Homologous Proteins of the Manganese Transporter PAM71 Are Localized in the Golgi Apparatus and Endoplasmic Reticulum

Chloroplast manganese transporter 1 (CMT1) and photosynthesis-affected mutant 71 (PAM71) are two membrane proteins that function sequentially to mediate the passage of manganese across the chloroplast envelope and the thylakoid membrane. CMT1 and PAM71 belong to a small five-member protein family in Arabidopsis thaliana. The other three, photosynthesis-affected mutant 71 like 3 (PML3), PML4 and PML5 are not predicted to reside in chloroplast membranes. In this study, the subcellular localization of PML3:GFP, PML4:GFP and PML5:GFP was determined using transient and stable expression assays. PML3:GFP localizes to the Golgi apparatus, whereas PML4:GFP and PML5:GFP are found in the endoplasmic reticulum. We also examined patterns of PML3, PML4 and PML5 promoter activity. Although the precise expression pattern of each promoter was unique, all three genes were expressed in the leaf vasculature and in roots. Greenhouse grown single mutants pml3, pml4, pml5 and the pml4/pml5 double mutant did not exhibit growth defects, however an inspection of the root growth revealed a difference between pml3 and the other genotypes, including wild-type, in 500 µM manganese growth conditions. Strikingly, overexpression of PML3 resulted in a stunted growth phenotype. Putative functions of PML3, PML4 and PML5 are discussed in light of what is known about PAM71 and CMT1.


Introduction
Plants, like all other living organisms, require manganese (Mn) as an activator of enzymes or as an integral component of protein complexes. The translocation of Mn from the soil into the aerial parts of plants requires the concerted action of a number of Mn transporters, which mediate uptake of Mn into root cells and facilitate subsequent Mn translocation from root to shoot [1,2]. At the cellular level, stably bound Mn is found in various locations, including the photosystem II (PSII) in chloroplasts, oxalate oxidase in cell walls, and Mn superoxide dismutase in mitochondria. Mn is also present in the Golgi apparatus (e.g., glycosyltransferases are activated by Mn 2+ ), in the endoplasmic reticulum and in the vacuole. Indeed, the vacuole serves as an intracellular sink when Mn is in excess and as a source when the Mn supply is limited [3]. The most prominent role of Mn is its involvement in the oxygen-evolving complex of PSII, which splits water into oxygen, protons and electrons. The electrons released from water in PSII are eventually transferred to NADP + via the cytochrome b 6 /f complex, plastocyanin, photosystem I and ferredoxin. Detailed analysis over the past several years has elucidated the structural basis for water oxidation and oxygen evolution in PSII by the catalytic Mn 4 CaO 5 cluster [4][5][6][7]. Protein sequence alignment of chloroplast manganese transporter 1 (CMT1), photosynthesis-affected mutant 71 (PAM71) and related proteins in Arabidopsis. Annotations were made as described previously [10]. Black and grey boxes indicate identical residues and conservative exchanges. Six putative transmembrane domains are indicated from TM1 to TM6, with the putative additional TM domain of CMT1 and PAM71 indicated by a dashed line. The conserved motifs E-x-G-D-(KR)-(TS) are highlighted by red lines, and the central loops are indicated by a dotted line. The putative chloroplast transit peptides are indicated by green letters and the putative secretory pathway transit peptide is indicated by magenta letters.

