Structural prediction of two novel human atypical SLC transporters, MFSD4A and MFSD9, and their neuroanatomical distribution in mice

Out of the 430 known solute carriers (SLC) in humans, 30% are still orphan transporters regarding structure, distribution or function. Approximately one third of all SLCs belong to the evolutionary conserved and functionally diverse Major Facilitator Superfamily (MFS). Here, we studied the orphan proteins, MFSD4A and MFSD9, which are atypical SLCs of MFS type. Hidden Markov Models were used to identify orthologues in several vertebrates, and human MFSD4A and MFSD9 share high sequence identity with their identified orthologues. MFSD4A and MFSD9 also shared more than 20% sequence identity with other phylogenetically related SLC and MFSD proteins, allowing new family clustering. Homology models displayed 12 transmembrane segments for both proteins, which were predicted to fold into a transporter-shaped structure. Furthermore, we analysed the location of MFSD4A and MFSD9 in adult mouse brain using immunohistochemistry, showing abundant neuronal protein staining. As MFSD4A and MFSD9 are plausible transporters expressed in food regulatory brain areas, we monitored transcriptional changes in several mouse brain areas after 24 hours food-deprivation and eight weeks of high-fat diet, showing that both genes were affected by altered food intake in vivo. In conclusion, we propose MFSD4A and MFSD9 to be novel transporters, belonging to disparate SLC families. Both proteins were located to neurons in mouse brain, and their mRNA expression levels were affected by the diet.


Introduction
Membrane-bound transporter proteins translocate molecules over cellular membranes. Most transporters belong to three major groups [1]; channels move solutes down their electrochemical gradient, primary active transporters use various energy-coupled mechanisms to generate an ion/solute gradient to constitute movements, and secondary active transporters utilise existing energy to translocate molecules [2]. The largest group of transporters in human are the solute carriers (SLCs) [

Material and methods Ethical considerations and animals
The study was approved by, and carried out in accordance with the recommendations of, the local ethical committee in Uppsala (Uppsala Djurförsöksetiska Nämnd, Uppsala district court, under the permit numbers C39/16 and C419/12). Adult C57Bl6/J mice (Taconic M&B, Denmark) were housed in accordance with the Swedish regulation guidelines (Animal Welfare Act SFS 1998: 56) and European Union legislation (Convention ETS123 and Directive 2010/63/ EU). The animals were euthanized during the light period by either cervical dislocation or transcardiac perfusion of anaesthetised animals.

Phylogenetic clustering of MFSD4 and MFSD9
Hidden Markov Models (HMM) were calculated for mammalian MFSD4A and MFSD9 protein sequences, using HMMbuild from the HMMER package [31]. The models were used to search the protein datasets obtained from Ensembl version 86 [32] listed in Table 1, to identify related proteins. Sequences were manually curated, and proteins originating from the same locus and pseudogenes were removed. The longest protein from each species was combined in a multiple PSI/TM tcoffee sequence alignment [33]. Subsequently, their relationship was inferred according to the Bayesian approach, as implemented in mrBayes 3.2.2 [34,35]. This was done to confirm the relations between the identified proteins. The analysis was run via the Beagle library (Ayres et al. 2012), on six chains (five heated and one cold), with two runs in parallel (n runs = 2), for a maximum of 2,000,000 generations.
Sequence similarities between orthologue proteins were calculated using global pairwise alignments, based on the Needleman-Wunsch algorithm [36]. The alignments were built on the annotated proteins listed in Table 2.
Proteins are clustered into SLC families based on homology, function, structure [37] and sequence identity [2]. MFSD4A cluster phylogenetically with MFSD4B and the SLC29 family, whereas MFSD9 is closest related to the SLC46 family, MFSD10, MFSD14A and MFSD14B [5]. Global alignments were performed according to the Needle approach [36] to calculate sequence identities as a way to study if MFSD4A and MFSD9 could belong to existing SLC families.

