Biochimica et Biophysica Acta 1863 (2016) 2859–2867

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta

j ourna l homepage: www.e lsev ie r .com/ locate /bbamcr

Erythroid cell mitochondria receive endosomal iron by a“kiss-and-run” mechanism

Amel Hamdi a,b,1, TariqM. Roshan a,1, TanyaM. Kahawita a,b, Anne B.Mason c, Alex D. Sheftel d,e, Prem Ponka a,b,⁎a Lady Davis Institute for Medical Research, Jewish General Hospital, Montreal, Quebec, Canadab Department of Physiology, McGill University, Montreal, Quebec, Canadac Department of Biochemistry, University of Vermont, Burlington, VT, USAd Spartan Bioscience Inc., Ottawa, Ontario, Canadae High Impact Editing, Ottawa, Ontario, Canada

⁎ Corresponding author at: Jewish General Hospital, LResearch, and the Department of Physiology, McGill UnivCatherine, Montréal, QC H3T 1E2, Canada.

E-mail address: prem.ponka@mcgill.ca (P. Ponka).1 Co-first author.

http://dx.doi.org/10.1016/j.bbamcr.2016.09.0080167-4889/© 2016 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 7 July 2016Received in revised form 24 August 2016Accepted 8 September 2016Available online 11 September 2016

In erythroid cells,more than 90%of transferrin-derived iron entersmitochondriawhere ferrochelatase inserts Fe2+

into protoporphyrin IX. However, the path of iron from endosomes to mitochondrial ferrochelatase remains elu-sive. The prevailing opinion is that, after its export fromendosomes, the redox-activemetal spreads into the cytosolandmysteriously finds itsway intomitochondria through passive diffusion. In contrast, this study supports the hy-pothesis that the highly efficient transport of iron toward ferrochelatase in erythroid cells requires a direct inter-action between transferrin-endosomes and mitochondria (the “kiss-and-run” hypothesis). Using a novelmethod (flow sub-cytometry), we analyze lysates of reticulocytes after labeling these organelles with differentfluorophores. We have identified a double-labeled population definitively representing endosomes interactingwith mitochondria, as demonstrated by confocal microscopy. Moreover, we conclude that this endosome-mitochondrion association is reversible, since a “chase” with unlabeled holotransferrin causes a time-dependentdecrease in the size of the double-labeled population. Importantly, the dissociation of endosomes frommitochon-dria does not occur in the absence of holotransferrin. Additionally, mutated recombinant holotransferrin, that can-not release iron, significantly decreases the uptake of 59Fe by reticulocytes and diminishes 59Fe incorporation intoheme. This suggests that endosomes, which are unable to provide iron to mitochondria, cause a “traffic jam” lead-ing to decreased endocytosis of holotransferrin. Altogether, our results suggest that a molecular mechanism existsto coordinate the iron status of endosomal transferrinwith its trafficking. Besides its contribution to thefield of ironmetabolism, this study provides evidence for a new intracellular trafficking pathway of organelles.

© 2016 Elsevier B.V. All rights reserved.

Keywords:IronTransferrinEndosomesMitochondriaErythroid cells

1. Introduction

Iron is a transition metal whose properties make it useful to vital bi-ologic processes in virtually all living organisms. Essential cellular func-tions include use of iron for oxygen transport, electron transfer, DNAsynthesis and innumerable other purposes [1–4]. However, under phys-iological conditions iron is virtually insoluble and can be extremely toxicwhen not properly shielded [5–7]. Hence, iron plays a role in the forma-tion of toxic oxygen radicals that damage various cell structures. There-fore, it is critical that we gain a better understanding of themechanismsinvolved in normal and abnormal intracellular iron trafficking.

ady Davis Institute for Medicalersity, 3755 Chemin Côte-Ste-

Developing erythroid cells, which have a capacity to obtain and pro-cess ironwith astonishing efficacy [8], are capable of concentrating iron(in the formof heme in the red cell) to approximately 7000-foldwhat ispresent in the plasma (as diferric transferrin [Fe2-Tf]). Additionally, thedelivery of iron into hemoglobin occurs extremely efficiently, since ma-ture erythrocytes contain about 45,000-fold more heme iron than non-heme iron [9]. These facts suggest that in erythroid cells the iron trans-port and heme biosynthetic machineries are fully integrated and arepart of the samemetabolic pathway that leads to a remarkably efficientproduction of heme. Delivery of iron to these cells occurs following thebinding of Fe2-Tf to its cognate receptors (TfR) on the cell membrane.The Fe2-Tf-TfR complexes are then internalized via endocytosis, andiron is released by a process which includes both acidification and theparticipation of the TfR [1,9–12]. Iron, following its reduction to Fe2+

by Steap3 (six-transmembrane epithelial antigen of the prostate [13]),likely together with ascorbate [14] is then transported acrossthe endosomal membrane by the divalent metal transporter 1, DMT1/Nramp2 [15].

2860 A. Hamdi et al. / Biochimica et Biophysica Acta 1863 (2016) 2859–2867

Following its egress from endosomes, iron is transported to intracel-lular sites of use and/or storage in ferritin, but this aspect of ironmetab-olism remains elusive or is at best controversial. It has been commonlybelieved that a low molecular weight intermediate chaperones ironfrom endosomes to mitochondria and other sites of utilization [16].However, this long sought iron binding intermediate, thatwould consti-tute the labile iron pool, has never been identified. In fact, earlier workfrom this laboratory demonstrated [17] that there was virtually nolow molecular weight iron pool in hemoglobin-synthesizing cells andthat iron present in this pool does not behave as an intermediate, thuscorresponding to an end product. Although it has recently been pro-posed that the ferrous iron in this pool is associated with glutathione[18], this conclusion was only based on chemical interactions. Hence,the relevance of this study to the physiological mechanisms involvedin the intracellular iron transport remains in doubt. Additionally, it hasbeen proposed that human poly(rC)-binding protein 1 (PCBP1) “chap-erones” iron into ferritin and non-heme iron proteins [19,20]. On theother hand, more recent reports [21,22] have proposed PCBP2, ratherthan PCBP1, to play an “iron-chaperone” role. This diversity is probablydue to different cell types and/or experimental strategies used. Howev-er, no direct evidence has been provided that PCBPs acquire iron fromtransferrin-endosomes, carry it in cytosol and, more crucially, deliverit to mitochondria.

