How the binding of human transferrin primes the transferrin … · How the binding of human transferrin primes the transferrin receptor potentiating iron release at endosomal pH Brian

How the binding of human transferrin primesthe transferrin receptor potentiatingiron release at endosomal pHBrian E. Eckenrotha, Ashley N. Steerea, N. Dennis Chasteenb, Stephen J. Eversea, and Anne B. Masona,1

aDepartment of Biochemistry, University of Vermont, 89 Beaumont Avenue, Burlington, VT 05405; and bDepartment of Chemistry, University of NewHampshire, Parsons Hall, Durham, NH 03824

Edited by Pamela J. Bjorkman, California Institute of Technology, Pasadena, CA, and approved June 23, 2011 (received for review April 12, 2011)

Delivery of iron to cells requires binding of two iron-containinghuman transferrin (hTF) molecules to the specific homodimerictransferrin receptor (TFR) on the cell surface. Through receptor-mediated endocytosis involving lower pH, salt, and an unidentifiedchelator, iron is rapidly released from hTF within the endosome.The crystal structure of a monoferric N-lobe hTF/TFR complex(3.22-Å resolution) features two binding motifs in the N lobeand one in the C lobe of hTF. Binding of FeNhTF induces globaland site-specific conformational changes within the TFR ectodo-main. Specifically, movements at the TFR dimer interface appearto prime the TFR to undergo pH-induced movements that alterthe hTF/TFR interaction. Iron release from each lobe then occursby distinctly different mechanisms: Binding of His349 to the TFR(strengthened by protonation at low pH) controls iron release fromthe C lobe, whereas displacement of one N-lobe binding motif, inconcert with the action of the dilysine trigger, elicits iron releasefrom the N lobe. One binding motif in each lobe remains attachedto the same α-helix in the TFR throughout the endocytic cycle.Collectively, the structure elucidates how the TFR accelerates ironrelease from the C lobe, slows it from the N lobe, and stabilizesbinding of apohTF for return to the cell surface. Importantly, thisstructure provides new targets for mutagenesis studies to furtherunderstand and define this system.

Human serum transferrin (hTF) is an 80-kDa bilobal glycopro-tein synthesized by hepatocytes and secreted into the serum

where it binds iron acquired from the diet (1). Ferric iron (Fe3þ)is held extremely tightly within one or both of the homologous Nand C lobes that comprise hTF for transport to cells throughoutthe body (2). Each lobe is composed of two subdomains (N1, N2and C1, C2), forming a cleft within which the hexa-coordinateFe3þ is bound to four amino acid ligands: one histidine, one as-partate, and two tyrosine residues. A synergistic anion, carbonate,anchored by a conserved arginine residue occupies the two re-maining coordination sites in each lobe. Large-scale rigid bodymovements (approximately 50°) of the subdomains are observedwhen each cleft opens and iron is released (3, 4).

Understanding the acquisition, distribution, and regulation ofiron has been an active and important area of research for manyyears, as indicated by numerous excellent reviews (1, 5–8). Theredox properties of the Fe3þ∕Fe2þ pair are indispensable to thephysiological functions of both electron and oxygen transport. Inthe absence of hTF, Fe3þ is highly insoluble and readily hydro-lyzed; reduction to Fe2þ can produce reactive oxygen species viathe Fenton reaction.

Crucial to efficient Fe3þ delivery to cells is a TF specific homo-dimeric receptor, TFR, which binds two hTF molecules (9, 10). Atype II transmembrane glycoprotein, full-length TFR is com-prised of a short N-terminal intracellular region (residues 1–67)containing an endocytosis motif (YXRF), a transmembrane re-gion (residues 68–88), and a stalk (residues 89–120) that connectsto the large hTF binding ectodomain (121–760) (11). Althoughthe stalk contains two disulfide bonds covalently linking the mono-mers, the TFR homodimer forms even in the absence of the stalk

region. The 3.2-Å crystal structure of the TFR ectodomain revealedthree distinct domains per monomer: the protease-like domain(domain I, 121–188 and 384–606), the apical domain (domain II,189–383), and the helical domain (domain III, 607–760) (12).

In normal plasma, hTF is only approximately 30% iron satu-rated with a distribution of approximately 27% diferric hTF(Fe2hTF), 23% monoferric N (FeNhTF), 11% monoferric C(FeChTF), and 40% apohTF (13). Fe2hTF preferentially bindswith nM affinity to the TFR at neutral pH, whereas hTF withoutiron (apohTF) binds poorly at this pH (6). The two monoferricspecies bind with a similar intermediate affinity demonstratingthat each iron-containing lobe in the context of full-length hTFcontributes equally to the binding isotherm (14, 15). Thus it ispossible to prepare stable complexes of either monoferric hTFspecies bound to the TFR that are physiologically relevant.

Iron within each lobe of hTF is transported into cells by recep-tor-mediated endocytosis in which lower pH approximately 5.6,the participation of the TFR, and an unidentified chelator withinthe endosome orchestrate the efficient and balanced release ofiron from each lobe of hTF (16–19). Prior to exiting the endo-some via the divalent metal transporter 1, Fe3þ must be reducedto Fe2þ. The ferrireductase Steap3 may be involved in this pro-cess (20). Crucial to the recycling of hTF, after iron is released,apohTF remains bound to the TFR at the lower pH within theendosome and returns to the cell surface where it dissociates[or is displaced by Fe2hTF (21)].