Predicted protein structures of human MFSD4A and MFSD9
Transmembrane segments (TMS) in the human MFSD4A and MFSD9 proteins were predicted using the three topology tools TMHMM server (v. 2.0) [38], Phobius prediction [39,40] and Sousi [41]. These topology tools incorporate parameters such as hydrophobicity, charge bias, helix lengths and signal peptide predictions [38,39] when making predictions. The tertiary structures were built using Swiss-Model, a fully automated homology program [42]. A structurally known MFS lactose permease from E. coli [43] (PDB ID code: 2v8n) was used as template for the MFSD4A model, and the YajR [44] (PDB ID code: 3WDO) MFS protein, found many gram-negative bacteria, was the templet used for the MFSD9 model. The alignments between the proteins of interest and their templates were manually inspected to verify that conserved MFS motifs, like the characteristic cytoplasmic loop between TMS 6 and 7 [45] were matched. Images of the tertiary structures were finalized using Swiss-Pdb Viewer [46] and coloured using Adobe Photoshop CS6.

mRNA extraction and reverse transcription
Adult male C57Bl6/J mice were sacrificed by cervical dislocation, and the following organs were dissected freshly, as described in [9]: brainstem, cerebellum, cortex, eye, heart, hippocampus, hypothalamus, intestine, kidney, liver, lungs, olfactory bulb, ovary, spinal cord, spleen, striatum, thalamus and thymus (male mice), and uterus (female mice). Blood was collected from male mice via cardiac puncture, mixed with EDTA (1.5mg/ml blood, VWR), and centrifuged to retrieve a pellet for RNA extraction. N = 5 per organ. The samples were mechanically homogenized in a bullet blender (Averill Park, USA) and RNA was extracted using Absolutely RNA Miniprep Kit (Agilent Technologies), following the manufacturer's instructions. The final concentrations were measured in a spectrophotometer (ND-1000, NanoDrop  Table 3. Primer sequences used in the quantitative real-time PCR analyses.

Gene Forward Reverse
Mfsd4a Technologies). Reverse transcription was performed using the Applied Biosystems High-Capacity RNA-to-cDNA kit (Invitrogen), following the manufacturer's manuals. 2μg RNA was used as template for each cDNA synthesis. RNA was extracted separately for each organ and individual. The cDNA from each organ was then pooled and diluted to 5ng/μl RNA in sterile water.
Primer design, quantitative-time PCR and data analysis Primers for mouse Mfsd4a and Mfsd9 and reference genes (marked Ã ) were designed in Beacon Design 8 (Premier Biosoft, Palo Alto), and listed in Table 3. qPCR mastermix contained 2x DreamTaq Buffer (Thermo scientific), 0.2μl 20mM dNTP, 0.05μl forward and reverse primer (100pmol/μl), 0.5μl of SYBR Green (1:50000; Invitrogen) in TE buffer (pH 7.8), 1μl Dimethyl sulfoxide (Sigma Aldrich) and 0.08μl DreamTaq polymerase (5U/μl, Thermo scientific). 5μl pooled cDNA per reaction was used as template. The measurements were run on iCycler realtime detection instrument (Bio-Rad Laboratories) according to following parameters: 30 sec at 95˚C initial denaturation, followed by 50 cycles of 10 sec at 95˚C, 30 sec at 55-61˚C (optimal temperature for each primer pair) and 30 sec at 72˚C, followed by a melting curve (+0.5˚C per cycle, 81 cycles at 10 sec intervals, starting from 55˚C). Each sample was run in triplicates. Negative controls were included on each plate. All experiments were repeated twice. Raw-data was collected from the MyIQ (Bio-Rad Laboratories) software. Primer efficiencies were calculated using LinRegPCR software and Grubbs test (GraphPad software) was performed to remove outliers. The GeNorm protocol [47] was used to detect stable reference genes, and their geometric mean was used to normalize the data. Gapdh, bTub, Rpl19, Cyclo and Actb were stably detected between samples, and subsequently used for the normalization. The sample with the highest gene expression for each transporter was set to 100%, and the relative expression level of each tissue was plotted (±SD) in the GraphPad Prism 5 software.

Western blot to study antibody binding
Antibody binding was verified by western blot on fractionated mouse brain tissue as previously described [9,15]. Protein amount was 100μg per well and the protein transfer was performed with the Trans-Blot1 Turbo™ Mini PVDF Transfer Packs and Trans-blot Turbo Transfer system (Bio-Rad), following the manufacturer's instructions. Anti-MFSD4A (1:100, rabbit, AV53395, Sigma-Aldrich) and anti-MFSD9 (1:50, goat, sc-247973, Santa Cruz) were used as antibodies, and a molecular weight marker (PageRuler Prestained, Thermo Fisher Scientific) was included as reference on each blot. HRP-coupled secondary antibodies (anti-rabbit and anti-goat (Invitrogen) dilution 1:10000) were added followed by chemiluminescent development using Clarity Western ECL Substrate (Bio-Rad). Staining was visualized using a CCD camera (Bio-Rad). Glycosylation sites for the mouse proteins were predicted using the NetO-Glyc 4.0 Server from CBS Predictions Servers [48].

