Insights into the Function of the Unstructured N-Terminal Domain of Proteins 4.1R and 4.1G in Erythropoiesis

Membrane skeletal protein 4.1R is the prototypical member of a family of four highly paralogous proteins that include 4.1G, 4.1N, and 4.1B. Two isoforms of 4.1R (4.1R135 and 4.1R80), as well as 4.1G, are expressed in erythroblasts during terminal differentiation, but only 4.1R80 is present in mature erythrocytes. One goal in the field is to better understand the complex regulation of cell type and isoform-specific expression of 4.1 proteins. To start answering these questions, we are studying in depth the important functions of 4.1 proteins in the organization and function of the membrane skeleton in erythrocytes. We have previously reported that the binding profiles of 4.1R80 and 4.1R135 to membrane proteins and calmodulin are very different despite the similar structure of the membrane-binding domain of 4.1G and 4.1R135. We have accumulated evidence for those differences being caused by the N-terminal 209 amino acids headpiece region (HP). Interestingly, the HP region is an unstructured domain. Here we present an overview of the differences and similarities between 4.1 isoforms and paralogs. We also discuss the biological significance of unstructured domains.


4.1R in the Erythrocyte Membrane Skeleton
The membrane skeleton, which underlies the erythrocyte plasma membrane, is made of a spectrin/actin lattice anchored to various transmembrane proteins via two specialized cytoskeletal proteins, 4.1R and red blood cell ankyrin, ankyrin-R [1]. 4.1R 80 stabilizes horizontal interactions between spectrin heterodimers (α2-β2) and short actin (∼14 molecules) filaments. Actin filaments interact with numerous accessory proteins, such as tropomyosin, myosin, tropomodulin, and adducin [1], which ensure reorganization of actin filaments. 4.1R 80 interacts also with the transmembrane protein, glycophorin C (GPC) and with the membrane-associated guanylate kinase (MAGUK) protein p55, which also acts as an erythrocyte scaffolding protein ( Figure 1).

p55
. p55 is a 55 kDa erythrocyte scaffolding protein that belongs to the membrane-associated guanylate kinase homologues (MAGUK) family (ID: Q00013). This protein is characterized by the presence of a PDZ (Postsynaptic density protein-95, Dlg (Drosophila disc large tumor suppressor), ZO-1 (Zonula Occludens-1)) domain, an SH (srchomology) 3 domain, and a catalytic inactive guanylate kinase like (GUK) domain, all of which function as proteinprotein interaction modules ( Figure 2) [8]. The number of p55 copies in the human erythrocytes is ∼80,000. p55 is also called Membrane Palmitoylated Protein 1 (MPP1) since  Figure 2: Organization of the R30/GPC/p55 ternary complex. The NMR structure of GPC peptide and PDZ domain complex has been previously reported [7] (PDB accession no. 2ejy). The HOOK domain is the 4.1R binding site for p55 [4].
International Journal of Cell Biology The LRRRY sequence mediates interaction between monomers and is located in an α-helix. The band 3 binding sequence LEEDY, which mediates interaction with 4.1R, is located in the loop structure [15] (PDB accession no. 1hyn). There is no information about the stoichiometry of band 3 binding to 4.1R.
cysteine residues in the GUK domain can be palmitoylated [8]. However, there is still no direct evidence for the expression of palmitoylated p55 in living cells. Although the function of p55 in erythrocytes has not been clarified, p55 seems essential for maintenance of polarity in neutrophils [9] and in hair cells [10,11]. Recently, NMR-based studies have enabled to characterize the 3D structural profile of the GPC peptide that interacts with the PDZ domain of p55 [7]. Mutational studies, based on the replacement of the phenylalanine residue in the EYFI motif with a cysteine residue (E 125 YCI), have provided us with structural information on GPC binding to p55. Thus, 4.1R 80 participates in the formation of two different ternary complexes in erythrocytes, the 4.1R 80 /GPC/p55/ complex and the 4.1R 80 /spectrin/actin complex. Ektacytometry studies have revealed that 4.1R plays a key role in controlling erythrocyte membrane mechanical properties. Indeed, resealed membranes prepared from erythrocytes totally or partially deficient in 4.1R 80 show a dramatic decrease in membrane stability (reviewed in [12]). Interestingly, addition of either purified 4.1R 80 or purified 10 kDa spectrin-actin binding domain of 4.1R 80 to unstable 4.1R-deficient membranes is able to restore mechanical stability to such membranes. This demonstrates unequivocally an essential role for 4.1R 80 and more specifically for a 21amino-acid peptide encoded by exon 16 in the spectrinactin binding domain, in maintaining membrane stability by promoting spectrin/actin interactions [1,12].

