Insights into the roles of the Sideroflexins / SLC56 family in iron homeostasis and iron-sulfur biogenesis

Sideroflexins (SLC56 family) are highly conserved multi-spanning transmembrane proteins inserted in the inner mitochondrial membrane in eukaryotes. Few data are available on their molecular function but, since their first description, they were thought to be metabolite transporters probably required for iron utilization inside the mitochondrion. Such as numerous mitochondrial transporters, sideroflexins remain poorly characterized. The prototypic member SFXN1 has been recently identified as the previously unknown mitochondrial transporter of serine. Nevertheless, pending questions on the molecular function of sideroflexins remain unsolved, especially their link with iron metabolism. Here, we review the current knowledge on sideroflexins, their presumed mitochondrial functions and the sparse but growing evidence linking sideroflexins to iron homeostasis and iron-sulfur cluster biogenesis. Since an imbalance in iron homeostasis can be detrimental at the cellular and organismal levels, we also investigate the relationship between sideroflexins, iron and physiological disorders. Investigating Sideroflexins’ functions constitutes an emerging research field of great interest and will certainly lead to main discoveries on mitochondrial physiopathology.


Sideroflexins from an historical point of view
The mitochondrion is at the crossroad of key metabolic pathways (energy metabolism, central carbon metabolism, one carbon metabolism, lipid, nucleotides and amino acids synthesis, etc.) and is a key player in cell fate and response to stress or infection. In order to ensure its essential functions within the cell, the mitochondrion requires a wide variety of enzymes and transporters. Among these proteins, sideroflexins (SFXN) form a family of recently discovered mitochondrial proteins whose cell functions are progressively being specified. The first mention of the name "sideroflexin" appeared in 2001 [1]. Since then, a few studies have been dedicated to SFXN proteins and, at the time we are writing this review, only 24 articles are retrieved in Pubmed using the keyword "sideroflexin". Pioneers in the SFXN field, Fleming et al. identified a mutation affecting the Sfxn1 gene in the flexedtail mouse and emitted the hypothesis that the loss of Sfxn1 was responsible for the sideroblastic anemia phenotype. However, it should be noticed that the causal link between the mutation in the Sfxn1 gene and the phenotype of flexed-tail mice has not been clearly established yet. It was even questioned following a study showing that flexed-tail mice also had a mutation of the Madh5/Smad5 gene, involved in the BMP pathway, which could explain the anemia and flexed-tail phenotype [2,3]. scheme of the SFXN1 protein and its conserved motifs. B. Alignment of human SFXNs protein sequences. Red amino acids are for high consensus levels (90%), the blue ones are for low consensus levels (50%). Meaning of symbols found in the consensus line: "!" is for Ile or Val, "$" is for Leu or Met, "%" is for Phe or Tyr," #" is anyone of Asn, Asp, Glu, Gln. Conserved motifs are shown and highlighted using an HMM logo created using Skyline (http://skylign.org/) with consensus colors for amino acids according to the ClustalX coloring scheme. SFXN1 topology was recently investigated by APEX and classical biochemical experiments [13][14][15]. Acoba et al. [15] performed detergent extraction and protease-protection assays on HEK human cells and confirmed that endogenous SFXN1 is a mitochondrial protein inserted in the inner mitochondrial membrane (IMM). Furthermore, evidence was given for the presence of N-terminus in the intermembrane space (IMS) but not in the matrix contrarily to what is predicted by a in silico analysis using Protter. According to biochemical data, the C-terminus seems to protrude in the matrix, in agreement with the previously proposed 5 transmembrane domains. However, our model is rather in agreement with a TM domain composed of six alpha helices and, if this predicted structure is correct, N and C termini could be in the same mitochondrial compartment (Figure 2). CryoEM structure of SFXN1 is thus needed to precise the three-dimensional structure of this carrier. Moreover, two recent studies investigated the mechanisms of SFXN1 mitochondrial import and shed light on the role of TIM22 and AGK2 in this process [16,15]. Evidence for a mitochondrial localization of SFXN are listed in the Table 1.
Because of their predicted structure, showing several hydrophobic alpha helices, and their mitochondrial location, sideroflexins were proposed to be mitochondrial metabolite transporters. Rat Sfxn3 was presumed to be a tricarboxylate carrier (TCC) and, later, Sfxn5 (also known as BBG-TCC ) was reported to transport citrate in vitro [17,18]. However, it is only recently that a function of mitochondrial serine transporter was reported for SFXN1 [8].
By a bioinformatic analysis, the S. cerevisiae Fsf1 (YOR271cp) was proposed to be a candidate alpha-isopropylmalate transporter but no experimental data ascertained this function [19]. Similarly, the predicted Fsf1 protein from Schizosaccharomyces pombe, Spac17g6.15c, is annotated as a serine transporter in the database Pombase (https://www.pombase.org/) based on its homology with human SFXN1 [20,21], although it has not been extensively studied.
Since mice lacking Sfxn1 present similar features to that observed in human syndromes caused by a lack of pyridoxine or ALAS2 mutation (X-linked sideroblastic anemia), it was also proposed that Sfxn1 transports pyridoxine (B6 vitamin) inside the mitochondria [1,22]. Since pyridoxine is the precursor of pyridoxal phosphate that serves as a cofactor for ALAS2 (the erythroid specific enzyme catalyzing the first step of heme biosynthesis), SFXN1 could thus directly regulate heme biosynthesis. However, it has been recently reported that human SFXN1 is not able to transport pyridoxine in vitro [8]. Even if we cannot exclude that SFXN1 functions in a complex that is not fully reconstituted in in vitro assays, SFXN1 may not be the carrier for pyridoxine. Mtm1p, SLC25A39 yeast homologue, was suggested to import pyridoxal 5'-phosphate inside the mitochondria [23,24]. However, the substrate specificity of the SLC25A39 carrier remains unknown [25].
Thus, the main role of Sfxn1 seems to be the mitochondrial serine import. Inside the mitochondrion, Serine can be catabolized by the serine hydroxymethyl transferase (SHMT2) into glycine, an amino acid necessary for ALA synthesis (see below). So, the lack of Sfxn1 would lead to decreased mitochondrial levels of serine and glycine leading to ALA synthesis impairment (see section 4). The confidence of the predicted model shown here is very high (with estimated TM-score=0.806). The model was built by trRosetta based on de novo folding, guided by deep learning restraints. iCn3D was used for the visualization of 3D structure [26]. A. SFXN1 predicted structure reveals several alpha helices and beta strands. N and C termini are labelled. The inlet shows the position of the HPDT motif (aa 80-83), located just after the fourth helix. B. Two views highlighting secondary structures (helices in red, beta sheets in green). C. Models for SFXN1 insertion in the inner mitochondrial membrane.