Although the structure of apohTF has been determined (4),and despite tremendous effort, a crystal structure of Fe2hTFhas been surprisingly elusive. Likewise, the hTF/TFR complex hasevaded all efforts at crystallization, though the structures of TFRalone (12), TFR complexed with the HFE protein (22), and TFRcomplexed with a portion (GP1) of the Machupo virus (23) havebeen reported. In 2004, a 7.5-Å cryo-EM model of the hTF/TFRcomplex provided important insights into the molecular associa-tion of hTF and TFR (24). The validity of this model has beentested by mutagenesis studies; the contributions of specific resi-dues in both hTF and the TFR to the binding isotherm have beenmeasured by surface plasmon resonance (SPR) or isothermaltitration calorimetry, as well as by cell binding studies (25).

Here, we report the crystal structure of an hTF∕TFR complexat a resolution of 3.22 Å. The improved resolution of the presentstructure reveals a number of unique features of the hTF∕TFRinteraction that have a direct impact on function and highlight the

Author contributions: B.E.E. and A.B.M. designed research; B.E.E., A.N.S., and S.J.E.performed research; B.E.E., N.D.C., and A.B.M. analyzed data; and B.E.E., A.N.S., N.D.C.,S.J.E., and A.B.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.pdb.org (PDB ID codes 3S9L, 3S9M, and 3S9N).1To whom correspondence should be addressed. E-mail: [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1105786108/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1105786108 PNAS ∣ August 9, 2011 ∣ vol. 108 ∣ no. 32 ∣ 13089–13094

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Mar

ch 1

9, 2

020

unique receptor-mediated mechanisms of iron release from eachlobe. Our structure provides previously undescribed targets forfuture studies to advance understanding of how the interactionbetween hTF and the TFR promotes iron release in a physiolo-gically relevant time frame.

ResultsOverview of Structure.The asymmetric unit contains two TFR sub-units (chains A and B) and two FeNhTF molecules (chains C andD), which represent two half-biological units. Each unit forms anindependent biological assembly across a crystallographic twofoldaxis, such that chains A and C associate with a symmetry-relatedAsym and Csym (Fig. 1 and Fig. S1 symmetry molecules designatedA′ or C′). Likewise, chain B and chain D form a separate biolo-gical assembly with a symmetry mate DB-BsymDsym (B′ or D′).The final model of FeNhTF in the complex is comprised of theN1, N2, and C1 subdomains. Unfortunately, insufficient electrondensity precluded placement of the C2 subdomain in the model.Each TFR monomer contained residues 121–758 with threeN-linked glycans at Asn251, 317, and 727, each fit with a singleN-acetylglucosamine moiety.

Extensive contacts between the two TFR monomers form thenoncovalent dimer burying significant surface area (approxi-mately 3;200 Å2), with the helical domains from each monomercontributing substantially to this interaction. The surface areaburied at the hTF/TFR interface (approximately 1;330 Å2) is lessthan half of the TFR dimer interface; the C lobe contributesapproximately 60% and the N lobe contributes approximately40% to this interface. Binding of FeNhTF to the TFR involvesthree primary interaction motifs. The regions of the N1 and N2subdomains that contact the TFR (helical and protease-like do-mains, respectively) are located on either side of the hinge regionof the N lobe. The C lobe of hTF only contacts the TFR (helicaldomain) through the C1 subdomain. Significantly, there is noexperimental evidence to support the involvement of the C2 sub-domain in binding to the TFR.

hTF N1-TFR Motif. The N1/TFR interface accounts for approxi-mately 57% of the total contact surface area between the N lobeand TFR. The contacts of this motif are more extensive than sug-gested by the cryo-EM model. Nonadjacent residues Arg50 inhelix α-2, Tyr68 and Tyr71 both in helix α-3, as well as Ala73

and Asn75 in a loop (residues 72–76) within the N1 subdomainof hTF are in contact with three residues in the helical domain ofthe TFR [Gly661 and Asn662 in αIII-3, and Glu664 in the αIII-3/αIII-4 loop (residues 663–667)] (Table S1 and Fig. 2A). Arg50 inhTF (not identified in the cryo-EM model) likely forms a saltbridge with Glu664. Tyr68 and Tyr71 from hTF hydrogen bondwith Gly661 and Asn662 in the TFR. The backbone oxygen ofAla73 engages in a hydrogen bond with the backbone nitrogenof Glu664, whereas the nitrogen of Ala73 interacts with the back-bone oxygen of Asn662 of the TFR. There is no clear pattern ofconservation of the residues in this motif (4) that accounts for thespecificity of the interaction between the N1 subdomain of hTFand the TFR.We suggest that the requirement for specific residuesis somewhat obviated by the presence of backbone interactions.This is consistent with the observation that mutation of eitherTFR residue Asn662 or Glu664 did not significantly affect bindingof Fe2hTF or apohTF to the TFR as measured by SPR (25).

hTF N2-TFR Motif. Two loops in the N2 subdomain of hTF interactwith the N-terminal region of the TFR ectodomain and accountfor approximately 43% of the N lobe/TFR interface (Table S1 andFig. 2B). Van der Waals and hydrophobic interactions occur be-tween two proline residues (142 and 145) in the first loop of hTF(139–145) and Asp125 and Tyr123 in the TFR, respectively. Tworesidues in the second hTF loop (154–167), Asp166 and Phe167,also contact Tyr123 in the TFR. The side chain of Asp166 mayhydrogen bond with both the backbone nitrogen of TFR Arg121and the side chain OH of Tyr123 in the TFR. Asp166 and Phe167were not predicted to interact with the TFR in the cryo-EMmod-el. The importance of three of the four residues in the first loop(142–145, referred to as the PRKP loop) in the binding of the N2subdomain has been unequivocally established (15) and is further

Fig. 1. Structure of the FeNhTF∕TFR complex. The biological TFR homodimer(TFR-TFR′, A-A′) with two FeNhTF (FeNhTF and FeNhTF0, C-C0) molecules boundis shown oriented with the cell surface at the bottom. The TFR homodimer iscolored according to the domains: The apical domain is blue, the protease-like domain is green, and the twomonomers of the helical domain are brownand tan. The Ca2þ bound within the apical domain of each TFR monomer isshown in yellow. The FeNhTF molecules are colored according to subdomain:N1 is gray, N2 is black, and C1 is purple. The bridge between the two lobes iscyan. The Fe3þ bound within each N lobe of hTF is shown in red. All figureswere prepared using Pymol (43).