In erythroid cells, more than 90% of iron enters mitochondria whereferrochelatase, the enzyme that inserts Fe2+ into protoporphyrin IX(PPIX), resides in thematrix side of the innermitochondrial membrane.Importantly, in these cells, strong evidence exists for specific targetingof iron toward mitochondria. This mitochondria-directed transport isdemonstrated in hemoglobin-synthesizing cells, in which iron acquiredfrom Tf continues to flow intomitochondria evenwhen the synthesis ofPPIX is suppressed either experimentally [17,23–25] or in hereditarysideroblastic anemia caused by defects in heme synthesis [26–28]. Ofnote, a compelling candidate for the transport of iron through themito-chondria towards ferrochelatase, mitoferrin, has been identified [29].

Based on the above considerations,we have formulated a hypothesisthat in erythroid cells a transient mitochondrion-endosome interactionis involved in iron translocation to its final destination [17,30]. We havecollected the following experimental evidence to support this hypothe-sis: 1) iron, delivered to mitochondria via the Tf-TfR pathway, is un-available to cytoplasmic chelators [31]; 2) Tf-containing endosomesmove to and contact mitochondria [31] in erythroid cells; 3) endosomalmovement is required for iron delivery to mitochondria [31,32]. Wehave also demonstrated that “free” cytoplasmic iron is not efficientlyused for heme biosynthesis and that the endosome-mitochondrion in-teraction increases chelatable mitochondrial iron [31].

As already mentioned, the substrate for the endosomal transporter,DMT1, is Fe2+, the redox form of iron which is also the substrate forferrochelatase. These facts make our hypothesis quite attractive, since

Fig. 1. 3D confocal microscopy demonstrates the association of transferrin-endosomes (green)mitochondria pairs are shown in three different orientations.

the “chaperone”-like function of endosomes may be one of the mecha-nisms that keeps the concentrations of reactive Fe2+ at extremely lowlevels in the oxygen-rich cytosol of erythroblasts and reticulocytes,preventing ferrous iron's participation in a dangerous Fenton reaction.

There is accumulating evidence for transfer ofmetabolites directly be-tween interacting organelles [33–36]. However, the role of endosomes indistributing intracellular iron is accepted without enthusiasm [20],misinterpreted [16] or simply ignored [37]. Hence,wedeemed it essentialto seek further evidence for our “kiss-and-run” hypothesis.

In the current study, we used 3D live confocal imaging of reticulo-cytes following their incubation with MitoTracker Deep Red (MTDR)and Alexa Green Transferrin (AGTf) and demonstrated transient inter-actions of endosomes with mitochondria. We also demonstrate theseinteractions by a novel method exploiting flow sub-cytometry to ana-lyze reticulocyte lysates labeled with MTDR and AGTf. This strategyidentified a population double-labeled with both fluorescent markersrepresenting endosomes interacting with mitochondria. FACS sortingfollowed by 2D confocal microscopy confirmed the association of bothorganelles in the double-labeled population. The results of these exper-iments aswell as those exploiting amutated recombinant Fe2-Tf unableto release iron efficiently further support the hypothesis that the effi-cient iron transfer to PPIX requires a transient interaction between theendosome and the mitochondrion.

2. Results

2.1. Three-dimensional confocal microscopy of live cells revealsendosome-mitochondria interactions

Using two-dimensional (2D) confocal microscopy, we have previ-ously shown that endosomes come in contact with mitochondria [31].To avoid any pseudo-interactions of these two organelles, i.e., the endo-some hovering over themitochondria, we have used a QuorumWaveFxspinning disc confocal system to capture three-dimensional (3D) livecell imaging of reticulocytes with fluorescently labeled mitochondriaand endosomes. An individual cell at different time intervals is present-ed in three different orientations (Fig. 1, A–C) and shows the proximityof endosomes and mitochondria at different angles (Link for Video;please see Supplementary Material).

2.2. Development of the flow sub-cytometry method to studyendosome-mitochondria interactions

We developed a new method, “flow sub-cytometry”, based on thestandard flow cytometry method, to analyze lysates obtained from re-ticulocytes with fluorescently labeled mitochondria (MTDR) andendosomes (AGTf). Employing this strategy, three distinct populations:endosomes, mitochondria, and a population double labeled with both

with mitochondria (red). At one time interval (“3”, marked by white arrows), endosome-

2861A. Hamdi et al. / Biochimica et Biophysica Acta 1863 (2016) 2859–2867

fluorescentmarkers have been identified (Fig. 2A). To confirm the inter-action between endosomes and mitochondria, we used FACS-sorting toisolate the population double-labeled with both fluorescent markers aswell as MTDR-labeled population (as a negative control), followed by2D examination using a confocal fluorescencemicroscope. Amultichan-nel protocol was used to scan the samples for AGTf and MTDR labeledendosomes and mitochondria, respectively. The colocalization of thetwo organelles was demonstrated by overlapping the green and thered images of endosomes and mitochondria, where the merged imagesrepresent both organelles together (Fig. 2B). This experiment has un-equivocally demonstrated the presence of the Tf-endosome interactingwith the mitochondrion in the double-labeled population (UR). Impor-tantly, the LR population contains exclusively MTDR-labeled mitochon-dria and the LL population is devoid of any fluorescence. Moreover, wehave shown that the size of the double-labeled population representingendosomes associatedwithmitochondria (UR quadrant) increases withincubation time with AGTf and reaches a plateau in ~20 min (Fig. 3).

A

B MTDR AGTf Merge

MTDR AGTf Merge

M

EM

UR(EM)

LR(M)

E

Fig. 2. Flow sub-cytometry method. (A) Using FACS analysis, four different populations inlysates of reticulocytes, previously labeled with MTDR and AGTf for 20 min, wereidentified: The lower left (LL) quadrant displays events devoid of all fluorescence (celldebris). The upper left (UL) quadrant displays events labeled only with AGTf and,therefore, represents the endosomal population. The lower right (LR) quadrant displaysevents labeled only with MTDR and, therefore, corresponds to a mitochondrial population(M). The upper right (UR) quadrant, labeled with both AGTf and MTDR, is referred to asthe double-labeled population representing endosomes interacting with mitochondria(EM). (B) Populations present in UR, LR and LL quadrants were isolated by FACS-sortingand then examined by 2D confocal microscopy. UR is composed only of AGTf-labeledendosomes that colocalize with MTDR-labeled mitochondria. LR contains exclusivelyMTDR-labeled mitochondria. The confocal images show four separate structures. Scalebar is 5 μm.