PML3 Localizes to the Golgi, and PML4 and PML5 Are Found in the Endoplasmic Reticulum
In order to determine the subcellular localization of PML3, PML4 and PML5, the open reading frames of the respective cDNAs were amplified and subsequently cloned upstream of the GFP reporter gene and downstream of the 35S promotor. The resulting constructs were then transiently expressed in Nicotiana benthamiana leaves ( Figure 2 and Figure S1). In none of these cases was the GFP signal found to overlap with chlorophyll autofluorescence (Figure 2). Thus, a chloroplast localization for PML3, PML4 and PML5 can be excluded, as already suggested by the in silico analysis (Figure 1). The pattern of PML3:GFP fluorescence displayed predominantly dot-like signals resembling those of the fluorescent Golgi marker GmMAN1:mCherry [26]. Indeed, when simultaneously expressed, fluorescence signals from both proteins PML3:GFP and GmMAN1:mCherry overlapped, thus indicating that PML3:GFP is targeted to the Golgi apparatus ( Figure 2A, Figure S1A). We also determined the subcellular localization of PML4:GFP and PML5:GFP. The GFP fluorescence of both protein fusions was distributed in a pattern resembling that of the ER marker AtWAK2:mCherry [26] and coincided exactly with the red fluorescence of this marker ( Figure 2B,C, Figure S1B), indicating that both proteins are targeted to the endoplasmic reticulum. We verified that the GFP fluorescence of PML5:GFP did not overlap with the red fluorescence of GmMAN:mCherry, effectively excluding the possibility of a dual localization ( Figure S1C). Protein sequence alignment of chloroplast manganese transporter 1 (CMT1), photosynthesisaffected mutant 71 (PAM71) and related proteins in Arabidopsis. Annotations were made as described previously [10]. Black and grey boxes indicate identical residues and conservative exchanges. Six putative transmembrane domains are indicated from TM1 to TM6, with the putative additional TM domain of CMT1 and PAM71 indicated by a dashed line. The conserved motifs E-x-G-D-(KR)-(TS) are highlighted by red lines, and the central loops are indicated by a dotted line. The putative chloroplast transit peptides are indicated by green letters and the putative secretory pathway transit peptide is indicated by magenta letters.

PML3 Localizes to the Golgi, and PML4 and PML5 Are Found in the Endoplasmic Reticulum
In order to determine the subcellular localization of PML3, PML4 and PML5, the open reading frames of the respective cDNAs were amplified and subsequently cloned upstream of the GFP reporter gene and downstream of the 35S promotor. The resulting constructs were then transiently expressed in Nicotiana benthamiana leaves ( Figure 2 and Figure S1). In none of these cases was the GFP signal found to overlap with chlorophyll autofluorescence (Figure 2). Thus, a chloroplast localization for PML3, PML4 and PML5 can be excluded, as already suggested by the in silico analysis (Figure 1). The pattern of PML3:GFP fluorescence displayed predominantly dot-like signals resembling those of the fluorescent Golgi marker GmMAN1:mCherry [26]. Indeed, when simultaneously expressed, fluorescence signals from both proteins PML3:GFP and GmMAN1:mCherry overlapped, thus indicating that PML3:GFP is targeted to the Golgi apparatus ( Figure 2A, Figure S1A). We also determined the subcellular localization of PML4:GFP and PML5:GFP. The GFP fluorescence of both protein fusions was distributed in a pattern resembling that of the ER marker AtWAK2:mCherry [26] and coincided exactly with the red fluorescence of this marker ( Figure 2B,C, Figure S1B), indicating that both proteins are targeted to the endoplasmic reticulum. We verified that the GFP fluorescence of PML5:GFP did not overlap with the red fluorescence of GmMAN:mCherry, effectively excluding the possibility of a dual localization ( Figure S1C).
To further confirm these findings, we generated stably transformed Arabidopsis lines expressing PML3:GFP, PML4:GFP and PML5:GFP under the control of the 35S promoter and named these individual lines Pro35S::PML3:GFP #1, Pro35S::PML4:GFP #1 and Pro35S::PML5:GFP #1, respectively ( Figure S2). Microsomal fractions from leaves of the three lines were subjected to centrifugation on sucrose-density gradients ( Figure 2D-F). Fractions 6 to 22 were subjected to immunoblot analysis employing three antibodies, including the GFP antibody. Antibodies against sterol methyltransferase 1 (SMT1), an integral membrane protein specific to the endoplasmic reticulum [27], and ADP-ribosylation factor 1 (ARF1), a protein that is found in both the Golgi and the trans-Golgi network [28], were used to detect endogenous proteins of the respective compartment ( Figure 2D,F). PML3:GFP was found in the same fractions as the ARF1 signal ( Figure 2D), while the GFP fusions expressed in the lines Pro35S::PML4:GFP and Pro35S::PML5:GFP were both found to comigrate with the SMT1 protein ( Figure 2E,F). These observations confirm the localization of PML3 to the Golgi apparatus and the assignments of PML4 and PML5 to the endoplasmic reticulum. To further confirm these findings, we generated stably transformed Arabidopsis lines expressing PML3:GFP, PML4:GFP and PML5:GFP under the control of the 35S promoter and named these individual lines Pro35S::PML3:GFP #1, Pro35S::PML4:GFP #1 and Pro35S::PML5:GFP #1, respectively ( Figure S2). Microsomal fractions from leaves of the three lines were subjected to centrifugation on sucrose-density gradients ( Figure 2D-F). Fractions 6 to 22 were subjected to immunoblot analysis employing three antibodies, including the GFP antibody. Antibodies against sterol methyltransferase 1 (SMT1), an integral membrane protein specific to the endoplasmic reticulum [27], and ADPribosylation factor 1 (ARF1), a protein that is found in both the Golgi and the trans-Golgi network [28], were used to detect endogenous proteins of the respective compartment ( Figure 2D,F). PML3:GFP was found in the same fractions as the ARF1 signal ( Figure 2D), while the GFP fusions expressed in the lines Pro35S::PML4:GFP and Pro35S::PML5:GFP were both found to comigrate with the SMT1 protein ( Figure 2E,F). These observations confirm the localization of PML3 to the Golgi apparatus and the assignments of PML4 and PML5 to the endoplasmic reticulum.

