated, which were subsequently averaged and filtered using the program DAMAVER (Volkov and Svergun 2003). For the corresponding filtered models, the envelope function was determined using the SITUS package (Wriggers 2012). The molecular mass was estimated from the excluded volume of the filtered model by dividing the respective values by 2 (Petoukhov et al. 2012). Theoretical scattering curves for the available LPOR homology models were calculated and fitted to the experimental SAXS data using the program CRYSOL (Svergun et al. 1995). LPOR homology models were built with YASARA Structure Version 16.6.24 Vriend 2014, 2015)