2.3. Holotransferrin causes the detachment of endosomes frommitochondria

To investigate whether the association of endosomes with mito-chondria is reversible, we incubated reticulocytes with both fluorescentmarkers for 20 min, washed the cells and reincubated with or withoutholotransferrin (Fe2-Tf) for 60 min. The presence of non-fluorescentFe2-Tf caused a time-dependent decrease in the size of the double-labeled population (Fig. 4A and B). Hence, AGTf-endosomes associatedwith mitochondria are replaced by incoming non-fluorescent Fe2-Tf-endosomes, indicating the reversibility of the interaction. AGTf-endosomes leaving mitochondria cannot be detected by FACS becauseof their small size. Importantly, this finding, together with the fact thatendosomes remain bound tomitochondria in the absence of Fe2-Tf, sug-gests that the Tf-endosome drives the detachment.

2.4. Intra-endosomal transferrin saturation governs interorganellarassociation

We investigated whether the degree of endosomal Tf saturationwith iron affects the dynamics of the endosome-mitochondrion associ-ation.When bafilomycin A1, a specific V-ATPase inhibitor known to de-crease acidification of endosomes [38], is added to reticulocytes, the sizeof the double-labeled population dramatically decreased (Fig. 5A).Bafilomycin A1 has been shown to block the incorporation of 59Fe intoheme from 59Fe-Tf-endosomes [32]. Additionally, hemin treatment ofreticulocytes also considerably decreased the size of the double-labeled population (Fig. 5B). Heme has been shown to feedback inhibitthe release of iron from Tf within reticulocyte endosomes [39] and, sim-ilarly as bafilomycin A1, can be expected to increase endosomal Tfsaturation.

The most likely explanation for these observations is thatbafilomycin A1 as well as hemin block Fe2+ export from endosomes al-ready associated with mitochondria, extending their residence time onthese organelles. This causes a “traffic jam” that decelerates the move-ment of endosomes containing fluorescent Tf towards mitochondria.Furthermore, the presence of heme during the reincubation of double-labeled reticulocyteswith unlabeled Fe2-Tf slowed down the disappear-ance of the double population (Fig. 5C), suggesting that the efficient dis-sociation of endosomes requires them to be iron-free.

2.5. Calmodulin antagonist (W-7) decreases the size of double-labeledpopulation

N-(6-aminohexyl)-5-chloro-1-naphthalene-sulfonamide (W-7), acalmodulin antagonist, is known to inhibit the microfilament motor,myosin, which is involved in intracellular movement of endosomesand causes significant inhibition of 59Fe incorporation from 59Fe-endosomes into heme [32]. A positive experiment presented in Fig. S2shows that W-7, as expected, inhibits dramatically the association ofendosomes with mitochondria.

2.6. Transferrinmutant unable to release iron (Holo-rTf) impairs the uptakeof 59Fe from 59Fe2-Tf

To assess whether endosomes unable to release iron affect intracel-lular iron trafficking, reticulocytes were preincubated with Holo-rTf at37 °C for 30 min, washed with cold PBS and incubated with 59Fe-Tf forthe indicated time intervals. After washing with cold PBS, total celland heme 59Fe-radioactivities were measured. The presence of Holo-rTf during incubation significantly inhibited 59Fe uptake by reticulocytesfrom wild-type 59Fe-Tf and also diminished 59Fe incorporation intoheme (Fig. 6). This result suggests that the failure to release iron fromHolo-rTf interferes with the process of endocytosis or endosomal traf-ficking, causing a decrease in the total 59Fe uptake by the reticulocytesas compared to controls.

0

0.31%

1

10.01%

2

16.10%

5

23.31%

10

27.75%

20

29.68%

30

31.47%

Time (min)

Mitochondria (MTDR)

En

do

som

es (

AG

Tf)

A

B

0

5

10

15

20

25

30

35

0 5 10 15 20 25 30

Per

cen

tag

e o

f to

tal p

op

ula

tio

n

Time (min)

Fig. 3. Flow sub-cytometry analysis of the endosome-mitochondria interaction. (A) Percentage ofmitochondria associatedwith endosomes increaseswith time. (B)Graphical presentationof results shown in (A). Error bars were determined from triplicate values.

2862 A. Hamdi et al. / Biochimica et Biophysica Acta 1863 (2016) 2859–2867

2.7. Endosomes interact with mitochondria also in erythroid progenitors

In this experiment, we used murine fetal liver (FL) cells, primarymouse erythroid cells that closely recapitulate several aspects of termi-nal maturation in vivo, in particular very active hemoglobinization [8]We deemed it important to ascertain whether endosomes interactwith mitochondria only in differentiating erythroid cells or if this phe-nomenon operates also before the onset of hemoglobin synthesis. Aspredicted, subjecting these cells to flow sub-cytometry as performedabove with reticulocytes, revealed a large pool of the double-labeledpopulation in FL cells induced to differentiate by erythropoietin. Impor-tantly, we also observed that the endosome-mitochondria population,although smaller than in erythropoietin-induced FL cells, is present al-ready in undifferentiated cells (Fig. S3).

3. Discussion

During differentiation, immature erythroid cells acquire vast amountsof iron at a “breakneck” rate. Proper coordination of iron delivery andutilization in heme synthesis is essential; disruption of this process likelyunderlies iron loading disorders such as sideroblastic anemia andmyelodysplastic syndrome with ringed sideroblasts [28]. On a per-cell

basis, the rate of heme synthesis in developing erythroid cells is at leastan order of magnitude higher than in the liver, the second highest hemeproducer in the body.

As described in the Introduction,mostmammalian cells take up ironvia receptor-mediated endocytosis by the TfR. The prevailing opinion isthat after its escape from endosomes, iron spreads into the cytosol [40]andmysteriously finds its way intomitochondria. Based on the fact thatTf-bound iron is extremely efficiently used for hemoglobin synthesis,targeted into erythroid mitochondria, and that no cytoplasmic irontransport intermediate has ever been identified in erythroid cells, wehave suggested a new hypothesis of intracellular iron transport [17,30]. This model proposes [17] that after iron is released from Tf in Hb-synthesizing cells, it would bypass the cytosol and be directly trans-ferred from protein to protein (probably in hydrophobic environments)until it reaches ferrochelatase in the mitochondrion.

Previouswork fromour laboratory has suggested that iron is directlytransferred from endosomes to mitochondria [31,32] likely throughinterorganellar interaction. However, this model has not been acceptedby many [16,20]. In the present study, we used 2D and 3D confocal mi-croscopy and developed a novel experimental strategy based on flowcytometry analysis of fluorescently labeled organelles (“flow sub-cyometry”) to provide evidence for our hypothesis.