Transgenic Arabidopsis Lines Expressing PML3:GFP Show a Stunted Growth Phenotype
The stably transformed Arabidopsis lines Pro35S::PML4:GFP #1 and Pro35S::PML5:GFP #1 possessed a wild-type-like phenotype ( Figure 3A), however Pro35S:PML3:GFP #1 showed an interesting leaf phenotype. The rosette of Pro35S::PML3:GFP #1 was much smaller than a wild-type rosette of the same developmental stage ( Figure 3B), and individual leaves exhibited a prominent constraint and bent morphology ( Figure 3C,D). A rather simple explanation for this growth phenotype could be that the transgene caused a mutation by its random insertion into the genome of Arabidopsis. To exclude this possibility, we searched for independent transgenic Pro35S::PML3:GFP lines that exhibited GFP expression. To this end, we isolated three further lines, named Pro35S::PML3:GFP #2, #3, and #4 ( Figure  S2). All lines were grown at the same time and visually inspected for their leaf phenotype. It turned out that they were smaller than wild-type of the same age; thus, it is likely that expression of PML3:GFP evokes the stunted growth phenotype. Moreover, two lines, Pro35S::PML3:GFP #2 and #3, displayed a similar curvature of individual leaves in agreement with observations made in Pro35S::PML3:GFP #1 ( Figure 3). It should be noted that not all leaves are equally affected and therefore it is maybe not surprising that Pro35S::PML3:GFP #4 did not show curvature of individual leaves ( Figure 3). inspected for their leaf phenotype. It turned out that they were smaller than wild-type of the same age; thus, it is likely that expression of PML3:GFP evokes the stunted growth phenotype. Moreover, two lines, Pro35S::PML3:GFP #2 and #3, displayed a similar curvature of individual leaves in agreement with observations made in Pro35S::PML3:GFP #1 ( Figure 3). It should be noted that not all leaves are equally affected and therefore it is maybe not surprising that Pro35S::PML3:GFP #4 did not show curvature of individual leaves ( Figure 3).

