*Alphaproteobacteria*

The phototrophic *Alphaproteobacteria* formed the most fragmented and diverse array of groups in the PS tree with representatives of the six orders *Rhodospirillales, Rhizobiales, Sphingomonadales, Rhodobacterales, Caulobacterales,* and *Rhodothalassiales*. Most significantly, the *Rhodobacterales,* together with *Sphingomonadales* and *Brevundimonas* (*Caulobacterales*), formed a major branch, according to BchXYZ-PufHLM, which was clearly distinct from all other phototrophic *Alphaproteobacteria* (Figures 1, 3 and 4). A deep branching point separated the *Sphingomonadales* and *Brevundimonas* from the *Rhodobacterales*. The relations of other *Alphaproteobacteria*, however, were more problematic because most of the species had long-distance lines with deep branching points and only a few species arranged in stable groups that could be recognized in both PS tree and RNA tree.

*Rhodobacterales.* The most recent and shallow divergences were seen in the phylogeny of the *Rhodobacterales*, which, in contrast to most other phototrophic *Alphaproteobacteria,* appeared as a young group that had differentiated later than others and was well established as a group in PS tree and RNA tree. It diversified quite fast in evolutionary terms and now represents one of the largest orders of phototrophic bacteria known. The following groups of *Rhodobacterales* were formed in the PS tree. With the exception of the *Rhodobacter* group and the *Rhodovulum* group they represent aerobic phototrophic bacteria.


*Sphingomonadales.* The *Sphingomonadales* formed a consistent lineage of aerobic phototrophic bacteria within all considered phylogenetic trees. *Sphingomonadaceae* with *Sphingomonas* and *Novosphingobium* species (likely also *Blastomonas,* see [3]) were forming one sub-branch and the *Erythrobacteraceae* with *Erythrobacter* and *Porphyrobacter* species (likely also *Erythromicrobium,* see [3]) a second one. There was support for the inclusion of *Erythrobacter marinus* into the *Sphingomonas* group from BchXYZ and BchXYZ-PufHLM phylogeny. In addition, *Erythrobacter marinus* contained PufC like *Sphingomonas* and *Novosphingobium* species but unlike other *Erythrobacteraceae*. According to the RNA tree, *Erythrobacter marinus* clustered with other *Erythrobacter* species, however, with low confidence in its position.

*Brevundimonas (Caulobacterales). Brevundimonas subvibrioides* represented an aerobic phototrophic bacterium, which clearly but distantly was linked to the *Sphingomonadales* branch according to the PS tree and RNA tree. *Brevundimonas* lacked PufC as the *Erythrobacteraceae* did. The deep branching point of *Brevundimonas* in the PS tree indicated that it was closest to the common ancestor of this branch.

The *Rhodobium*/*Hoeflea* group. A most deeply branching stable lineage in the PS tree was found within the radiation of the *Alphaproteobacteria* and was represented by the *Rhodobium*/*Hoeflea* group with *Rhodobium orientis*, *Hoeflea phototrophica*, *Labrenzia alexandrii,* and *Oceanibaculum indicum* (Figures 2 and 3). Despite the formation of a coherent group according to the PS tree, the species had different, though unsupported positions in the RNA tree (Figures 1 and 4). According to 16S rRNA, *Hoeflea phototrophica* (*Rhizobiales*, *Phylobacteriaceae*) had a deeply branching unsupported position; *Labrenzia alexandrii* (*Rhodobacterales*, *Rhodobacteraceae*) also had an unsupported position that was linked at the basis to *Brevundimonas* and the *Spingomonadales*; *Rhodobium orientis (Rhizobiales, Rhodobiaceae)* was found together with *Afifella* in a poorly rooted distinct branch; *Oceanibaculum indicum* (*Rhodospirillales, Rhodospirillaceae*) appeared distantly associated with *Rhodocista*, *Skermanella,* and the *Acetobacteraceae* (Figure 1). The photosynthesis of the *Rhodobium*/*Hoeflea* group represented one of the most ancient lines among the purple bacteria, and the most recent divergence (between *Labrenzia* and *Oceanibaculum*) was rooted much deeper as the basic divergence of the *Rhodobacterales* branch (Figure 3). In addition, there is no close relative to the photosynthesis system among other known phototrophic bacteria, which is a clear indication of the very ancient origin of photosynthesis in this lineage of phototrophic bacteria. If we do trust the phylogenetic reliability of the 16S rRNA system, we should assume quite early genetic transfers of major parts or the complete photosystem from an ancient ancestor within the *Rhodobium* lineage to the other bacteria. Alternatively, as the species of this branch formed poorly rooted lines in the RNA tree, the differences between PS and RNA tree might be explained by unresolved relationships and not correctly rooted positions of these bacteria in the RNA tree.