B

A

Fig. 4.Holotransferrin causes the detachment of endosomes frommitochondria. (A) Flow sub-cytometry analysis of reticulocytes preincubated (20min)withMTDRand AGTf, followedbytheir incubation (60min)with orwithout Fe2-Tf. (B)Graphical presentation of the results shown in (A) as the percentage decrease from100% (Time 0, the start of the “chase” experiment)in the presence or absence of holotransferrin as compared to control. ■, Fe2-Tf; ▲, minus Tf. Error bars were determined from triplicate values.

2863A. Hamdi et al. / Biochimica et Biophysica Acta 1863 (2016) 2859–2867

Employing 3D confocal system, we acquired video images, allowingus to capture fluorescent Tf-endosomes interacting with labeled mito-chondria in space, excluding the possibility of “pseudo-interactions”caused by a coincidental overlapping. This technique very clearly re-vealed contacts of endosomes with mitochondria in 3D space (Fig. 1).

Using a flow sub-cytometry to analyze lysates obtained from reticulo-cytes with fluorescently labeled mitochondria (MTDR) and endosomes(AGTf), we have identified three distinct populations: endosomes, mito-chondria, and a population double-labeledwith both fluorescentmarkers(Fig. 2A). The double-labeled population was sorted by FACS and shownby2Dconfocalmicroscopy to be composed of endosomes associatedwithmitochondria (Fig. 2B). The size of the double-labeled population in-creases with the incubation time and reaches a plateau at approximately20 min (Fig. 3).

The reversibility of the endosome-mitochondria association wasdemonstrated by the experiment inwhichwe first preincubated reticu-locyteswith both fluorescentmarkers for 20min followed by their incu-bation (“chase”) with or without unlabeled holotransferrin (Fig. 4A andB). The presence of unlabeled Fe2-Tf caused a time-dependent decreasein the size of the double-labeled population indicating the reversibilityof the endosome-mitochondria contact. Importantly, in the absence ofholotransferrin, endosomes remain associated with mitochondria.Taken together, our finding indicates that Tf-endosome drives thedetachment.

Both bafilomycin A1 and heme interferewith iron release from Tf andcause a significant decrease in the size of the double-labeled population(Fig. 5A and B). These agents obstruct Fe2+ export from endosomes al-ready associated with mitochondria, extending the residence time ofendosomes onmitochondria, thus blocking themovement of endosomescontaining fluorescent Tf towards mitochondria. Additionally, heminslows down the dissociation of labeled endosomes from mitochondria(Fig. 5C). Hence, our results suggest that a molecular mechanism existsto coordinate the iron status of endosomal Tf with its trafficking.

Moreover, the presence of a Tf mutant unable to release iron (Holo-rTf) impairs the uptake of 59Fe from 59Fe2-Tf (Fig. 6), suggesting that ef-ficient endocytosis and/or endosomal trafficking requires the transfer ofiron from endosomes to mitochondria. In other words, the endosomalmitochondrial interface plays a role in regulating the cellular iron up-take, probably by controlling Tf endocytosis.

Our finding of endosomes interacting with mitochondria in undiffer-entiated fetal liver cells (Fig. S3) is of considerable importance. It suggeststhat the endosome-mitochondrion interaction may be a common mech-anism involved in iron delivery for heme synthesis. However, since FLprogenitors are committed to synthesize hemoglobin, further experi-ments with non-erythroid cells are needed to establish whether thekiss-and-run mechanism is generally involved in the transport of Tf-borne iron to mitochondria. If the so-called kiss-and-run mechanism ofiron delivery tomitochondria proves to be ubiquitous, thenmitochondria

A

B

C

0

5

10

15

20

25

30

35

0 5 10 15 20

Per

cen

tag

e o

f to

tal p

op

ula

tio

n

Time (min)

Control

Bafilomycin A1

Control

Hemin

0

10

20

30

40

50

60

0 5 10 15 20 25 30

Per

cen

tag

e o

f to

tal p

op

ula

tio

n

Time (min)

Control

Hemin

0

20

40

60

80

100

0 10 20 30 40 50 60

Per

cen

tag

e o

f d

ou

ble

-lab

eled

p

op

ula

tio

n

Time (min)

Fig. 5. Bafilomycin A1 and hemin decrease the association of endosomeswithmitochondria.(A) The percentage of the double-labeled population after treatment with bafilomycin. ■,Control; ♦, bafilomycin A1 (100 nM). Results presented are representative of 6 replicateexperiments. (B) The percentage of the double-labeled population after treatment withhemin. ■, Control; ▲, hemin (5 μM). Results presented are representative of 4 replicateexperiments. (C) Hemin prevents the dissociation of labeled endosomes frommitochondria. ■, Control; ▲, hemin (5 μM). Error bars were determined from triplicatevalues.

CP

M /

500

mill

ion

ret

icu

locy

tes

TotalHeme

Time (min)fTr-oloHlortnoC

0

500

1000

1500

2000

2500

3000

3500

4000

4500

5000

5500

6000

6500

7000

0 5 10 20 0 5 10 20

***

**

Fig. 6.Holo-rTf inhibits 59Fe incorporation into reticulocytes and heme from 59Fe2-Tf. Errorbars were determined from triplicate values. ⁎⁎ p ≤ 0.01; ⁎⁎⁎ p ≤ 0.001.

2864 A. Hamdi et al. / Biochimica et Biophysica Acta 1863 (2016) 2859–2867

may be the entry site of iron into the cell. In this context it needs to bepointed out that the only known function of internalized Tf is to deliveriron to its intracellular destination [31]. Hence, the described endosomemitochondria association facilitates interorganellar iron transfer.

Our study is relevant to sideroblastic anemias, a heterogeneousgroup of disorders, characterized by mitochondrial iron overload in de-veloping red blood cells [26–28]. The unifying characteristic of allsideroblastic anemias is the ring sideroblast, a pathological erythroidprecursor containing excessive deposits of non-heme iron inmitochon-dria with perinuclear distribution creating a ring appearance.

Themajority of congenital sideroblastic anemias are X-linked;most ofthem are caused bymutations in the gene encoding erythroid-specific 5-aminolevulinate synthase [26–28]. However, there is a certain proportionof patients with hereditary sideroblastic anemia who exhibit autosomal

recessive inheritance; some such patients have a defect in the geneencoding the erythroid-specific mitochondrial carrier protein, SLC25A38[41]. It has been suggested that SLC25A38 may translocate glycine intomitochondria, a proposal recently confirmed by Fernández-Murrayet al. [42] and Lunetti et al. [43].