Fig. 2. hTF N-lobe-TFR interaction motifs (also see Table S1). (A) FeNhTF re-sidues involved in the N1 interaction motif (gray). Arg50, Tyr68, Tyr71, Ala73,and Asn75 within the N1 subdomain of hTF are close to three residues in thehelical domain of the TFR (Gly661, Asn662, and Glu664). Residues Leu72 andPro74, although involved in the N1 interaction motif, have been omittedfor clarity. (B) FeNhTF residues involved in the N2 interaction motif (black).The space filling representation of the N2 motif emphasizes that the predo-minant mode of interaction is van der Waals compared with the H-bondingnetwork for the N1 motif.

13090 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1105786108 Eckenroth et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

9, 2

020

supported by sequence alignments. The conservation of this re-gion of hTF correlates with the ability of a given TF to specificallybind to the human TFR (4). Importantly, the N2 motif appears tolack ionic bonds and has the fewest contacts making it relativelyweak in comparison to the N1 and C1 motifs.

hTF C1-TFR Motif. Because of conformational changes in the TFRas a result of hTF binding, this motif differs in a number ofrespects from the cryo-EMmodel. Specifically, the predicted net-work of four salt bridges does not exist. However, as predicted,there are multiple contacts including extensive interactions be-tween helix α-1, strand β-2, and a loop in the C1 subdomain ofhTFand the helical domain of the TFR, including αIII-2 and αIII-3, as well as the C-terminal region (Table S1 and Figs. 3A and 4A).In contrast to the antiparallel helical interactions between αIII-1and αIII-3 of the TFR with HFE, helix α-1 and strand β-2 fromthe C1 subdomain of hTF lie perpendicular to these TFR helices(Fig. 3B). Van der Waals, side chain to backbone, and side chainto side chain interactions all make significant contributions to thebinding of the C1 motif. Among the side chain interactions is asingle salt bridge formed between Arg651 of the TFR, previouslyshown to be critical for binding of either Fe2hTF or apohTF (25),and the highly conserved Asp356 of hTF (Fig. 3A). The backboneoxygen of hTF Cys368, residing in strand β-2, also lies within hy-drogen bonding distance of the ϵ nitrogen of Arg651. Arg646 inthe TFR is within 5.5 Å of the backbone carbonyl of hTF residues356, 359, and 366 as well as 3.4 Å from the γ-oxygen of Ser359.We note that Arg646 in the TFR resides in the canonical RGD(Arg646, Gly647, and Asp648) sequence previously shown to becritical to hTF binding (26). No obvious interactions with hTFand TFR residues Gly647 or Asp648 are observed in our struc-ture. Similar to the N1 motif, the side chain to backbone and vander Waals interactions may account for the limited conservationof C1 motif residues among TFs (4).

Kinetics of Iron Release. His349 in hTF (Fig. 4), identified as apH-inducible switch responsible for iron release from the C lobein the presence of the TFR (27, 28), forms the N-terminal cap ofhTF helix α-1. At pH 7.5, His349 interacts through both hydrogenbonds and van der Waals interactions with at least two residues inthe C-terminal portion of the TFR including Asp757 and Asn758.Although the weak density for the terminal Phe760 precludes itsinclusion in the model, it could also potentially interact withHis349. Intriguingly, His349 in our structure of the FeNhTF∕TFRcomplex is situated at the convergence of structural elementsfrom the two TFR monomers discussed below (Fig. 4A) and isshifted 5 Å (approximately one helical turn) from its positionin the cryo-EM model of Fe2hTF∕TFR (Fig. 4B).

The effect of substituting His349 in the C lobe has been furtherevaluated by determining rate constants for iron release from theH349A mutant in the Fe2hTF∕TFR complex by monitoringchanges in the intrinsic tryptophan fluorescence of hTF. Previouskinetic studies were carried out in a FeChTF background (28). Sig-nificantly, iron release from this H349A Fe2hTF∕TFR complex ispreceded by a conformational change with k ¼ 23.7� 4.6 min−1.The iron release rate constants are k1N ¼ 6.7� 0.3 min−1 for theN lobe and k2C ¼ 0.61� 0.02 min−1 for the C lobe (Fig. S2).

Conformational Changes in the TFR as a Result of hTF Binding. Thebinding of hTF results in the translation of the apical and pro-tease-like domains of the TFR and the reorientation of the mono-mers within the homodimer. These changes are revealed bysuperimposing a TFR monomer from the HFE/TFR (PDB IDcode 1DE4) and TFR alone (PDB ID code 1CX8) structures onour FeNhTF∕TFR structure. The calculated changes in the meanrms show the effect of reorientation at the dimer interface in aligand-dependent manner (Fig. 5, Inset). Site-specific changes perresidue within a TFR subunit are highlighted by a positional com-

parison (Fig. 5). The TFR in the FeNhTF complex is very similarto the TFR in the HFE complex (black line), but varies consider-ably in comparison to the unliganded TFR (red line).