The *Rhodopseudomonas*/*Bradyrhizobium* group. In the PS tree, the *Rhodopseudomonas*/*Bradyrhizobium* group formed one of the most deeply branching lines distinct from other *Rhizobiales*. Both *Rhodopseudomonas* and *Bradyrhizobium* lacked PufC. According to 16S rRNA phylogeny, *Rhodopseudomonas* and *Bradyrhizobium* formed a sister branch to the photosynthetic *Methylobacterium* species, distant to other *Rhizobiales* (*Blastochloris* and *Rhodoplanes*, *Methylocella* and *Rhodoblastus, Prosthecomicrobium* and *Rhodomicrobium*).

The *Rhodopila* group. Another distinct group was represented by the *Acetobacteraceae* and supported by both RNA tree and PS tree with species of *Rhodopila*, *Acidiphilum*, *Paracraurococcus,* and *Rubritepida*.

The *Rhodospirillum* group. According to the PS tree, species of *Rhodospirillum*, *Pararhodospirillum,* and *Roseospirillum parvum* formed a group to which *Rhodospira trueperi* appeared distantly linked. In the RNA tree, *Caenispirillum* was included in this group, while in the PS tree, it had a separate position and formed a branch together with *Rhodovibrio* species, which, in turn, appeared as an isolated line at the basis of the *Alphaproteobacteria* within the RNA tree.

In addition to these groups, several separate lineages were represented by single genera of *Fulvimarina, Rhodothalassium*, *Prosthecomicrobium,* and *Afifella* in both PS and RNA trees (Figures 1 and 3). Thus, their phylogenetic positions remained unclear. While *Methylocella* specifically associated with *Rhodoblastus* in both RNA tree and PS trees, the following groupings were not well supported or had different positions in PS tree and RNA tree:


#### *3.4. Distribution of PufC*

The cytochrome associated with the photosynthetic reaction center is an important component in many of the PS-II type photosynthetic bacteria. As a more peripheral part of the photosynthetic reaction center, the cytochrome may be more easily replaced by alternative electron transport systems, and this obviously happened in a number of phylogenetic lineages of the *Alphaproteobacteria* and *Betaproteobacteria* (Supplementary Table S1). The general presence of the reaction center cytochrome PufC in phototrophic purple bacteria and the absence in quite a few distinct groups of the *Alphaproteobacteria* and *Betaproteobacteria* strongly suggested that PufC independently has been lost several times. PufC was absent in*Rhodoferax fermentans*(but present in the related*Rhodoferax antarcticus*), in*Rhodospirillum rubrum* (but present in the related *Pararhodospirillum photometricum*), in *Bradyrhizobium* and *Rhodopseudomonas* species (but present in other *Rhizobiales*), in *Brevundimonas*, *Porphyrobacter,* and *Erythrobacter* species (but present in *Erythrobacter marinus* and *Sphingomonadaceae*). It was also absent in one of the major *Rhodobacterales* branches of the PS tree, including the *Rhodobacter*/*Rhodobaca* group and the *Loktanella*/*Sulfitobacter* group. The presence/absence of PufC in species of the *Alphaproteobacteria* was congruen<sup>t</sup> with the photosynthesis phylogeny. The presence of PufC supported the inclusion of *Rhodobaculum* into the *Rhodovulum* group, and its absence in *Roseisalinus* was in accord with its inclusion into the *Loktanella*/*Sulfitobacter* group according to the PS tree. Following the PS tree and the presence of PufC, *Erythrobacter marinus* fitted into the *Sphingomonadaceae* (rather than into the *Erythrobacteraceae*).

**Figure 4.** Linear comparison tree of 16S rRNA gene sequences (right) and BchXYZ-PufHLM sequences (left) of phototrophic bacteria.

#### *3.5. Phylogenetic Aspects of Aerobic Anoxygenic Photosynthesis*

As oxygen was absent from the earth during the first billion years of life, in which the basic concepts of photosynthesis are expected to have evolved and as oxygenic photosynthesis using two di fferent consecutive photoreactions is considered a late event in the evolution of photosynthesis, the roots of anoxygenic, as well as oxygenic photosynthesis, are to be found in the ancient anoxic environments [30,31]. We assume that with the onset of oxygenic photosynthesis, the basic concepts of photosynthesis, as known today, have already been established. In fact, oxygen evolution by photosynthesis using water as electron source is dependent on the use of two di fferent consecutive photosynthetic reactions that have evolutionary ancestors among anoxygenic phototrophic bacteria: a type-I photosynthesis in ancestors of *Heliobacteria*, *Chlorobi*, and *Chloracidobacterium* and a type-II photosynthesis in ancestors of *Chloroflexi* and all phototrophic purple bacteria [31].

The appearance of oxygenic photosynthesis approx. 3 billion years ago was a revolution in ecology. It drastically changed the environmental conditions on earth, and over approx. 2 billion of years caused the gradual increase of the atmospheric oxygen content to the actual level [32]. Quite likely, during this transition period, the radiation of the purple bacteria diverged to its full extension.