Expression Patterns of PML3, PML4 and PML5 in Arabidopsis Tissues
From the results so far, we concluded that PML4 and PML5 localize to the same compartment and thus might have redundant functions in the cell. The expression pattern revealed that PML4 and PML5, as well as PML3, are expressed in photosynthetic and in non-photosynthetic tissues ( Figure  4A and Figure S3). To define the expression patterns more specifically, promoter-driven reporter gene constructs were assembled by fusing a 1.3-kb fragment upstream of the PML3 coding sequence and 1.2-kb segments of the upstream regions of PML4 and PML5, respectively, to the uidA reporter gene. The constructs were designed to exclude the ATG of the endogenous genes; more precisely, the PML3 fragment comprises −1021 bp to +315 bp from the transcription initiation site, the PML4 fragment comprises −971 bp to +237 bp from the transcription initiation site, and the PML5 fragment comprises −1116 bp to +94 bp from the transcription initiation site (sequences of the fragments are shown in Table S1). Transgenic lines harboring the individual constructs were generated and named ProPML3::GUS, ProPML4::GUS and ProPML5::GUS. We found β-Glucuronidase (GUS) activity in rosette leaves and in roots of the transgenic line ProPML3::GUS, PML3 promoter-driven GUS expression was strong in lateral roots, throughout the leaf and particularly in the vasculature and in anthers ( Figure 4B). Both ProPML4::GUS and ProPML5::GUS lines showed similar yet distinct GUS activity patterns. In both lines, GUS activity was found in roots and leaves; however, PML4-driven

Expression Patterns of PML3, PML4 and PML5 in Arabidopsis Tissues
From the results so far, we concluded that PML4 and PML5 localize to the same compartment and thus might have redundant functions in the cell. The expression pattern revealed that PML4 and PML5, as well as PML3, are expressed in photosynthetic and in non-photosynthetic tissues ( Figure 4A and Figure S3). To define the expression patterns more specifically, promoter-driven reporter gene constructs were assembled by fusing a 1.3-kb fragment upstream of the PML3 coding sequence and 1.2-kb segments of the upstream regions of PML4 and PML5, respectively, to the uidA reporter gene. The constructs were designed to exclude the ATG of the endogenous genes; more precisely, the PML3 fragment comprises −1021 bp to +315 bp from the transcription initiation site, the PML4 fragment comprises −971 bp to +237 bp from the transcription initiation site, and the PML5 fragment comprises −1116 bp to +94 bp from the transcription initiation site (sequences of the fragments are shown in Table  S1). Transgenic lines harboring the individual constructs were generated and named ProPML3::GUS, ProPML4::GUS and ProPML5::GUS. We found β-Glucuronidase (GUS) activity in rosette leaves and in roots of the transgenic line ProPML3::GUS, PML3 promoter-driven GUS expression was strong in lateral roots, throughout the leaf and particularly in the vasculature and in anthers ( Figure 4B). Both ProPML4::GUS and ProPML5::GUS lines showed similar yet distinct GUS activity patterns. In both lines, GUS activity was found in roots and leaves; however, PML4-driven GUS expression was found mainly in minor veins, and PML5-driven GUS expression was predominantly detected in the main veins of adult leaves ( Figure 4C,D). The ProPML4::GUS line expressed GUS particularly in root hairs, while in line ProPML5::GUS, the GUS activity was found in the root stele ( Figure 4C,D). Moreover, the patterns of PML4 and PML5 promoter expression differed from each other in flower tissues ( Figure 4C,D). Only weak GUS activity in the receptacle tissue of ProPML5::GUS was detected, whereas petals of ProPML4::GUS displayed strong GUS activity. Taken together, this analysis shows that PML4 and PML5 are expressed in distinct tissues of roots, leaves and flowers, which suggests that they might have partially redundant functions in plants.
expressed GUS particularly in root hairs, while in line ProPML5::GUS, the GUS activity was found in the root stele ( Figure 4C,D). Moreover, the patterns of PML4 and PML5 promoter expression differed from each other in flower tissues ( Figure 4C,D). Only weak GUS activity in the receptacle tissue of ProPML5::GUS was detected, whereas petals of ProPML4::GUS displayed strong GUS activity. Taken together, this analysis shows that PML4 and PML5 are expressed in distinct tissues of roots, leaves and flowers, which suggests that they might have partially redundant functions in plants.