Preparation of mouse brain sections
See [9] for precise procedures regarding fixation, paraffin embedding and sectioning. In brief, adult male mice C57Bl6/J were anesthetized by i.p. injection of 0.01mg/g body weight sodium Pentobarbital (Apoteket Farmaci, Sweden). The tissue was fixed by transcardiac perfusion with 4% formaldehyde (Histolab, Sweden). Brains were stored in 4% formaldehyde over night before sectioning. For DAB staining, the fixed brains were mounted in 4% agarose (VWR) and cut into 70μm coronal sections using a Leica VT 1200 S vibratome (Leica Microsystems). For fluorescent immunohistochemistry the tissue was embedded in paraffin [9] and cut into 7μm coronal sections using a HM355S microtome (Thermo Scientific).

Mice exposed to altered nutritional intake, followed by mRNA expression measurements
Since both Mfsd4a and Mfsd9 were detected in mouse brain, with expression in brain areas implicated in food intake and its regulation, hypothalamus, pituitary gland, cortex, striatum, thalamus, brainstem and spinal cord were selected for RNA expression analysis. Male mice were divided into three groups receiving three different diets, as described in [9]. 1) Standard chow (control diet), consisting of 5.0% fat, 21.0% protein, and 51.5% carbohydrates (R3, Lantmännen), 2) standard chow ad libitum, and starved for 24h prior euthanasia and 3) fed highfat western diet (HFD), containing 21.0% fat, 17.2% protein, 43.0% carbohydrates (R638, Lantmännen) for eight weeks to induce obesity, all according to [9]. No deviant behaviours in the mice were observed. At dissection day, the obese mice had a 38.0% ± 9.0% mean weight gain compared to a 12.0% ± 2.3% increase in controls. N = 4 per diet for each brain area. All animals had access to water ad libitum. mRNA extraction, cDNA synthesis and qPCR were done as described above.
For the analysis, the geometric means of the three most stable reference genes (Gapdh, H3a and Actb) according to GeNorm calculations were used for normalization of the ct-values±SD. The values from control samples for each area were set to 1, and the test group sample values were stated as relative values. Differences in gene expression between groups were analysed by unpaired t-tests, followed by Bonferroni correction for multiple testing (Graphpad Prism 5). Significance levels were set to Ã p 0.0493, ÃÃ p 0.00998, ÃÃÃ p 0.001.

MFSD4A and MFSD9 have orthologues in several vertebrates
Several proteomes (yeast, roundworm, fruit fly, zebrafish, three-spined stickleback, lizard, chicken, mouse and human) were scanned with HMMs. The identified proteins were combined in a phylogenetic tree to identify orthologue proteins to human MFSD4A and MFSD9. Both proteins were conserved in all vertebrate data sets studied ( Fig 1A). The hits obtained Abbreviations: ac, anolis lizard; dr, zebrafish; ga, stickleback; gg, chicken; mm, mouse. Neither of the proteins was found in yeast, roundworm or fruit fly. MFSD4B was identified as orthologue to MFSD4A. Global pairwise alignments were run to calculate protein identities between human and animal orthologues, as listed in (B). The human protein sequence for MFSD4A (C) and MFSD9 (D) were depicted, where the 12 transmembrane segments (TMS) were underlined (TMS1-12). The first 6 TMS constitute the Ndomain which is connected to the C-domain (TMS7-12) via a long cytoplasmic loop (grey). Both the N and C terminals were predicted to be localised on the cytoplasmic side of the membrane.
https://doi.org/10.1371/journal.pone.0186325.g001 from the yeast, roundworms and fruit fly proteomes did not clustered in proximity with the human proteins. They were discarded as orthologues, even if they still were considered to be related proteins. MFSD4B (synonymous names: KIAA1919 and NAGLT1) was identified as a related protein in all species in the MFSD4A searches. Globally, protein sequence identities were high between the orthologues (Fig 1B), suggesting similar protein function among species. Amongst the organisms examined, human and mouse shared the highest sequence identity, with 84.0% for MFSD4A, and 72.5% for MFSD9. When comparing human MFSD4A with MFSD4B, sequence identity amounted to 20.0% (Fig 1B).