80
Mitochondrial translation is not dramatically impaired in the absence of a functional SFXN1 protein, 81 nevertheless a slight decrease in cytochrome b translation was reported in this study.

82
No decrease neither in the quantity of mtDNA nor in the mitochondrial mass was seen in SFXN1

83
KO cells, thus a general defect in mitochondrial biogenesis can be excluded [8,15]

96
Deficiencies of mitochondrial respiration and/or RC activity were also reported for other SFXN, as 97 summarized in Table 2. For example, SFXN2 knockout led to a decreased activity of CII-CIII and CIV 98 [9]. As no specific impairment in complex III activity has been described nor in SFXN2 nor in SFXN4

116
Hence OCM, through the folate cycle, links serine catabolism to purine and nucleotides biosynthesis.

117
Liver, kidney and blood are tissues with high OCM activity, however OCM role is not restricted to 118 these organs but present in all human tissues including brain [46]. Actually, defective one-carbon 119 metabolism during embryonic development is responsible for neural tube defects.

120
Whereas Jurkat cells lacking SFXN1 proliferate as wild-type cells do, their proliferation rate is 121 markedly reduced in a medium lacking serine but is normal in the absence of glycine that can be 122 provided by the catabolism of serine [8]. A lower proliferative rate compared to that of wild-type

163
Iron is an essential cofactor for several enzymes involved in redox reactions due to its ability to

212
Heme biosynthesis (Figures 4, 6) is a pathway comprising eight steps, among which four arise inside 213 the mitochondrion (e.g. the first and the last three steps). The rate limiting enzyme of this process is The first evidence for a link between sideroflexins and iron metabolism came from a study of 223 the flexed-tail mouse, which harbors a mutation in a locus containing the Sfxn1 gene [1]. Mice mutant for Sfxn1 displayed sideroblastic anemia, microcytic anemia and hypochromic erythrocytes.
Based on the annotation of SFXN as transporters of metabolites required for iron metabolism,