Root Elongation in pml3 Is Distinct from Other Genotypes in 500 µM MnSO4
We isolated the insertion alleles pml3, pml4 and pml5 ( Figure S4), and named the mutant lines accordingly. From our observations so far, we also aimed to generate a double mutant pml4/pml5 to test whether PML4 and PML5 have a redundant function. Single and double mutant plants grown in the greenhouse did not display differences compared to wild-type plants ( Figure S4). Thus, we examined the mutants at an earlier stage, with a focus on roots, because PML3, PML4 and PML5 expression is higher in roots than in rosette leaves ( Figure S3). Seeds of wild-type, pml3, pml4, pml5, and pml4/pml5 were sterilized and grown for ten days to inspect for a rosette-leaf phenotype and for root phenotype. The five genotypes were grown on medium containing 5 µM MnSO4, 50 µM MnSO4, and 500 µM MnSO4 ( Figure 5A).

Root Elongation in pml3 Is Distinct from Other Genotypes in 500 µM MnSO 4
We isolated the insertion alleles pml3, pml4 and pml5 ( Figure S4), and named the mutant lines accordingly. From our observations so far, we also aimed to generate a double mutant pml4/pml5 to test whether PML4 and PML5 have a redundant function. Single and double mutant plants grown in the greenhouse did not display differences compared to wild-type plants ( Figure S4). Thus, we examined the mutants at an earlier stage, with a focus on roots, because PML3, PML4 and PML5 expression is higher in roots than in rosette leaves ( Figure S3). Seeds of wild-type, pml3, pml4, pml5, and pml4/pml5 were sterilized and grown for ten days to inspect for a rosette-leaf phenotype and for root phenotype. The five genotypes were grown on medium containing 5 µM MnSO 4 , 50 µM MnSO 4 , and 500 µM MnSO 4 ( Figure 5A).
As a read-out for a rosette-leaf phenotype, we determined chlorophyll content. The chlorophyll content of the five genotypes is subjected to little fluctuation when plants were grown in 5 µM and 50 µM MnSO 4 , indicating that Arabidopsis can grow well within a certain range of Mn. The overall chlorophyll content went down when plants were subjected to 500 µM MnSO 4 and a comparison between the five genotypes revealed that they behave similarly ( Figure 5B, Table S2). As a read-out for a root phenotype, we determined the length of the primary root and observed that this parameter is also subjected to little fluctuation when plants were grown in 5 µM and 50 µM MnSO 4 ( Figure 5A,C). The overall root length increased when plants were subjected to 500 µM MnSO 4 ( Figure 5C); in another study, this tendency was observed when Arabidopsis plants were treated with 1250 µM MnSO 4 [30]. Interestingly, a comparison of the root length revealed that roots of pml3 grown in 500 µM MnSO 4 were significantly more elongated than roots of wild-type, pml4, pml5 and pml4/pml5 ( Figure 5C, Table S3) in 500 µM MnSO 4 and in any other condition. To investigate whether this effect can be inverted, plants were grown on medium with the MnSO 4 concentration set to 50 nM ( Figure S5). Plant growth was not drastically impaired, indicating that internal Mn stores possibly support plant growth to a certain degree. The overall root length decreased slightly in this condition (Table S3), in agreement with a trend described by Gruber et al. [31]. However, we could not detect any difference between the genotypes regarding root length and chlorophyll content ( Figure S5). Taken together, we concluded that the 500 µM MnSO 4 treatment caused an intracellular excess of Mn which influences root elongation and that PML3 is involved in this process. Consistent with other results, this finding indicates that the functionality of PML3 is distinct from that of PML4 and PML5.  Tables S2 and S3. Poorly germinated seeds were excluded from the analysis.
As a read-out for a rosette-leaf phenotype, we determined chlorophyll content. The chlorophyll content of the five genotypes is subjected to little fluctuation when plants were grown in 5 µM and 50 µM MnSO4, indicating that Arabidopsis can grow well within a certain range of Mn. The overall chlorophyll content went down when plants were subjected to 500 µM MnSO4 and a comparison between the five genotypes revealed that they behave similarly ( Figure 5B, Table S2). As a read-out  Tables S2 and S3. Poorly germinated seeds were excluded from the analysis.