Predicted protein structures for human MFSD4A and MFSD9
The number and structure of possible transmembrane segments (TMS) were assessed based on primary amino acid sequence using TMHMM [38], Phobius [39,40] and Sousi [41]. All three tools provided similar results concerning the amino acids spanning the membrane. For MFSD4A, all three programs predicted 12 TMS, in accordance with most MFS proteins [11,13], with N and C terminals on the inside. TMHMM predicted MFSD9 to have 10 TMS, where only eight met the requirement for highest probability, whereas Phobius and Sousi predicted MFSD9 to have 10 TMS.
Homology models were built using the SWISS-MODEL program [42], in which a structurally known MFS transporter was used as template for each model. Global model quality estimation indicates the reliability of models on a scale range from 0-1, 1 representing total reliability. The MFSD4A model reached a quality score value of 0.40. MFSD9 reached a model quality score of 0.47. Fig 1C and 1D depicts a schematic representation of the proteins topology, where the TMS are detailed according to the homology models. The three dimensional model for MFSD4A resulted in 12 TMS (Fig 2A and 2B), which correlate well with the TMS identified in the secondary structure prediction. Likewise, for MFSD9 12 TMS were identified (Fig 2C and 2D), unlike the 10 TMS previously predicted based on primary amino acid sequence. However, after comparing the TMS identified by the three topology tools with the homology model for the MFSD9 models, TMS3 and 8 from the secondary models corresponded to four TMS in the homology models, providing a final 12 TMS structures, with N and C terminals on the cytoplasmic side of the membrane. For MFSD4A, all except TMS 9 and 12 were incomplete helices, with disrupted intermediate sections, while for MFSD9, TMS 4, 5, 6 and 11 were predicted to be incomplete helices. In both models, TMS 1, 4, 7 and 10 appear to be closest the substrate pore, and they were predicted to contain a long cytoplasmic loop between TMS 6 and 7 (Fig 2A and 2C), two traits common for MFS protein structures [45]. Both peptides folded into a cylinder, through which molecules possibly could be transferred (Fig 2B and 2D).

Mfsd4 and Mfsd9 mRNA expression in both central and peripheral regions
To gain a comprehensive understanding of Mfsd4a and Mfsd9 expression in mice, mRNA levels were measured using quantitative real-time PCR (qPCR). Both genes were expressed in the nervous system and in peripheral tissues (Fig 3). In general, Mfsd4a had higher relative expression in central than peripheral areas. Mfsd4a was relatively high expressed (normalized relative expression ± SD) in cerebellum, hippocampus and hypothalamus, with slightly lower levels in brainstem and cortex (Fig 3A). Variation in mRNA levels between the peripheral organs measured was larger, with highest relative levels in intestines and kidneys, and lowest in the heart, liver and spleen. Mfsd9 was also detected throughout all organs tested (Fig 3B), where no discrepancy could be detected between CNS and peripheral expression. However, there was variation within CNS and the periphery, where, for example, the brainstem sample expressed higher Mfsd9 mRNA levels than striatum, and kidney had higher relative Mfsd9 expression than the eyes. High levels of Mfsd9 were also detected in intestine and kidney.

Western blot to study antibody binding
Commercially available antibodies were used for protein staining. Antigen binding was confirmed for each antibody by western blot run on fractionated mouse brain. The blot for MFSD4A produced two bands, one at 40 kDA and one at 56 kDA in the mouse brain (Fig 4A), which corresponded with predicted sizes of the isoforms having 56kDA (Ensembl number: ENSMUSP00000125558 and ENSMUSP00000107989) and 40kDA (Ensembl number: ENSM USP00000039635) as molecular weights. MFSD9 bound at 60kDa (Fig 4B), corresponding to Ensembl number: ENST00000258436.9, having a molecular weight at 51kDA. The band was slightly larger than predicted, likely due to post-translational modifications such as glycosylations. MFSD9 was predicted to contain 12 possible glycosylation sites, as found by analysis via NetOGlyc 4.0 [48]. MFSD4A was predicted to have one possible glycosylation site. The western blot indicated that both antibodies were specific for their respective target.

Identified MFSD4A and MFSD9 staining in mouse brain neurons
To characterize distribution of both MFSD4A and MFSD9 in mouse brain tissue, fluorescent immunohistochemistry was performed, including the neuronal marker NeuN [49] and the astrocytic marker GFAP [50]. Both MFSD4A and MFSD9 co-localised with NeuN ( Fig 4C and  4D), but not with GFAP (Fig 4E and 4F).