In our opinion, at least four factors are responsible for the pathogen-esis of ring sideroblast formation in congenital sideroblastic anemias:(i) As extensively discussed in this report, the efficient delivery of ironto ferrochelatase in erythroid cells requires the direct interaction ofendosomes with mitochondria. (ii) Mitochondrial iron cannot be ade-quately utilized due to the lack of PPIX. (iii) Erythroid cells, but notnon-erythroid cells, are equipped with a negative feedback mechanismin which ‘uncommitted’ heme inhibits iron acquisition from Tf [30]. Al-though themolecularmechanism of this inhibition is still unknown, thelack of heme plays an important role in mitochondrial iron accumula-tion in erythroid cells. (iv) Iron can leave erythroid mitochondria onlyafter being inserted into PPIX [24].

As already alluded to, there is no consensus regarding the mecha-nism of iron transfer from plasma Tf to mitochondria. One of the pro-posed intermediates in the intracellular iron path is the iron storageprotein, ferritin [44,45]. Nevertheless, a strong argument against thisclaim was collected and reported [46]. Somewhat surprisingly, it wassuggested in a recent report [47] that the autophagic turnover of ferritinvia NCOA4 is a critical process for regulating intracellular iron bioavail-ability. However, Darshan et al. [48] demonstrated that the conditionaldeletion of the ferritin heavy chain in adult mice did not cause any de-crease in hematocrit or hemoglobin levels, strongly indicating that ferri-tin is not involved in hemoglobinization. In this context it is pertinent tomention that the nuclear receptor coactivator 4 (NCOA4), also known asARA70, was initially identified as a specific coactivator for the androgenreceptor in human prostate cells [49]. Additionally, another study [50]has provided evidence that NCOA4 knockout is not lethal and NCOA4-null mice have only a mild microcytic hypochromic anemia. These find-ings concur with our previous and current studies demonstrating thatthe efficient utilization of iron for hemoglobin synthesis requires a di-rect contact of endosomes with mitochondria.

Finally, recent experimental support for the kiss-and-run hypothesiscame from a study that investigated the interaction between several mi-tochondrial heme biosynthetic enzymes and proteins involved in iron ho-meostasis [51]. This study revealed a partnership between ferrochelatase(amitochondrial protein) and TfR (an endosome associated protein) thatmay bemediated by a yet to be identified bridge protein. It will be worthexamining whether one of the PCBPs could serve as such protein [19,21,52]. Medlock et al. [45] further state, “(their) data…support a model oftransient protein-protein interactions within a dynamic proteincomplex.” Our model, shown in Fig. 7, suggests a potential outer

Fig. 7. Scheme of iron delivery to mitochondria for heme biosynthesis in erythroid cells. Iron is released from transferrin within endosomes by a combination of Fe3+ reduction by Steap3(likely when it is still bound to the transferrin receptor) and a decrease in pH (~pH 5.5); following this, Fe2+ is transported through the endosomal membrane by DMT1. Our modelproposes that DMT1 conveys Fe2+ to an outer mitochondrial membrane protein. Our preliminary studies suggest that the voltage-dependent anion channel (VDAC) interacts with theintracellular loops of DMT1 (Hamdi A, Roshan TM, Sheftel AD, Ponka P. Interaction of transferrin-endosomes with mitochondria: Implications for iron transport to ferrochelatase inerythroid cells [abstract]. Blood, 2015; 126(23): 407). VDAC, or possibly another protein, then delivers Fe2+ to mitoferrin 1 (Mfrn1) which, in turn, supplies Fe2+ to ferrochelatase (FC)that inserts it into protoporphyrin IX (PPIX). Maroon and red balls stand for ferric and ferrous iron, respectively. ALA, 5-aminolevulinic acid; ALA-S2, erythroid-specific ALA synthase.

2865A. Hamdi et al. / Biochimica et Biophysica Acta 1863 (2016) 2859–2867

mitochondrial membrane protein(s) that may interact with the intracel-lular loop of DMT1 for the delivery of iron to ferrochelatase.

In summary, our study provides strong evidence that the highly effi-cient delivery of iron to heme in hemoglobin-synthesizing cells requiresthe direct interaction of transferrin-endosomes with mitochondria. Tothe best of our knowledge, this is the first evidence of the interorganellarinteractionhaving the functional consequence of facilitating the transportof amicronutrient. A full understanding of themolecularmechanisms in-volved in the endosome-mitochondria systemwill require the identifica-tion of the “signal(s)” that direct iron-carrying endosomes towardsmitochondria, the players involved in the docking of endosomes tomito-chondria and the “signal(s)” that determine the detachment of iron-freeendosomes from mitochondria. We believe that the current work pro-vides a strong platform to pursue such studies.

4. Materials and methods

4.1. Materials

MitoTracker Deep Red 633 (MTDR) and Alexa Green Transferrin 488(AGTf) were purchased from Molecular Probes (Invitrogen, Carlsbad,CA). Protease inhibitor was purchased from Roche (Penzberg,Germany). Pronase, heparin and bafilomycin A1 were purchased fromBioshop (Burlington, Ontario). Holotransferrin (Fe2-transferrin),avertin, phenylhydrazine, thiozole orange and N-(6-Aminohexyl)-5-chroro-1-napthalenesulfonamide hydrochloride (W7) were purchasedfrom Sigma-Aldrich (St. Louis, MO). Fluorescent mounting media werepurchased from Dako (Glostrup, Denmark). Microslides and micro cov-erslipswere purchased fromVWR (West Chester, PA).MinimumEssen-tial Media (MEM) were purchased from Invitrogen (Gibco® LifeTechnologies™, Carlsbad, CA) supplementedwith 25mMHEPES. Mito-chondrial isolation buffer contains 220 mM mannitol, 60 mM sucrose,20 mM HEPES, pH 7.4, 80 mM KCl, 2 mM magnesium acetate, 0.5 mMEGTA and protease inhibitors. The human mutant recombinant Fe2-Tfunable to release iron (K206E/K534A) [53,54] was generated in Dr.Mason's laboratory.

4.2. Animals

All the animal studies performed in this work were approved by theLady Davis Institute animal care committee following the guidelines ofthe Canadian Council on Animal Care.

CD1 and C57BL/6 mice (females; 6 months old) retired breederswere obtained from Charles River Laboratories (Wilmington, MA) andwere housed in microisolator cages.

4.3. Isolation of reticulocytes

Adult, female CD1 female mice were injected with neutralizedpheynylhydrazine intraperitoneally at the dose of 50 mg/kg/day forthree consecutive days. Mice were then anesthetized after 2–3 days fol-lowing the last injection by using 1 mL of 2.5% avertin. Blood was col-lected via direct cardiac puncture using heparinized syringe andresuspended in cold PBS. Cells were then washed with ice-cold PBSthree times. By using this technique we obtained 30–45% reticulocytesas verified by thiozole orange staining.