Discussion
The genome of Arabidopsis contains three sequences that code for proteins with high similarity to PAM71 and CMT1, two previously characterized Mn transporters localized in distinct chloroplast membranes [8][9][10]. Like PAM71 and CMT1, the three predicted protein sequences contain the two highly conserved E-x-G-D-(KR)-(TS) motifs, with two negatively charged acidic residues in TM1 and TM4, which provide a suitable environment for the passage of Mn 2+ ions and maybe other cations through membranes. The greatest diversity within this protein family is found in their N-terminal regions (Figure 1), which were proposed to be responsible for targeting the respective protein to the correct membrane [23]. In the present study, we were able to localize PML3, PML4 and PML5 in the endomembrane system of Arabidopsis.
The PML3 protein sequence contains an N-terminal region predicted to serve as signal peptide for the secretory pathway (Figure 1), and indeed PML3 could be localized to the Golgi apparatus ( Figure 2, Figure S1), most likely in the Golgi membrane. PML3, together with its homologs, belongs to the UPF0016 family of membrane proteins, and the yeast and human members of this family were found to be located in the Golgi membrane as well [32]. Recent findings support a role for the human member transmembrane protein 165 (TMEM165) in Ca 2+ /Mn 2+ import into the Golgi. Indeed, Mn 2+ is known to be an essential cofactor for many Golgi-localized glycosyltransferases [33] and glycosylation abnormalities have been described in humans with mutations in TMEM165, which result in a rare genetic condition called Congenital Disorder(s) of Glycosylation [34]. Plant glycosyltransferases with functions in glycoprotein biosynthesis are also present in the Golgi [35,36]. In addition, glycosyltransferases in plants are involved in the biosynthesis of non-cellulosic polysaccharides of the cell wall [35], and some of them have been shown to depend on Mn 2+ [37]. It is perhaps not surprising that the pml3 insertion line does not show a drastic altered phenotype compared to wild-type ( Figure 5, Figure S4), as at least one other protein, the P-type ATPase ECA3, is known to be located in the Golgi membrane and possibly plays a role in the transport of Mn into this organelle [19]. ECA3 is expressed in all major organs, and particularly in the vasculature of primary and lateral roots, as well as in leaves and flowers [19,20]. Its expression thus overlaps with PML3 expression, which was found in all major organs, especially in the vasculature of leaves and in roots (Figure 4). The growth of the eca3 mutant is impaired when grown in Mn deficiency [19,20], whereas in pml3, only a subtle root phenotype appeared ( Figure 5). It appears that primary root elongation in pml3 plants is induced more in 500 µM MnSO 4 conditions ( Figure 5). Because the root length of pml3 plants was not changed in comparison to wild-type in lower MnSO 4 concentrations, we assume that PML3 is involved in specific processes during Mn excess, perhaps loading Mn into the Golgi. It is believed that Mn stored in the Golgi can be remobilized by NRAMP2 in Mn deficient conditions [21,22]. In addition, an interesting phenotype was observed when PML3:GFP was overexpressed. A stunted growth phenotype occurred in PML3:GFP over-expressor lines (Figure 3), with some leaves being out of shape. The underlying molecular mechanism evoking this phenotype is unknown, however one speculation is that massive Mn content in the Golgi apparatus of the over-expressor lines might alter glycosylation reactions, which eventually affects leaf shape and plant growth. Although highly speculative, this hypothesis certainly deserves deeper investigation in the future.
The subcellular localization of PML4 and PML5 was also determined and, unlike PML3, both proteins were found to reside in the endoplasmic reticulum (Figure 2, Figure S1). The importance of the endoplasmic reticulum in N-glycosylation of glycoproteins and in the biosynthesis of glycans for the cell wall matrix is well known [38,39]. Furthermore, the peptidyl serine O-α-galactosyltransferase (SGT1), which is involved in O-glycosylation, localizes to the endoplasmic reticulum, and was shown to require Mn 2+ as a cofactor [40]. Mn transport into the endoplasmic reticulum is mediated by ECA1, and the eca1 mutant showed increased sensitivity towards high Mn; in fact, it failed to elongate the root hairs under these conditions, presumably through impairment in growth tip processes [18]. We propose that PML4 and PML5 might act to fine tune Mn allocation into the endoplasmic reticulum of specific cell types, e.g., those of the root hair (PML4) or root stele (PML5) (Figure 4), as they are dispensable for primary metabolism. The interplay of various Mn transporters within single cells as well as on different tissue levels is complicated by the fact that some of them also transport other cations like calcium, or indirectly act on calcium homeostasis. Thus, the challenge of future work will be to elucidate the precise substrate specificity of PML3, PML4 and PML5 and perhaps other Mn transporters by employing heterologous expression and reconstitution assays. Furthermore, it will be interesting to learn more about the sophisticated network of Mn transporters, perhaps through the generation of triple or even higher order mutant lines.