Immunohistochemistry staining of MFSD4A in mouse brain
As both proteins were located to neurons, the global neuroanatomical distribution was analysed with DAB immunohistochemistry on 70μm coronal brain sections. Fig 5 displays representative MFSD4A staining, in whole brain sections at bregma 0.38mm; -0.70mm; -1.22mm and -5.80mm (Fig 5A-5D), and selected magnified areas (Fig 5E-5L). Specific layered staining for MFSD4A in cortex layer 5 (Fig 5E), and in the magnified cortex-picture projections were visible (Fig 5F). Labelling of interspersed cells was present in hypothalamic areas around the third ventricle (3V) (Fig 5G), globus pallidus (Fig 5H), amygdala nuclei, starting in the cortexamygdala transition zone (Fig 5I), and in the CA2 and CA3 region in the hippocampus ( Fig  5J). Moreover, distinctly layered staining was observed in the cerebellar cortex, where MFSD 4A clearly marked the Purkinje cell layer and their dendritic trees reaching the molecular layer in the paraflocculus (Fig 5K). Staining was seen in the plasma membrane of the soma, as well as in the neuronal projections (Fig 5F and 5I-5K). Finally, dense MFSD4A staining was observed in the facial nuclei of the brain stem (Fig 5L).

Immunohistochemistry staining of MFSD9 in mouse brain
MFSD9 displayed a more abundant staining pattern from rostral to caudal parts of the mouse brain, as showed at bregma 0.38mm; -0.34mm; -1.58mm and -5.80mm (overview images: Fig  6A-6D; with adjacent magnified micrographs Fig 6E-6L). Densely interspersed staining of MFSD9 was found in the matrix of the striatum (Fig 6E). Evenly distributed staining was visualized throughout cortex (Fig 6F) with stained possible projections in cortex layer 4 and 5 ( Fig  6G). MFSD9-positive neurons were found throughout the hypothalamus, around third ventricle ( Fig 6H) and in thalamic cells, (Fig 6I). Immunostaining was detected in the fields CA2 and CA3 in hippocampus, where projections from the pyramidal cell layer towards deeper hippocampal layers could be observed in CA2 stretching into the dorsal part of CA3 (Fig 6J). In brainstem, MFSD9 staining was observed in a group of large cells, in which the staining was distributed evenly throughout the soma and projections (Fig 6K). In the cerebellar cortex, the Purkinje cell layer distinctly stained positive for MFSD9 (Fig 6L).
Expression levels of Mfsd4a and Mfsd9 were affected by altered nutrient intake As several MFSD genes are known to respond to altered nutritional status, mice were exposed to three food paradigms, normal chow, food restriction and high-fat diet (HFD). mRNA levels for Mfsd4a and Mfsd9 were measured in the following brain areas: hypothalamus, pituitary gland, cortex, striatum, thalamus, brainstem and spinal cord. For Mfsd4a (Fig 7A), no alterations in transcription levels were detected in hypothalamus or thalamus. In pituitary gland, showed a band at 60 kDa (expected size at 51 kDa) in the mouse brain. 7μm coronal adult mouse brain sections were stained for MFSD4A and MFSD9, together with the neuronal marker NeuN and astrocytic marker GFAP, and representative staining is shown. Both MFSD4A (green in C) and MFSD9 (red in D) staining co-localised with NeuN. No overlay was detected between MFSD4A (green in E) and GFAP (red in E), or MFSD9 (red in F) and GFAP (green in F). DAPI was used as a nuclei marker and stained in blue.
Mfsd9 was affected by the food paradigms (Fig 7B). Food deprivation increased the expression levels in hypothalamus (p = 0.0005), whereas HFD provided no effects. In pituitary gland, starvation caused a reduction of the expression (p = 0.0082), while HFD had no effects. The   Fig 6. Neuroanatomical distribution of MFSD9. MFSD9 histology was displayed using colorimetric staining on 70μm coronal sections from adult mouse brains. Overview micrographs (A-D), with magnifications (E-L) are showed. MFSD9 staining in striatum (CPu) (E), cortex (F) with possible projections stained in layer 4 and 5 (G) and hypothalamic areas around third ventricle (3V) (H). In thalamus cell bodies were marked by MFSD9 (I). In hippocampus, both soma and projections were detected in CA2, while only soma was seen in CA3 (J). A close up of cells in brainstem (K) and the Purkinje cell layer (L) showed prominent staining. Bregma regions were according to [51], and scale bars represent 100μm.
https://doi.org/10.1371/journal.pone.0186325.g006 cortex responded with down-regulation to starvation (p = 0.0040), while HFD enhanced the Mfsd9 levels (p = 0.030). No significant changes were seen either in striatum or thalamus. Furthermore, both starvation and HFD up-regulated the Mfsd9 levels in brainstem (starvation, p = 0.0002; HFD, p = 0.0050), whereas spinal cord was resistant toward food deprivation, but reduced by HFD (p = 0.034). In conclusion, both genes were affected by the nutritional status in vivo.