229
we and others have tried to monitor the consequences of the loss of SFXN on iron cellular levels.
230 Table 3 summarizes the experimental evidence for an iron imbalance in the absence of SFXN.   requires Gly import through SCL25A38 on the one hand, and ALA export on the other hand,

289
presumably through the same transporter (Figure 6) [74]. SFXN1 was shown to be a Serine 290 transporter in vivo [8]. Intramitochondrial Ser would be catabolized by SHMT2 into Gly and 5,10-  Figure S1, Appendix A).  transporter attributed to SFXN. Following its import into the mitochondrion, Ser can be converted in protoporphyrins into which iron is incorporated in the final step of heme synthesis catalyzed by 341 FECH (Figure 6). As SFXN1 is presumed to be the mitochondrial transporter of Ser, its loss could 342 increase cellular Ser and lower Gly levels. Indeed, in Jurkat and K562 SFXN1 KO cells, the cellular 343 Ser/Gly ratio was increased and associated to increased cellular Ser levels but decreased Gly levels 344 [8]. In agreement with an imbalance in serine levels upon SFXN1 loss, HEK SFXN1 KO cells also have 345 increased cellular Ser levels and Ser/Gly ratio but no decrease in Gly cellular levels were reported 360 SFXN2 has been recently described in HEK293 cells to have a key role in iron metabolism, mainly 361 in heme synthesis [9]. High levels of iron have been shown in mitochondria in SFXN2 knockout 362 HEK293 cells. Also, a decreased activity of Complexes II-IV but not of the Complex I was noticed.

363
Complex I subunits contain Fe-S clusters, in contrast to Complex IV, which is mainly composed by 364 heme groups. Complexes II and III contain both Fe-S clusters and heme groups (Figure 2). Thus, as 365 no effect in Complex I was detected, and no decrease in Frataxin (FXN), a mitochondrial enzyme 366 required for the Fe-S cluster formation, nor in ALAS2, the enzyme that catalyzes the first step of the 367 heme biosynthetic pathway, was reported, it was concluded that SFXN2 mutants affected heme neither the levels of ISC-containing proteins nor those of ALAS1 have been assessed in this study. It 370 is surprising because ALAS2 is the erythroid specific form and ALAS1 the housekeeping one.

371
We propose few possibilities to explain SFXN2 knockout cells phenotype. The lack of SFXN2 372 could either lead to an impaired ALA export or no mitochondrial import of protoporphyrin (PPIX) 373 for the last step of the heme pathway. A defective mitochondrial export of the heme groups is another 374 plausible explanation. Finally, other options could be possible as an interaction of SFXN2 with BCS1L, 375 a chaperone anchored to the inner mitochondrial membrane that is required for proper assembly of 376 the Complex III (see section 2.2 for more details). In all those cases, an intramitochondrial iron 377 accumulation is presumed. All those possibilities, and others, must be studied to be able to clarify the  its availability for those purposes [8]. The lack of SFXN1 activity can be overcome by SFXN2 and 409 SFXN3 but not by SFXN4 [8].

476
It would thus be interesting to study FSP1 activity in SFXN1 KO cells.  (Table 4). One IRE of high quality and a second one of low quality are found 517 respectively at the end of the SFXN1 coding sequence and in the 3' UTR (Figure 8)

561
Accumulation of biometals can be detrimental and may promote protein aggregation. Hence,

562
Amyloid beta peptide (A), which forms toxic aggregates in the brain of patients who suffered from iron-dependent regulation of SFXN, as discussed above, and the known regulation of -Syn by IRPs.

575
Hence, an IRE is found in the 5'UTR of -Syn mRNAs and IRP-mediated translational inhibition is levels. However, we did not find putative IRE in SFXN3 transcripts, as stated above.

588
Whether SFXN are able to regulate ferroptosis is also an important concern, because ferroptosis

599
To conclude, further investigations must be undertaken to precisely specify the role of SFXN1 600 and its homologues in brain biometals homeostasis and neurodegeneration.