Plasmid Construction and Plant Material
For subcellular localization studies, the sequences encoding PML4 and PML5 were amplified using specific primer combinations (Table S4) and single-stranded Arabidopsis cDNAs. The amplified cDNAs were ligated into pENTR-TOPO (Life Technologies, Carlsbad, CA, USA), and recombined into pB7FWG2 [41], yielding the plasmids pPro35S::PML4:GFP and pPro35S::PML5:GFP. The plasmid pPro35S::PML3:GFP was constructed from appropriate primers (Table S4), the backbone of pB7FWG2 and a template cDNA using the Gibson Assembly Cloning Kit (catalog number E5510S, New England Biolabs, Ipswich, MA, USA) in accordance with the manufacturer's instructions.
For study promoter activities, fragments located upstream of the PML3, PML4 and PML5 coding regions (Table S3) were amplified from Arabidopsis genomic DNA using specific primer combinations (Table S4) and ligated into pENTR-TOPO. The entry clones were subsequently recombined into pKGWFS7 [41] upstream of the uidA reporter gene, yielding the plasmids pProPML3::GUS, pProPML4::GUS and pProPML5::GUS. Thus, the constructs were designed to use the ATG of the uidA reporter gene.
All plasmids were transformed into Agrobacterium tumefaciens strain GV3101. Agrobacterium strains harboring pPro35S::PML3:GFP, pPro35S::PML4:GFP or pPro35S::PML5:GFP were either infiltrated into N. benthamiana leaves or used for stable transformation of Arabidopsis by the floral dip method. Individual transgenic Arabidopsis plants were selected on the basis of their resistance to ammonium glufosinate (100 mg L −1 ). Agrobacterium strains harboring pProPML3::GUS, pProPML4::GUS or pProPML5::GUS were also used for stable transformation of Arabidopsis. In this case, individual transgenic Arabidopsis plants were selected on the basis of their resistance to kanamycin (50 µg mL −1 ).
Arabidopsis mutant lines were obtained from the European Arabidopsis Stock Centre (NASC) and named pml3 (N402563=GK-027F07), pml4 (N664220=SALK_143524) and pml5 (N438509=GK-402B01), respectively. Genotyping of all lines was performed using PCR with appropriate primer combinations (Table S3). Expression of PML3, PML4 and PML5 in the different genotypes was analyzed by RT-PCR using an appropriate primer combination (Table S4)