4.4. Cell culture of primary mouse fetal liver cells

Primary erythroid cells were cultured as described [8]. Briefly, cellswere grown from FL cells that were obtained from embryonic day12.5 embryos of wild type C57BL/6 mice. FL cells were grown inserum-free StemPro-34mediumplus Nutrient Supplement (Invitrogen,Carlsbad, CA) and 2 mM L-glutamine, with 100 ng/mLmurine recombi-nant stem cell factor (SCF), 10−6 M synthetic glucocorticoid dexameth-asone (Dex), 40 ng/mL insulin-like growth factor 1 (IGF-1), 0.2% (vol/vol) gentamicin, 0.2% amphotericin, and 2 U/mL human recombinanterythropoietin (Epo). FL cells were kept either in an undifferentiatedstate or induced to terminal differentiation (48 h) by re-suspendingthem in differentiationmedium (StemPro-34) containing Nutrient Sup-plement, 10 U/mL Epo, 2 mM L-glutamine, 4 × 10−4 IU/mL insulin,3 × 10−6 M RU486, and 1 mg/mL iron-saturated human transferrin(Fe2-Tf; Sigma, St. Louis, MO).

2866 A. Hamdi et al. / Biochimica et Biophysica Acta 1863 (2016) 2859–2867

4.5. Confocal microscopy

Reticulocytes were labeled with MTDR and AGTf and kept at 4 °Cuntil their analysis by spinning disc confocal microscopy. The tempera-ture of the chamber was kept at 37 °C. One-milliliter samples of reticu-locytes suspension were resuspended in 50 μL of medium withoutphenol red. Cells were then mounted on a pre-warmed (30 °C) heatedstage and were allowed to equilibrate for 15 min before imaging. Retic-ulocytes were not kept for more than 10 min in the chamber after ex-posing to laser. Three-dimensional images were acquired over time.Optical sections of 0.5 μm were acquired in z-plane through a ±2.5-μm (total of 5.0-μm stack). Single endosome was tracked in a cell in dif-ferent planes from the start of endocytosis to its interaction with mito-chondria and leaving in a different plane. Microscopy was performedwith a Zeiss LSM 5 Pascal confocal inverted microscope. Live cell imag-ing of reticulocytes was done with Wave FX Spinning Disc Confocal(SDC) microscopy by Quorum Technologies Inc. (Guelph, Ontario). Ve-locity 3D Image Analysis Software from PerkinElmer was used to ana-lyze live cell imaging. Rapid imaging of these instruments makes itpossible to determine the 3D trajectory of an intra-cellular object.

4.6. Flow sub-cytometry

Reticulocytes were washed three times with ice-cold PBS, spundown at 800 g for 5min and pellet was re-suspended inMEM. Reticulo-cytes were then incubated with 500 nMMTDR for 30 min at 37 °C in ashaking water-bath following which they were washed 3-times withice-cold PBS and centrifuged at 800g for 5 min at 4 °C. Cells were re-suspended in MEM containing 500 nM AGTf and kept on ice for30 min in dark. After incubating with AGTf samples were divided andincubated at 37 °C for the indicated time intervals. The samples werethen immediately placed on ice and washed three-times with ice-coldPBS to stop endocytosis. The samples were treated with pronase(1mg/mL) on ice to remove any fluorescent probe bound to TfR presenton the cell surface. The cells were then washed three-times with ice-cold PBS, re-suspended in 1 mL of mitochondrial isolation buffer andthen subjected to four cycles of rapid freeze (liquid nitrogen) andthaw. The samples were spun at 800 g for 10 min at 4 °C and superna-tants were analyzed by flow cytometer (Fig. S1). Fluorescence signalswere determined using Becton Dickinson FACS Caliber (BD). AGTf 488and MTDR 633 were determined by using a 530/30 and a 660/20 bandpass filter, respectively. Data was analyzed using FCS Express V3 soft-ware. Double labeled population were also sorted on a FACSAria sorter(BD) followed by 2D examination in confocal microscopy. The effectsof bafilomycin A1, hemin and W7 were examined by their addition toMTDR pre-labeled reticulocytes followed by the incubation with AGTf.

4.7. Iron uptake and incorporation into heme

59Fe2-Tf was made from 59FeCl3 (PerkinElmer, Santa Clara, USA;2 μCi) as described previously [55,56]. The acid precipitation methodwas used tomeasure 59Fe in heme and non-heme fractions as describedpreviously [23,31,32]. Reticulocytes were collected by centrifugation(4000g for 30 s), lysed in water and then boiled in 1 mL of 0.2 M HCL.Samples were the transferred to an ice-bath and 59Fe-heme-containingproteins were precipitated with ice-cold 7% trichoroacetic acid (TCA;Bioshop Canada Inc., Burlington, Canada) solution, for 2 h and collectedby centrifugation (2800g for 5 min at 4 °C). Precipitated proteins (con-taining 59Fe in heme) and supernatants (containing non-heme 59Fe)were separated in different tubes and radioactivity in heme and non-heme fraction was measured in a Packard Cobra gamma counter(PerkinElmer, Santa Clara, USA.

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.bbamcr.2016.09.008.

Author contributions

A.H. and T.M.R. performed research, analyzed data, and wrote themanuscript; T.M.K. performed research and analyzed data; A.B.M. andA.D.S. analyzed data, and wrote the manuscript; and P.P. conceivedthe study, analyzed data, and wrote the manuscript.

Conflict of interest

The authors declare no competing or financial interests.

Funding

The research of P.P. was funded by grants from the Canadian Insti-tutes of Health Research (MOP-14100 and MOP-126064).

Transparency document

The Transparency document associated with this article can befound, in online version.

Acknowledgments

The authors are grateful to Dr. Heidi McBride for her valuable adviceand helpful suggestions during this research. The authors also thankDrs. Julie Desbarats, Daniel Garcia-Santos, Matthias Schranzhofer,MarcMikhael and Volker Blank for their useful comments and ChristianYoung and Dr. Lilian Amrein for their technical assistance with FACSanalyses and confocal microscopy.

References

[1] M.W. Hentze, M.U. Muckenthaler, B. Galy, C. Camaschella, Two to tango: regulationof mammalian iron metabolism, Cell 142 (2010) 24–38.

[2] A.D. Sheftel, A.B. Mason, P. Ponka, The long history of iron in the universe and inhealth and disease, Biochim. Biophys. Acta 1820 (2012) 161–187.

[3] S. Guo, D.M. Frazer, G.J. Anderson, Iron homeostasis: transport, metabolism, and reg-ulation, Curr. Opin. Clin. Nutr. Metab. Care (2016).