The most dramatic change in the TFR structure as a result ofhTF binding is observed in the loop containing one of three gly-cosylation sites, Asn317 (Fig. 5, designated as TFR-TFR′ + C1motif). Specifically, the helical domain and C terminus from oneTFRmonomer interacts with the loop containing the glycosylatedAsn317 from the other TFR monomer, as well as with His318(Table S2). Phe316 is shifted by 8 Å and His318 flips, bringingit to within 5 Å of the C terminus of the other TFR monomer(in comparison to a distance of 17.5 Å in unliganded TFR) (Fig. 4).Although the interaction between nearby Gln320 and Ser638 isunchanged, a number of rearrangements occur at the TFR dimerinterface. Specifically, binding of hTF causes two Trp residues fromthe helical domain of one TFR monomer (Trp641 and 740) to un-dergo significant changes in packing with the other TFRmonomer.Thus, Trp641 shifts from interacting with the side chain of Phe316to interact with the backbones of Asn317 and His318. Trp740changes from backbone packing and a carbonyl hydrogen bondwith Pro314 to hydrogen bonding with the backbone of Gly469and π-stacking with Phe316, and His318 and Tyr470.

As in the HFE/TFR structure, binding of hTF causes a rotationalong the TFR dimer interface bringing four histidines (His475 inthe protease-like domain and His684 in the helical domain fromeach TFR monomer) into proximity (Fig. S3). The formation ofthis histidine cluster is another significant change in the TFRstructure as a result of hTF binding (Fig. 5, designated TFR-TFR′ interface). Other changes are observed in the regions ofthe TFR involved in binding the various hTFmotifs. Additionally,we suggest that conformational changes near residue Trp528 in

Fig. 3. hTF C1-TFR interaction motif (see Table S1). (A) Residues in the C1subdomain that contact the TFR are in purple. The space filling representa-tion of the C1 motif emphasizes the predominant van der Waals and packinginteractions. The carbonyl oxygen of Gly617 in hTF could hydrogen bondwiththe ϵ-nitrogen of Arg629 of the TFR (omitted for clarity). Note that criticalhTF residue His349 is also not depicted in this representation for enhancedclarity, but is clearly shown in Fig. 4. (B) Comparison of hTF and HFE (22) in-teractions with TFR αIII-1 and αIII-3 (brown). Secondary structural elements ofthe C1 subdomain (helix α-1 and strand β-2) that interact with TFR αIII-1 andαIII-3 are shown in purple (as in A). HFE secondary structural elements (helicesα-1 and α-2) that interact with TFR αIII-1 and αIII-3 are shown in cyan.

Eckenroth et al. PNAS ∣ August 9, 2011 ∣ vol. 108 ∣ no. 32 ∣ 13091

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Mar

ch 1

9, 2

020

the TFR are attributable to its proximity to the bridge betweenthe N and C lobes of hTF.

Metal Binding Site in the TFR. The interface between the apicaldomain and the protease-like domain features a metal ion withoctahedral coordination involving residues Thr310, Phe313,Glu465, and Glu468. Although significant anomalous scatteringis atypical of occupancy by Ca2þ or Mg2þ (Fig. S4), analysis ofour recombinant TFR by inductively coupled plasma-mass spec-trometry (ICP-MS) revealed the presence of a single Ca2þ permonomer.

The N1 and C1 Motifs Remain Attached Throughout the Endocytic Cy-cle. The N1 and C1 subdomains are directly connected by a sevenresidue bridge (332–338) that joins the lobes of hTF. Least-squares superposition of the Fe2 rabbit TF structure onto theFeNhTF from our structure indicates that the position of theN1 and C1 subdomains of hTF, relative to each other, doesnot appear to change as a function of iron status of the C lobe(Fig. S5). Likewise, superposition of the apohTF structure ontothe FeNhTF from our structure indicates that the position of theN1 and C1 subdomains of hTF, relative to each other, does notappear to change as a function of iron status of the N lobe. Inaddition to multiple interactions of the N1 and C1 subdomainswith αIII-3 in the TFR, a salt bridge between Arg308 (N1) andAsp376 (C1) observed in all three structures (FeNhTF from ourstructure, Fe2 rabbit TF and apohTF) may help maintain theirorientation relative to each other.

The Bridge Between the N and C Lobes. Of probable functional rele-vance, the position of the seven amino acid bridge between the N1and C1 subdomains in our structure and the bridge in the Fe2 rabbit

TF structure is very similar; in contrast, the bridge in the apohTFstructure is in a very different orientation, implying that the loss ofiron from the N lobe results in movement of the bridge residues.

DiscussionThe crystal structure of the FeNhTF∕TFR complex in combina-tion with previous structures suggests a mechanistic basis for thekinetic differences in iron release from each lobe of hTF in thepresence of the TFR. The structure of the complex clearly showsthat the TFR is altered by hTF binding. The relative positions ofthe N1 and C1 subdomains of hTF appear to remain constantthroughout the cycle, indicating that the N2 and C2 subdomainsmust move to accommodate the approximately 50° cleft openingand the release of iron from each lobe. Moreover, some insightinto how apohTF is stabilized by and remains bound to the TFRthroughout the endocytic cycle is provided.