Discussion
The MFS [52] constitutes the largest superfamily of phylogenetically related secondary active and facilitative transporters [11,45]. In humans, these types of transporters constitute the solute carrier (SLC) family, of which approximately 30% contain the MFS fold [5]. Among the SLCs of MFS type, there is a subclass of orphan permeases denoted MFSD proteins. We identified MFSD4A and MFSD9 orthologues in several species that could be used when searching the proteins functions, and we predicted the three dimensional structures of human MFSD4A and MFSD9. Moreover, we elucidated the neuroanatomical distribution of the transporters in mice to get a comprehensive understanding of their distribution patterns. Since MFSD4A and MFSD9 are probable transporters involved in nutritional uptake and/or ion transport, we analysed the impact food deprivation and high-fat diet had on their expression levels.
Homologous MFS proteins are recognized in all organismal phyla [11,52], and the human MFSD proteins have identified orthologues in bacteria, archaea and eukaryotic domains [8,9,16,53]. MFSD4A and MFSD9 were estimated to be comparably young proteins, as relative proteins were only identified in vertebrates, of which orthologues from human and mice shared around 80% of the amino acids. This suggests that both proteins emerged late in evolution, likely to perform specialized tasks in higher species. As SLCs with high sequence similarity usually have similar substrate profiles [54], it is plausible they also share expression patterns. The high sequence identity between the human and mouse proteins suggests that the proteins share function, which makes mice a well-suited model to use when searching for the transporters' mechanisms. Since orthologues were found in chicken and zebrafish as well, there are more good alternative animal models that could be used when elucidating the proteins functions.
We identified MFSD4A and MFSD4B as related and they shared 20% of the amino acids, meaning they meet the criteria for belonging to the same SLC family [2,37]. As family members, it is possible that they share a common substrate and mechanism. As MFSD4B is a known sodium dependent sugar transporter [21,55], it is possible that also MFSD4A transports sugars. MFSD9 shares both phylogenetic branching [5] and more than 20% sequence identity with SLC46A1, MFSD10, and MFSD14B, suggesting they could all be members of the SLC46 family. However, they possess different substrate profiles. SLC46A1 is a folate transporter [56], while MFSD10 (TETRAN) transports organic anions [18].The substrate for MFSD14B is unknown, but due to its high sequence identity (67.7%) with the predicted sugar transporter MFSD14A [57], it is presumed that MFSD14B transports sugar. The substrate for MFSD9 remains to be revealed.
MFSD4A and MFSD9 are similar in sequence and phylogeny with known SLCs, suggesting they are transporters as well. To assess this issue, we used online topology prediction tools to identify possible transmembrane segments (TMS), followed by homology modelling to predict their structures. The topology tools rely on the properties of the amino acid sequence, such as hydrophobicity, charge bias and helix lengths [38,39] to calculate possible TMS that could span the membrane, whereas the homology models were built in comparison to a structurally determined protein. The optimal set-up for homology modelling would be to use orthologue proteins as templates, but since MFSD4A and MFSD9 lack known related proteins in bacteria, and have no orthologue with determined structure, this cannot be done. Therefore, structurally known MFS proteins were used as templates, and even though the overall sequence identities were low, the conserved MFS domains constitutes good hallmarks when aligning the sequences to the templates [58], and the models were considered reliable.
For MFSD4A, all structural models suggested 12 TMS, and for MFSD9 the prediction tools identified only 10 TMS, while the homology model identified 12 TMS. Under such circumstances the results from the homology modelling holds higher validity, as the protein of interest is aligned against a protein with known structure. As an example, the characteristic long cytoplasmic loop between TMS 6 and 7 was identified in both MFSD4A and MFSD9, and it aligned well with the templates. When comparing the amino acids in the 10 predicted TMS in the secondary models for MFSD9 with the 12 TMS found in the homology model, it was evident that all tools identified the same TMS. However, in the tertiary model, TMS 3 and 8 were divided into four helices. Both models contained incomplete