Protoplast Isolation and Fluorescence Microscopy
Agrobacterium-mediated infiltration of N. benthamiana was performed as described [43]. Protoplasts were isolated from leaf tissue 48 h after infiltration [43] and fluorescent signals were detected using an Axio-Imager fluorescence microscope (Carl Zeiss, Oberkochen, Germany). Two plasmids expressing marker proteins for either the Golgi (pG-rk expressing GmMAN1:mCherry) or the endoplasmic reticulum (pER-rk expressing AtWAK2:mCherry) were chosen for co-infiltration, because their specificity is well established [26,44]. The GFP fluorescence was excited at 470 ± 40 nm and the emission recorded at 525 ± 50 nm. Chlorophyll autofluorescence was excited at 450-490 nm, and emission was recorded at >515 nm. The mCherry fluorescence of the marker proteins was excited at 560 ± 40 nm and the emission recorded at 630 ± 75 nm. The fluorescence signals obtained in the different channels were overlaid using ImageJ (version 1.51j).

Microsomal Preparation and Western Blot Analysis
For the isolation of microsomal fractions, transgenic Arabidopsis lines were grown for 4-5 weeks in a growth chamber under a 12 h/12 h light-dark cycle at 100 µmol photons m −2 s −1 and 22 • C/18 • C. Leaf homogenization and microsomal fraction enrichment by differential centrifugation was performed as previously described [45]. Isolated microsomal fractions were layered on top of a continuous 20-50% (w/v) sucrose density gradient and centrifuged for 16 h at 4 • C and 100,000× g. After centrifugation, 500 µL fractions were collected form the top of the tube and analyzed by SDS-PAGE and Western blotting [45]. Individual proteins were detected using antisera raised against GFP (Life Technologies A6455, 1:2000 dilution), SMT1 (Agrisera AS07 266, Vannas, Sweden; 1:500 dilution) and ARF1 (Agrisera AS08 325, 1:1000 dilution) in combination with anti-rabbit IgG horseradish peroxidase (HRP) (Sigma Aldrich, St. Louis, MO, USA 1:25,000 dilution) and the Pierce Enhanced Chemiluminescence System (Thermo Fisher Scientific, Waltham, MA, USA).

Real-time PCR Analysis
Total RNA was isolated from the root and shoot systems of Arabidopsis wild-type plants using TRIzol reagent (InVitrogen). For real time qRT-PCR, SYBR Green Supermix (Bio-Rad) was used, and PCR was performed with the iQ5 multi-color real-time PCR detection system (Bio-Rad). Quantification of relative expression levels was performed using the comparative cycle threshold (C T ) method [46].
Boxplots were generated in Excel 2016 and a one-way ANOVA with post-hoc Tukey HSD test was performed [52].

Conclusions
The main goal of the present study was to determine the subcellular localization(s) of the three PAM71 homologs: PML3, PML4 and PML5. The localization of PML3 to the Golgi, and of PML4 and PML5 to the endoplasmic reticulum, is clearly different from the chloroplast localizations of PAM71 and CMT1 in a leaf cell. Logically, PAM71 and CMT1 must both be present in chloroplast membranes at the same time. In contrast, the functions of the other three proteins do not necessarily have to be coordinated within individual cells. In agreement with their subcellular localization in the endomembrane system, expression of PML3, PML4 and PML5 is not restricted to photosynthetic cells, e.g., they are also expressed in non-photosynthetic cells of roots or flowers. The three proteins do not seem to be essential for plant growth and development, however a subtle root growth phenotype was observed in the pml3 mutant. Overall, root length increased in plants exposed to 500 µM Mn, and this phenomenon is boosted in pml3. Taken together, the cellular function of PML3 is distinct from that of PML4 and PML5 and is also evident from the leaf phenotype of the PML3 over-expressor line. We suspect that PML3 at the Golgi membrane plays a role in balancing excess manganese.