Conformational Changes in TFR Induced by hTF Binding. A crucialfinding is that FeNhTF binding at pH 7.4 repositions the TFRdomains within each monomer, priming the TFR homodimerto undergo movements when the hTF/TFR complex encountersendosomal pH. Specifically, and as observed for the binding ofHFE (22), binding of hTF to each TFR monomer causes a rota-tion at the dimer interface that brings four histidines (His475 ineach protease-like domain and His684 in each helical domain)into proximity (Fig. S3). We suggest that binding of two hTF mo-lecules is probably required to fully prime the TFR dimer. As pre-viously suggested (22), endosomal pH triggers a chain reaction atthe TFR dimer interface. Specifically, protonation of the histi-dine cluster would result in movement of the TFR, perturbing(but not severing) the interaction with bound hTF. This histidinecluster would be sensitive to pH changes that occur within theendosome; however, its location deep within the TFR dimer in-terface might restrict direct solvent access. We suggest that near-by solvent accessible Arg680 could serve as a proton shuttle fromthe endosomal milieu to the histidine cluster to circumvent thisrestriction. Because the N1 and C1 subdomains both remainbound to the same TFR helix (αIII-3), even a small pH-inducedmovement of the TFR could impact the stability of the iron bind-ing cleft within each lobe. As detailed below, other elements ineach lobe are simultaneously undergoing pH-induced changesthat promote iron release. In contrast, protonation of both theTFR and HFE histidine residues completely disrupts the HFE/TFR interaction below pH 6.0 (22).

Fig. 4. Intersection formed between apical domain (blue) and protease-likedomain (green) of one TFR monomer (TFR′), the helical domain (brown-tan)of the other TFR monomer (TFR) (Table S2), and the C1 subdomain (purple) ofhTF. (A) Our crystal structure of TFR in complex with FeNhTF. The maps shownare for the anomalous difference Fourier for the data collected at 0.98 Åcontoured at 3 sigma (red) and a simulated annealing composite omitmap at 1 sigma (blue). (B) Overlay of A (darker shades) and the cryo-EM com-plex (1SUV) (24) after least-squares superposition using the TFR molecule(chain A). Secondary structural elements are labeled for clarity. Note thatorientation has changed relative to Fig. 1, such that the cell surface is atthe top. The Ca2þ bound within the apical domain of each TFR monomeris shown in yellow.

Fig. 5. Plot of the root mean squared deviation calculated using CNS forchain A from the complex compared with a single chain from the receptoralone [red—1CX8 (12)] and when in complex with HFE [black—1DE4 (22)].(Inset) The table shows the mean rms for both chains of the TFR dimer aftersuperposition of the single chain. P, A, and H refer to the protease-like, apical,and helical domains of the TFR, respectively.

13092 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1105786108 Eckenroth et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

9, 2

020

Significantly, binding of hTF causes surface exposed His318 inthe apical domain of each TFR subunit to flip into the intersec-tion formed by the two TFR monomers and the C1 subdomain ofhTF; His318 moves nearly 18 Å relative to the unliganded TFRstructure (see below). Collectively, this region of the TFR influ-ences both protein stability and iron release from hTF. The ad-verse affect of the N317D TFR mutant on expression level andiron release support its importance (29).

Kinetics of Iron Release from Each Lobe of hTF.Accurately describingthe mechanism of iron release from each lobe has been challen-ging due to the number of variables (the TFR, endosomal pH, achelator, salt, and cooperativity between lobes). We have recentlyreported a complete set of kinetic rate constants for conforma-tional changes and iron release from hTFat pH 5.6 (�TFR) usingan array of recombinant hTF constructs (19). Kinetic studies fromFe2hTF at pH 5.6 (with EDTA as chelator) in the absence of theTFR indicate that 96% of the time iron is released quickly from theN lobe (17.7 min−1), followed by slow release from the C lobe(0.65 min−1) (Fig. S2A) (19). Additionally, in the absence of theTFR there is clearly cooperativity between the two lobes of hTF.Specifically, iron release from the N lobe is sensitive to the C lobe(although the reverse is not true) (19, 30). Fitting of the kineticdata from the Fe2hTF∕TFR complex has allowed us to estimatethat approximately 65% of the time iron is released first from the Clobe (k1C ¼ 5.5 min−1) and 35% of the time from the N lobe first(k1N ¼ 2.8 min−1) (19) (Fig. S2B). Together with previous TFstructures, the present FeNhTF∕TFR structure provides insightsthat help to explain the kinetic behavior of each lobe of hTF.

Iron Release from the C Lobe. Iron release from the C lobe of hTF inthe absence of the TFR is extremely slow and unaffected by the Nlobe (18). The C lobe features a triad of residues (Lys534-Arg632-Asp634) that appears to control the rate constant of iron releasein the absence of the TFR (31, 32). Iron release from the C lobein the presence of the TFR proceeds by a different mechanismand is 7- to 11-fold faster than in the absence of the TFR (19).Recent studies have demonstrated that iron release from the Clobe is dictated by His349, but only when hTF is bound to theTFR (28). Based on the cryo-EM structure, it was predicted thata hydrophobic patch (TFR residues Trp641 and Phe760) interactswith His349 and stimulates iron release by stabilizing the apohTF/TFR complex (27). Because of the 5-Å shift of helix α-1 of the Clobe in our structure, His349 actually lies in the intersectionformed between the two TFR monomers and the C1 subdomainof hTF and is positioned to interact with several C-terminal re-sidues (Asp757-Phe760) of the TFR (but not with Trp641).

The critical role of His349 as the driving force of TFR stimu-lated iron release from the C lobe is clearly demonstrated by thenewly determined kinetics of iron removal from the H349A mu-tant in the Fe2hTF∕TFR complex whereby the rate constant foriron removal from the C lobe is reduced 12-fold (k2C ¼ 7.2 versus0.61 min−1). In contrast to the Fe2hTF∕TFR control, which re-quires that both pathways be included in the fit, the data for theH349A mutant fit only to the single upper pathway (N lobe fol-lowed by the C lobe) (Fig. S2). Interestingly, the rate constant foriron release (k2C ¼ 0.61� 0.02 min−1) from the C lobe of theH349A Fe2hTF∕TFR complex is essentially identical to the rateconstant of k2C ¼ 0.65� 0.06 min−1 for iron release from the Clobe of Fe2hTF in the absence of the TFR (Fig. S2A). We suggestthat, in wild-type hTF, protonation of His349 at pH 5.6 converts aweak hydrophobic interaction with Phe760 at the C terminus ofthe TFR into either a stronger cation-π interaction with Phe760or a salt bridge with Asp757, causing a conformational change inthe C lobe and accelerating iron release from this lobe.