[4] N.C. Andrews, Forging a field: the golden age of iron biology, Blood 112 (2008)219–230.

[5] M.D. Scott, J.W. Eaton, F.A. Kuypers, D.T. Chiu, B.H. Lubin, Enhancement of erythro-cyte superoxide dismutase activity: effects on cellular oxidant defense, Blood 74(1989) 2542–2549.

[6] P. Brissot, M. Ropert, C. Le Lan, O. Loreal, Non-transferrin bound iron: a key role iniron overload and iron toxicity, Biochim. Biophys. Acta 1820 (2012) 403–410.

[7] G.J. Anderson, F. Wang, Essential but toxic: controlling the flux of iron in the body,Clin. Exp. Pharmacol. Physiol. 39 (2012) 719–724.

[8] M. Schranzhofer, M. Schifrer, J.A. Cabrera, S. Kopp, P. Chiba, H. Beug, E.W. Mullner,Remodeling the regulation of iron metabolism during erythroid differentiation toensure efficient heme biosynthesis, Blood 107 (2006) 4159–4167.

[9] D.R. Richardson, P. Ponka, The molecular mechanisms of the metabolism and trans-port of iron in normal and neoplastic cells, Biochim. Biophys. Acta 1331 (1997)1–40.

[10] I. De Domenico, D. McVey Ward, J. Kaplan, Regulation of iron acquisition and stor-age: consequences for iron-linked disorders, Nat. Rev.Mol. Cell Biol. 9 (2008) 72–81.

[11] D.J. Lane, A.M. Merlot, M.L. Huang, D.H. Bae, P.J. Jansson, S. Sahni, D.S. Kalinowski,D.R. Richardson, Cellular iron uptake, trafficking and metabolism: key moleculesand mechanisms and their roles in disease, Biochim. Biophys. Acta 1853 (2015)1130–1144.

[12] B.E. Eckenroth, A.N. Steere, N.D. Chasteen, S.J. Everse, A.B.Mason, How the binding ofhuman transferrin primes the transferrin receptor potentiating iron release atendosomal pH, Proc. Natl. Acad. Sci. U. S. A. 108 (2011) 13089–13094.

[13] R.S. Ohgami, D.R. Campagna, E.L. Greer, B. Antiochos, A. McDonald, J. Chen, J.J. Sharp,Y. Fujiwara, J.E. Barker, M.D. Fleming, Identification of a ferrireductase required forefficient transferrin-dependent iron uptake in erythroid cells, Nat. Genet. 37(2005) 1264–1269.

[14] D.J. Lane, S. Chikhani, V. Richardson, D.R. Richardson, Transferrin iron uptake is stim-ulated by ascorbate via an intracellular reductive mechanism, Biochim. Biophys.Acta 1833 (2013) 1527–1541.

[15] M.D. Fleming, C.C. Trenor 3rd, M.A. Su, D. Foernzler, D.R. Beier, W.F. Dietrich, N.C.Andrews, Microcytic anaemia mice have a mutation in Nramp2, a candidate irontransporter gene, Nat. Genet. 16 (1997) 383–386.

[16] Z.I. Cabantchik, Labile iron in cells and body fluids: physiology, pathology, and phar-macology, Front. Pharmacol. 5 (2014) 45.

2867A. Hamdi et al. / Biochimica et Biophysica Acta 1863 (2016) 2859–2867

[17] D.R. Richardson, P. Ponka, D. Vyoral, Distribution of iron in reticulocytes after inhibi-tion of heme synthesis with succinylacetone: examination of the intermediates in-volved in iron metabolism, Blood 87 (1996) 3477–3488.

[18] R.C. Hider, X. Kong, Iron speciation in the cytosol: an overview, Dalton Trans. 42(2013) 3220–3229.

[19] H. Shi, K.Z. Bencze, T.L. Stemmler, C.C. Philpott, A cytosolic iron chaperone that de-livers iron to ferritin, Science 320 (2008) 1207–1210.

[20] C.C. Philpott, M.S. Ryu, Special delivery: distributing iron in the cytosol of mamma-lian cells, Front. Pharmacol. 5 (2014) 173.

[21] I. Yanatori, Y. Yasui, M. Tabuchi, F. Kishi, Chaperone protein involved in transmem-brane transport of iron, Biochem. J. 462 (2014) 25–37.

[22] I. Yanatori, D.R. Richardson, K. Imada, F. Kishi, Iron export through the transporterFerroportin 1 is modulated by the iron chaperone PCBP2, J. Biol. Chem. 291(2016) 17303–17318.

[23] J. Borova, P. Ponka, J. Neuwirt, Study of intracellular iron distribution in rabbit retic-ulocytes with normal and inhibited heme synthesis, Biochim. Biophys. Acta 320(1973) 143–156.

[24] P. Ponka, A. Wilczynska, H.M. Schulman, Iron utilization in rabbit reticulocytes. Astudy using succinylacetone as an inhibitor or heme synthesis, Biochim. Biophys.Acta 720 (1982) 96–105.

[25] L.M. Garrick, K. Gniecko, Y. Liu, D.S. Cohan, J.A. Grasso, M.D. Garrick, Iron distributionin Belgrade rat reticulocytes after inhibition of heme synthesis with succinylacetone,Blood 81 (1993) 3414–3421.

[26] S.S. Bottomley, Congenital sideroblastic anemias, Curr. Hematol. Rep. 5 (2006)41–49.

[27] M.D. Fleming, Congenital sideroblastic anemias: iron and heme lost in mitochondri-al translation, Hematol. Am. Soc. Hematol. Educ. Prog. 2011 (2011) 525–531.

[28] M. Cazzola, L. Malcovati, Diagnosis and treatment of sideroblastic anemias: from de-fective heme synthesis to abnormal RNA splicing, Hematol. Am. Soc. Hematol. Educ.Prog. 2015 (2015) 19–25.

[29] G.C. Shaw, J.J. Cope, L. Li, K. Corson, C. Hersey, G.E. Ackermann, B. Gwynn, A.J.Lambert, R.A. Wingert, D. Traver, N.S. Trede, B.A. Barut, Y. Zhou, E. Minet, A.Donovan, A. Brownlie, R. Balzan, M.J. Weiss, L.L. Peters, J. Kaplan, L.I. Zon, B.H.Paw, Mitoferrin is essential for erythroid iron assimilation, Nature 440 (2006)96–100.

[30] P. Ponka, Tissue-specific regulation of iron metabolism and heme synthesis: distinctcontrol mechanisms in erythroid cells, Blood 89 (1997) 1–25.