helices, suggesting a lower hydrophobicity index for those helices. This does however not mean that these proteins lack the standard helical configuration. This could be due to hydrophilic residues pointing toward the water filled transport pore or that these hydrophilic residues were shielded from the lipid bilayer by other parts of the protein. As TMS 1, 4, 7 and 10 directly constitute the transport path, and are located in the core of the transporter [45], they can be amphipathic depending on the substrate of the transporter. As these four TMS in the MFSD4A predicted structure mainly consisted of hydrophobic and neutral amino acids, it suggests a non-polar substrate, possible sugars as for MFSD4B [21,55]. For MFSD9, TMS 1, 4, 7 and 10 contained hydrophobic, hydrophilic and neutral amino acids, suggesting it could translocate charged substrates. Finally, to our knowledge, there is no conclusive evidence supporting a 10 TMS model for MFS proteins, even though it has been investigated [11]. Consequently, we suggest both MFSD4A and MFSD9 to have 12 TMS, composed of the N and C domains.
Commercially available antibodies were used for protein staining, and there are several ways to verify antibodies specificity. A refined way is to use blocking peptides followed by measurements of reduced antibody binding. There are also the possibilities to create knockout mice or knockout cells to show that the antibodies have no cross reactivity. However, for MFSD4A and MFSD9 there were, at present, neither blocking peptides available, nor any confirmed knockout mice or knockout cells to utilise. The specificity of antibodies can also be studied using siRNA knockdown strategies in cultured cells, but both MFSD4A and MFSD9 maintain low protein levels in rodent cell cultures and the proteins cannot be detected. Consequently, we used western blot to ensure the antibodies' accuracy. With western blot it was verified that both antibodies bound epitopes on proteins having correct predicted sizes, but crossreactivity cannot be excluded. The histology displays the likely distribution of MFSD4A and MFSD9.
Based on the results presented herein, MFSD4A and MFSD9 are suggested to be novel SLC transporters, belonging to disparate SLC families. That Mfsd4a and Mfsd9 were highly expressed in intestines and kidneys increases the possibility of involvement in nutritional regulation. Since approximately half of all known Slc genes present in mouse brain areas are involved in the regulation of food intake and energy production [14], the distribution of MFSD4A and MFSD9 was studied in mouse brain. This showed specific protein staining in food controlling areas such as brainstem [59], hypothalamus [60] and striatum [61], and it corresponded well with measured relative mRNA levels. Since MFSD4A and MFSD9 probably transport nutrients we decided to study if and how they were affected by changed food intake in food regulatory areas. 24 hours of food deprivation was analysed to study acute effects, while 8 weeks of HFD mirror how the genes were affected by food-induced obesity in mice.
Diet changes like these can in mice, as in humans, cause the metabolic syndrome [62], with effects like hyperinsulinemia, hyperglycaemia and hypertension. This could contribute to altered mRNA levels and the effects measured could be a resolute of other confounding factors. In general, both genes had similar response to the diets, with some exceptions: in hypothalamus, Mfsd4a was unaffected, whereas there was a prominent increase of Mfsd9 after food deprivation. This up-regulation could imply a required increase in uptake of certain molecules. Due to the phylogenetic and sequential resemblance between MFSD9 and MFSD14A, the Mfsd9 increase could be a response to the diminished sugar intake, as MFSD14A is a predicted sugar transporter [57]. Both the arcuate nucleus [63,64] and ventromedial nucleus [64,65] of the hypothalamus contain specialized glucose-sensitive cells, and perhaps MFSD9 is involved in this regulation. In striatum it was the opposite; Mfsd4a was reduced by high-fat diet, while Mfsd9 remained normal. Such reduction of expression upon exposure to HFD suggests that the system has reached satiety, and that the cells abolish the uptake mechanisms.
In conclusion, that human MFSD4A and MFSD9 proteins were predicted to have the MFS structural appearance, and that they phylogenetically cluster with SLCs suggest they could function as transporters. That protein staining and mRNA expression were detected in food regulatory areas, and that transcriptional changes were measured after altered food intake, suggests involvement in nutrient intake or regulation.