Iron Release from the N Lobe. In the absence of the TFR, iron re-lease from the N lobe relies on protonation of a pair of lysines

(Lys206 and Lys296 referred to as the dilysine trigger) on oppos-ing sides of the binding cleft that form a hydrogen bond at neutralpH and literally trigger cleft opening at endosomal pH (33). Therelease of iron from the N lobe is further accelerated by bindingof anions to Arg143, a recently identified kinetically significantanion binding (KISAB) site in the PRKP loop in the N2 subdo-main (34). Attachment of both the N1 and N2 subdomains to theTFR limits access to this KISAB site, hinders cleft opening, andresults in a rate of iron release that is 6- to 15-fold slower than inthe absence of the TFR (19). Given that the N1 and C1 subdo-mains maintain their positions relative to each other in both theFe2 rabbit TF and apohTF structures, the N2 subdomain must dis-engage from the TFR to allow the cleft to open. We suggest that apH-induced movement of the TFRmay help destabilize binding ofthe N2 subdomain, which is relatively weak. Rearrangement of thePRKP loop in the N lobe then pulls the N2 subdomain away fromthe TFR allowing the cleft to open and release of iron.

Stabilization of the apohTF/TFR Complex. The return of apohTF tothe cell surface is a distinctive feature of the endocytic cycle.Therefore, release of iron from each lobe of the hTF/TFR com-plex must be effectively balanced with stabilization of the apocomplex. The significant conformational changes associated withcleft opening and iron release from each lobe must be accommo-dated. The FeNhTF∕TFR structure provides a molecular basis forthe stabilization motifs within each lobe. Although some details ofthe interactions of the N1 and C1 subdomains with the TFR maychange during the endocytic cycle, many of the interactions withinthese two binding motifs are probably preserved (Table S1); stu-dies showing that mutation of TFR residues Asn629, Gly647,Phe650, or Arg651 affected binding of both Fe2hTF and apohTFare consistent with this idea (25). In the C lobe, additional stabi-lization of the apo conformation is imparted by protonation of thetwo histidine residues (His349 and His318) at endosomal pH,which strengthens this region through potential cation-π interac-tions with Phe760 and Trp641 of the TFR, respectively. The un-anticipated movement of His318 provides additional detail as tohow the C1 subdomain remains bound to the TFR throughout theendocytic cycle. As revealed by the apohTF structure (4), the apoconformation of the N lobe is secured by a salt bridge betweenAsp240 in the N2 subdomain and Arg678 in the C1 subdomain.Additionally, the PRKP loop is connected to the bridge betweenthe N1 and C1 subdomains by a disulfide bond between Cys137 inthe N2 subdomain and Cys331 in the N1 subdomain. Significantly,in the apohTF structure, movement of the PRKP loop and the dis-ulfide bond (Cys137-Cys331) repositions the bridge bringing it clo-ser to the protease-like domain of the TFR to possibly furtherstabilize the apo conformation in a pH-dependent manner.

In conclusion, the crystal structure of the FeNhTF∕TFR com-plex at neutral pH reveals a number of unique aspects of thisdynamic system, allowing a more accurate description of the in-teractions that control iron release from each lobe in the presenceof the TFR. Obviously, a single, static crystal structure cannotprovide absolute temporal resolution, as multiple events are oc-curring nearly simultaneously as the pH surrounding theFe2hTF∕TFR complex changes during endocytosis. Nevertheless,the functional significance of the induced conformationalchanges in the TFR structure is evidenced by its direct participa-tion in promoting iron release from the C lobe and hindering itfrom the N lobe. The structure advances our understanding of theimportant interactions, the role of the TFR, and provides pre-viously undescribed information to drive future work.

Materials and MethodsProduction and Purification of FeNhTF and the Soluble Portion of the TFR.Recombinant nonglycosylated monoferric hTF, designated FeNhTF, containsmutations preventing iron acquisition by the C lobe (Y426F/Y517F) andglycosylation (N413D and N611D) and is produced in a BHK cell expressionsystem (35). Likewise, the glycosylated ectodomain of the TFR (residues

Eckenroth et al. PNAS ∣ August 9, 2011 ∣ vol. 108 ∣ no. 32 ∣ 13093

BIOCH

EMISTR

Y

Dow

nloa

ded

by g

uest

on

Mar

ch 1

9, 2

020

121–760) is produced and secreted into the tissue culture medium of trans-fected BHK cells (29). Both constructs contain a 6X-His tag at the N terminus,followed by a factor Xa cleavage site (IEGR), and are purified from the med-ium using nickel affinity chromatography followed by gel filtration. Thehomogeneity of each preparation is evaluated by SDS-PAGE. The histidinetags have not been removed from either construct.

The hTF/TFR complex is formed in the presence of excess FeNhTF and iso-lated by passage over a Sephacryl S300HR gel filtration column (15). Followingconcentration to 20 mg∕mL in 100 mM NH4HCO3, crystals are grown at 20 °Cby the hanging drop vapor diffusion method. The protein solution is mixed ata 2 to 1 ratio with reservoir solution containing 100 mM Hepes pH 7.5, 4–6%PEG 3350, 200 mMMgCl2, and 5–20% 1,2-propanediol. Pale pink crystals withdimensions from 50 to approximately 500 μm developed in 1 to 10 d.