[31] A.D. Sheftel, A.S. Zhang, C. Brown, O.S. Shirihai, P. Ponka, Direct interorganellartransfer of iron from endosome to mitochondrion, Blood 110 (2007) 125–132.

[32] A.S. Zhang, A.D. Sheftel, P. Ponka, Intracellular kinetics of iron in reticulocytes: evi-dence for endosome involvement in iron targeting to mitochondria, Blood 105(2005) 368–375.

[33] Y. Elbaz-Alon, E. Rosenfeld-Gur, V. Shinder, A.H. Futerman, T. Geiger, M. Schuldiner,A dynamic interface between vacuoles and mitochondria in yeast, Dev. Cell 30(2014) 95–102.

[34] C. Honscher, M. Mari, K. Auffarth, M. Bohnert, J. Griffith, W. Geerts, M. van der Laan,M. Cabrera, F. Reggiori, C. Ungermann, Cellular metabolism regulates contact sitesbetween vacuoles and mitochondria, Dev. Cell 30 (2014) 86–94.

[35] H.M. McBride, Open questions: seeking a holistic approach for mitochondrial re-search, BMC Biol. 13 (2015) 8.

[36] C. Ungermann, vCLAMPs-an intimate link between vacuoles andmitochondria, Curr.Opin. Cell Biol. 35 (2015) 30–36.

[37] A.R. Bogdan, M. Miyazawa, K. Hashimoto, Y. Tsuji, Regulators of iron homeostasis:new players in metabolism, cell death, and disease, Trends Biochem. Sci. 41(2016) 274–286.

[38] M.E. Finbow, M.A. Harrison, The vacuolar H+-ATPase: a universal proton pump ofeukaryotes, Biochem. J. 324 (Pt 3) (1997) 697–712.

[39] P. Ponka, H.M. Schulman, J. Martinez-Medellin, Haem inhibits iron uptake subse-quent to endocytosis of transferrin in reticulocytes, Biochem. J. 251 (1988) 105–109.

[40] A. Jacobs, Low molecular weight intracellular iron transport compounds, Blood 50(1977) 433–439.

[41] D.L. Guernsey, H. Jiang, D.R. Campagna, S.C. Evans, M. Ferguson, M.D. Kellogg, M.Lachance, M. Matsuoka, M. Nightingale, A. Rideout, L. Saint-Amant, P.J. Schmidt, A.Orr, S.S. Bottomley, M.D. Fleming, M. Ludman, S. Dyack, C.V. Fernandez, M.E. Samuels,Mutations inmitochondrial carrier family gene SLC25A38 cause nonsyndromic autoso-mal recessive congenital sideroblastic anemia, Nat. Genet. 41 (2009) 651–653.

[42] J.P. Fernandez-Murray, S.V. Prykhozhij, J.N. Dufay, S.L. Steele, D. Gaston, G.K.Nasrallah, A.J. Coombs, R.S. Liwski, C.V. Fernandez, J.N. Berman, C.R. McMaster, Gly-cine and folate ameliorate models of congenital sideroblastic anemia, PLoS Genet.12 (2016), e1005783.

[43] P. Lunetti, F. Damiano, G. De Benedetto, L. Siculella, A. Pennetta, L. Muto, E. Paradies,C.M. Marobbio, V. Dolce, L. Capobianco, Characterization of human and yeast mito-chondrial glycine carriers with implications for heme biosynthesis and anemia, J.Biol. Chem. (2016).

[44] M.C. Bessis, J. Breton-Gorius, Iron particles in normal erythroblasts and normal andpathological erythrocytes, J. Biophys. Biochem. Cytol. 3 (1957) 503–504.

[45] B. Vaisman, E. Fibach, A.M. Konijn, Utilization of intracellular ferritin iron for hemo-globin synthesis in developing human erythroid precursors, Blood 90 (1997)831–838.

[46] P. Ponka, D.R. Richardson, Can ferritin provide iron for hemoglobin synthesis? Blood89 (1997) 2611–2613.

[47] J.D. Mancias, L. Pontano Vaites, S. Nissim, D.E. Biancur, A.J. Kim, X. Wang, Y. Liu, W.Goessling, A.C. Kimmelman, J.W. Harper, Ferritinophagy via NCOA4 is required forerythropoiesis and is regulated by iron dependent HERC2-mediated proteolysis,Elife 4 (2015).

[48] D. Darshan, L. Vanoaica, L. Richman, F. Beermann, L.C. Kuhn, Conditional deletion offerritin H in mice induces loss of iron storage and liver damage, Hepatology 50(2009) 852–860.

[49] S. Yeh, C. Chang, Cloning and characterization of a specific coactivator, ARA70, forthe androgen receptor in human prostate cells, Proc. Natl. Acad. Sci. U. S. A. 93(1996) 5517–5521.

[50] R. Bellelli, G. Federico, A. Matte, D. Colecchia, A. Iolascon, M. Chiariello, M. Santoro, L.De Franceschi, F. Carlomagno, NCOA4 deficiency impairs systemic iron homeostasis,Cell Rep. 14 (2016) 411–421.

[51] A.E. Medlock, M.T. Shiferaw, J.R. Marcero, A.A. Vashisht, J.A. Wohlschlegel, J.D.Phillips, H.A. Dailey, Identification of the mitochondrial heme metabolism complex,PLoS One 10 (2015), e0135896.

[52] D.J. Lane, D.R. Richardson, Chaperone turns gatekeeper: PCBP2 and DMT1 form aniron-transport pipeline, Biochem. J. 462 (2014) e1–e3.

[53] A.N. Steere, S.L. Byrne, N.D. Chasteen, A.B. Mason, Kinetics of iron release from trans-ferrin bound to the transferrin receptor at endosomal pH, Biochim. Biophys. Acta1820 (2012) 326–333.

[54] A.B. Mason, P.J. Halbrooks, J.R. Larouche, S.K. Briggs, M.L. Moffett, J.E. Ramsey, S.A.Connolly, V.C. Smith, R.T. MacGillivray, Expression, purification, and characterizationof authentic monoferric and apo-human serum transferrins, Protein Expr. Purif. 36(2004) 318–326.

[55] J. Martinez-Medellin, H.M. Schulman, The kinetics of iron and transferrin incorpora-tion into rabbit erythroid cells and the nature of stromal-bound iron, Biochim.Biophys. Acta 264 (1972) 272–274.

[56] P. Ponka, H.M. Schulman, Acquisition of iron from transferrin regulates reticulocyteheme synthesis, J. Biol. Chem. 260 (1985) 14717–14721.