During adaptation to oxic conditions, quite a number of anoxygenic phototrophic purple bacteria may have gained the ability to perform under both anoxic and oxic conditions by maintaining the strict regulation of biosynthesis of the photosynthetic apparatus and its repression by oxygen. An example of such bacteria is found in *Rhodobacter* species performing anoxygenic photosynthesis under anaerobic conditions in the light and aerobic respiration under oxic conditions in the dark [33–36]. During further evolution, some of these phototrophic bacteria may have lost the ability to build up the photosynthetic apparatus in the absence of oxygen and, in contrast, required oxygen for the formation of the photosynthetic apparatus [37]. In bacteria, such as the anaerobic phototrophic *Rhodobacterales*, that have already been adapted to arrange themselves with certain levels of oxygen, this could have been a small step in modifying the oxygen-response in the formation of the photosynthetic apparatus. As a result, in various phylogenetic branches, aerobic anoxygenic phototrophic bacteria have evolved, of which *Erythrobacter* and *Roseobacter* species at present are the most well-known examples [37–39].

Aerobic representatives of phototrophic purple bacteria that require oxygen for the synthesis of bacteriochlorophyll and the photosynthetic apparatus were found in a number of well-defined groups. The *Haliaceae* of the *Gammaproteobacteria* and the *Erythrobacteraceae* and *Sphingomonadaceae* of the *Sphingomonadales* at present exclusively contain aerobic representatives. In addition, isolated lines of single representatives of aerobic phototrophic bacteria were found with species of *Fulvimarina*, *Gemmatimonas, Polynucleobacter,* and *Brevundimonas*. In the *Rhodobium*/*Hoeflea* group, several aerobic phototrophic bacteria joined the anaerobic phototrophic *Rhodobium*. In the *Acetobacteraceae,* several aerobic representatives were found together with the anaerobic phototrophic *Rhodopila*. A larger group of the aerobic phototrophic *Rhizobiales* with *Methylobacterium*, *Methylocella*, *Prosthecomicrobium,* and *Bradyrhizobium* was related to the anaerobic phototrophic *Rhodopseudomonas* and *Rhodoblastus* species. Among the *Rhodobacterales,* the groups around *Loktanella*/*Sulfitobacter*, *Dinoroseobacter*/*Jannaschia,* and *Roseivivax*/*Roseovarius* represented branches with aerobic species.

With the exception of the aerobic phototrophic *Rhodobacterales*, most of the aerobic phototrophic bacteria represented ancient phylogenetic lineages. This was especially the case for the *Sphingomonadales*, *Brevundimonas* (*Caulobacterales*), the *Haliaceae,* and those within the *Rhodobium*/*Hoeflea* group. As traces or small levels of oxygen already were present at the time when anaerobic phototrophic purple bacteria presumably diversified approx. 2.5 billion years ago, the aerobic phototrophic purple bacteria could have developed in parallel, which would explain the deep divergence of some of the lineages of aerobic phototrophic *Proteobacteria*.

The pattern of distribution of aerobic phototrophic bacteria among the phototrophic purple bacteria strongly suggested that aerobic phototrophic purple bacteria evolved from anaerobic ancestors in independent and multiple events. The deep branching points of some lineages indicated their early divergence from the anaerobic phototrophic ancestors. The phylogeny suggested that in the

*Rhodobium*/*Hoeflea* group, photosynthesis of aerobic representatives evolved from an anaerobic ancestor with a common root with *Rhodobium*. Known representatives of aerobic phototrophic *Sphingomonadales*, *Cellvibrionales (Haliaceae*), *Caulobacterales* (*Brevundimonas*), and *Gemmatimonas* could be present-day survivors of ancient anaerobic phototrophic relatives not known so far or extinct.

The development of aerobic phototrophic *Rhodobacterales* was considered to be a more recent event, as these bacteria and their photosynthesis were much younger compared to most other phototrophic bacteria (Figure 3). It has been calculated by molecular clock calculations using sequences of representative genes that the divergence of the last common ancestor of *Roseobacter* and *Rhodobacter* dates to approx. 900 Myr ago (+/- 200 Myr) [40]. At that time, the oxygen content of the earth's atmosphere almost had reached present-day levels [32], and it is tempting to assume that aerobic phototrophic lineages branched off from anaerobic phototrophic *Rhodobacterales* under these conditions. Though this is considered the more likely scenario, alternatively, the common ancestor of the *Rhodobacterales* could have been an aerobic phototrophic bacterium. This scenario could find support in the common roots of *Rhodobacterales* with the aerobic phototrophic *Sphingomonadales* and *Brevundimonas* but implies that the ancestors of *Sphingomonadales* and *Brevundimonas* also were aerobic phototrophic bacteria, which is not necessarily the case. It would also imply that aerobic phototrophic *Rhodobacterales* transformed back to perform anaerobic photosynthesis, which from an evolutionary and ecological perspective appears quite unlikely. Therefore, we assume that aerobic phototrophic representatives of *Rhodobacterales*, *Sphingomonadsales,* and *Brevundimonas* evolved independently from anaerobic phototrophic ancestors.