Data Processing, Structure Solution, and Refinement. Integration of diffractionimages and data scaling were performed using HKL2000 (36). Molecular re-placement solutions were found using Phaser (37) within CCP4 version 6.2(38) and utilized TFR monomer (PDB ID code 1CX8) (12), the hTF N-lobe(PDB ID code 1A8E) (39), and C1/C2 subdomains of apohTF (PDB ID code2HAU) (4) as search models (Table S3). A clear molecular replacement solutionfor two TFR monomers was found with P4322 as the space group, each form-ing the biological TFR homodimer with a symmetry mate. As shown inTable S4, three independent models of FeNhTF in the complex were gener-ated using datasets derived from multiple crystallization and cryoprotectionconditions and contain most of the N1, N2 and the C1 subdomains. Prelimin-ary solutions were improved by rigid body refinement with Refmac5.5 (40)using the N lobe and C1 subdomain solutions as well as the three domains ofthe ectodomain of the TFR. Structure refinement was performed with Crys-tallography and NMR System (CNS) (41) version 1.2 and model building usingCoot (42). A single Fe3þ and the synergistic carbonate were clearly observed

in each FeNhTF (Fig. S4). No density was observed for the 6X-His tag, thefactor Xa cleavage site, or the first two to three residues of the N terminus.The final model of each TFR in the complex contained residues 121–758 withthree N-linked glycans at Asn251, 317, and 727 each fitted with a singleN-acetylglucosamine moiety. Because the carbohydrate composition for BHKexpressed TFR is unknown, the remaining density could not be unambigu-ously built. Again no density for the 6X-His tag and the factor Xa cleavagesite at the N terminus of this construct was observed in the model of theTFR. The model derived from the 3.22-Å data was refined to an R factorof 27.2% (Rfree of 31.4%) with 99.5% in the preferred and allowed regionsof the Ramachandran plot and 0.1% outliers (Table S4).

ACKNOWLEDGMENTS. We thank the University of Vermont Center for X-RayCrystallography; Alexei Soares at the Brookhaven National Laboratory “Mail-In Data Collection Program,” data were measured at beamline X25 of theNational Synchrotron Light Source with financial support from the Officesof Biological and Environmental Research and of Basic Energy Sciences ofthe US Department of Energy, and from the National Center for ResearchResources (NCRR) of the National Institutes of Health (NIH) (P41RR012408);Chae Un Kim and Irina Kriksunov at Macromolecular Diffraction at CornellHigh Energy Synchrotron Source (CHESS) (MacCHESS) Beamline F2. Someof the work is based upon research conducted at the CHESS, which is sup-ported by the National Science Foundation (NSF) and the NIH/NIGMS underNSF Award DMR-0225180, using the MacCHESS facility, which is also sup-ported by NIH Award RR-01646 through National Center for Research Re-sources. This work was funded by US Public Health Service R01 DK21739(to A.B.M.) and R01 GM-20194 (to N.D.C.). Support for B.E.E. and A.N.S.came from Hemostasis and Thrombosis Training Grant 5T32HL007594 issuedto Dr. K. G. Mann at the University of Vermont by National Heart, Lung andBlood Institute. A.N.S. is currently funded by an American Heart AssociationPredoctoral Fellowship (10PRE4200010).

1. Andrews NC, Schmidt PJ (2007) Iron homeostasis. Annu Rev Physiol 69:69–85.2. Harris DC, Aisen P, Loehr TM (1989) Physical biochemistry of the transferrins. Iron

Carriers and Iron Proteins (VCH, New York), pp 239–351.3. Hall DR, et al. (2002) The crystal and molecular structures of diferric porcine and rabbit

serum transferrins at resolutions of 2.15 and 2.60 A, respectively. Acta Crystallogr DBiol Crystallogr 58:70–80.

4. Wally J, et al. (2006) The crystal structure of iron-free human serum transferrin pro-vides insight into inter-lobe communication and receptor binding. J Biol Chem281:24934–24944.

5. Eisenstein RS (2000) Iron regulatory proteins and the molecular control of mammalianiron metabolism. Annu Rev Nutr 20:627–662.

6. Aisen P, Enns C, Wessling-Resnick M (2001) Chemistry and biology of eukaryotic ironmetabolism. Int J Biochem Cell Biol 33:940–959.

7. Hentze MW, Muckenthaler MU, Andrews NC (2004) Balancing acts: Molecular controlof mammalian iron metabolism. Cell 117:285–297.

8. Hentze MW, Muckenthaler MU, Galy B, Camaschella C (2010) Two to tango: Regula-tion of mammalian iron metabolism. Cell 142:24–38.

9. Trowbridge IS, Omary MB (1981) Human cell surface glycoprotein related to cellproliferation is the receptor for transferrin. Proc Natl Acad Sci USA 78:3039–3043.

10. Aisen P, Sarkar B (1983) Interactions of transferrin with cells. Biological Aspects ofMetals and Metal-Related Diseases (Raven, New York), pp 67–80.

11. McClelland A, Kuhn LC, Ruddle FH (1984) The human transferrin receptor gene:genomic organization, and the complete primary structure of the receptor deducedfrom a cDNA sequence. Cell 39:267–274.

12. Lawrence CM, et al. (1999) Crystal structure of the ectodomain of human transferrinreceptor. Science 286:779–782.

13. Williams J, Moreton K (1980) The distribution of iron between the metal-binding sitesof transferrin in human serum. Biochem J 185:483–488.

14. Mason AB, et al. (2005) Mutational analysis of C-lobe ligands of human serumtransferrin: Insights into the mechanism of iron release. Biochemistry 44:8013–8021.

15. Mason AB, et al. (2009) A loop in the N-lobe of human serum transferrin is critical forbinding to the transferrin receptor as revealed by mutagenesis, isothermal titrationcalorimetry, and epitope mapping. J Mol Recognit 22:521–529.

16. Klausner RD, Ashwell G, van Renswoude J, Harford JB, Bridges KR (1983) Binding ofapotransferrin to K562 cells: Explanation of the transferrin cycle. Proc Natl Acad SciUSA 80:2263–2266.

17. Bali PK, Zak O, Aisen P (1991) A new role for the transferrin receptor in the release ofiron from transferrin. Biochemistry 30:324–328.

18. Bali PK, Aisen P (1992) Receptor-induced switch in site-site cooperativity during ironrelease by transferrin. Biochemistry 31:3963–3967.

19. Byrne SL, Chasteen ND, Steere AN,Mason AB (2010) The unique kinetics of iron releasefrom transferrin: The role of receptor, lobe-lobe interactions, and salt at endosomalpH. J Mol Biol 396:130–140.

20. Ohgami RS, et al. (2005) Identification of a ferrireductase required for efficienttransferrin-dependent iron uptake in erythroid cells. Nat Genet 37:1264–1269.

21. Leverence R, Mason AB, Kaltashov IA (2010) Noncanonical interactions between serumtransferrin and transferrin receptor evaluated with electrospray ionization massspectrometry. Proc Natl Acad Sci USA 107:8123–8128.

22. Bennett MJ, Lebron JA, Bjorkman PJ (2000) Crystal structure of the hereditary haemo-chromatosis protein HFE complexed with transferrin receptor. Nature 403:46–53.

23. Abraham J, Corbett KD, Farzan M, Choe H, Harrison SC (2010) Structural basis forreceptor recognition by New World hemorrhagic fever arenaviruses. Nat StructMol Biol 17:438–444.

24. Cheng Y, Zak O, Aisen P, Harrison SC, Walz T (2004) Structure of the human transferrinreceptor-transferrin complex. Cell 116:565–576.

25. Giannetti AM, Snow PM, Zak O, Bjorkman PJ (2003) Mechanism for multiple ligandrecognition by the human transferrin receptor. PLoS Biol 1:341–350.

26. Dubljevic V, Sali A, Goding JW (1999) A Conserved RGD (Arg-Gly-Asp) motif in thetransferrin receptor is required for binding to transferrin. Biochem J 341:11–14.

27. Giannetti AM, et al. (2005) The molecular mechanism for receptor-stimulated ironrelease from the plasma iron transport protein transferrin. Structure 13:1613–1623.

28. Steere AN, et al. (2010) Evidence that His349 acts as a pH-inducible switch to acceleratereceptor-mediated iron release from the C-lobe of human transferrin. J Biol InorgChem 15:1341–1352.

29. Byrne SL, et al. (2006) Effect of glycosylation on the function of a soluble, recombinantform of the transferrin receptor. Biochemistry 45:6663–6673.

30. Byrne SL, Mason AB (2009) Human serum transferrin: A tale of two lobes. Urea gel andsteady state fluorescence analysis of recombinant transferrins as a function of pH, time,and the soluble portion of the transferrin receptor. J Biol Inorg Chem 14:771–781.

31. Halbrooks PJ, et al. (2005) Composition of pH sensitive triad in C-lobe of human serumtransferrin. Comparison to sequences of ovotransferrin and lactoferrin providesinsight into functional differences in iron release. Biochemistry 44:15451–15460.

32. James NG, et al. (2009) Inequivalent contribution of the five tryptophan residues in theC-lobe of human serum transferrin to the fluorescence increase when iron is released.Biochemistry 48:2858–2867.

33. He QY, Mason AB, Tam BM, MacGillivray RTA, Woodworth RC (1999) Dual role ofLys206-Lys296 interaction in human transferrin N-lobe: Iron-release trigger andanion-binding site. Biochemistry 38:9704–9711.

34. Byrne SL, Steere AN, Chasteen ND, Mason AB (2010) Identification of a kineticallysignificant anion binding (KISAB) site in the N-lobe of human serum transferrin.Biochemistry 49:4200–4207.

35. Mason AB, et al. (2004) Expression, purification, and characterization of authenticmonoferric and apo-human serum transferrins. Protein Expr Purif 36:318–326.

36. Otwinowski Z, Minor W, Carter CW, Sweet RM (1997) Processing of x-ray diffractiondata collected in oscillation mode. Methods in Enzymology, Part A (Academic,San Diego), pp 307–326.

37. McCoy AJ, et al. (2007) Phaser crystallographic software. J Appl Crystallogr 40:658–674.38. Winn MD, et al. (2011) Overview of the CCP4 suite and current developments. Acta

Crystallogr D Biol Crystallogr 67:235–242.39. MacGillivray RTA, et al. (1998) Two high-resolution crystal structures of the recombi-

nant N-lobe of human transferrin reveal a structural change implicated in iron release.Biochemistry 37:7919–7928.

40. Winn MD, Isupov MN, Murshudov GN (2001) Use of TLS parameters to model aniso-tropic displacements inmacromolecular refinement.Acta Crystallogr D Biol Crystallogr57:122–133.

41. Brunger AT, et al. (1998) Crystallography & NMR system: A new software suite formacromolecular structure determination.Acta Crystallogr D Biol Crystallogr 54:905–921.

42. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. ActaCrystallogr D Biol Crystallogr 60:2126–2132.

43. Delano WL (2002) The PyMOL Molecular Graphics System..

13094 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1105786108 Eckenroth et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 1

9, 2

020