Band 3.
Membrane stability is also controlled in part by band 3-ankyrin-spectrin interaction (as shown in Figure 1). Band 3 is a 102 kDa 14-transmembrane protein which mediates exchange of HCO 3 − and Cl − and is therefore referred to as anion exchanger 1 (AE1) [1] (ID: P02730). It is expressed at 1,200,000 molecules/cell. It forms dimers that assemble into tetramers, each tetramer binding to one molecule of ankyrin. This is the base for the organization of the band 3-ankyrin-spectrin complex [13]. 4.1R binds to the I 386 RRRY and L 343 RRRY sequences in band 3-cytoplasmic domain [14]. Although the crystal structure of the N-terminal cytoplasmic domain of band 3 has been reported, this structure is putative as the Nterminal 55 residues, including the L 343 RRRY sequence, were missing in the crystal [15]. The results indicate that band 3 has four 4.1R binding sites. The stoichiometry of band 3 binding to 4.1R is still unknown. The importance of band 3 in membrane architecture results from its role in anchoring the spectrin network through interaction with the scaffold protein ankyrin. We have demonstrated that 4.1R 80 modulates band 3 interaction with ankyrin [16]. We have characterized a similar function for 4.1R 80 in modulating ankyrin interaction with CD44, a single transmembrane protein which acts as receptor for hyaluronic acid [17].
The absence of 4.1R, ankyrin, or spectrin or selected mutations in these proteins result in alterations in erythrocyte shape and mechanical properties (reviewed in [1,12]). We have demonstrated that 4.1R interacts with membrane protein analogues in zebrafish (Danio rerio) using in vitro binding assays [18,19] 4.1R forms multimolecular complexes with transmembrane proteins and membraneassociated proteins, such as spectrin and actin [1]. Such complexes, which are critical for maintaining structural stability in red blood cells, could well be involved in other functions in nonerythroid cells, such as, for example, signal transduction at sites of cell-cell and/or cell-matrix contacts.
4.1R 80 (ID: P11171), present at approximately 200,000 copies per erythrocyte, can be extracted by high salt treatment of inside-out vesicles (IOVs), which correspond to erythrocytes membranes depleted of spectrin and actin. Based on its 622-amino-acid composition (reviewed in [1,12]), the predicted molecular weight of 4.1R is only ∼70 kDa, the discrepancy with the apparent molecular weight resulting in part from the unstructured domains of 4.1R. Limited αchymotryptic digestion of 4.1R generates four polypeptides: a 30 kDa N-terminal membrane-binding domain, a 16 kDa domain, a 10 kDa spectrin-actin binding domain, and a 22/24 kDa C-terminal domain (reviewed in [1,12]). A 4.1R isoform expressed in erythroblasts, but not in mature erythrocytes, contains an extra N-terminal 209 amino acids headpiece (HP) region. The apparent molecular weight of this 4.1R isoform in SDS-PAGE is ∼135 kDa, and it is therefore referred to as 4.1R 135 . However, its theoretical molecular weight is ∼100 kDa. This discrepancy results from the unstructured state of the HP region [23].

Unstructured N-Terminal and Structured 30 kDa FERM
Domains of 4.1R 135 . We calculated the disorder probability of the N-terminal HP region and the FERM domain using the PrDOS software (http://prdos.hgc.jp/cgi-bin/top.cgi) [26]. A value greater than 0.5 reflects a disordered structure, with a probability of false prediction of 5% or less. Our analysis indicates a highly disordered structure for the HP region (amino acids 1-209) that contrasts with a highly ordered structure for the 30 kDa FERM domain (amino acids 210-507). Of particular note, while the overall 209aa HP region adopts a disordered structure, a short polypeptide (amino acids 70-80), corresponding to a previously identified Ca 2+dependent CaM-binding site [27,28], does not ( Figure 4).
We experimentally demonstrated that the HP is an unfolded region by SDS-PAGE, size exclusion chromatography (SEC), and dynamic light scattering (DLS). The theoretical molecular weight of 4.1R HP (RHP) is 23 kDa but we estimate its apparent molecular weight as 55 kDa by SDS-PAGE [29]. Furthermore, SEC analysis reveals that RHP is eluted between IgG (150 kDa) and albumin (68 kDa) on a Sephacryl S-300 column. While the theoretical molecular weights of the proteins corresponding to amino acids 1-507 of 4.1R 135 (RHP-R30) and to R30 (30 kDa FERM domain) are 56 kDa and 32 kDa, respectively, they migrate as polypeptides of >100 kDa and 35 kDa, respectively, on SDS-PAGE [29]. By DLS measurements, the hydrodynamic diameters of RHP, RHP-R30 and R30 are 7.6, 9.4, 5.6 nm, respectively (Nunomura, W., Shiba, K. and Takakuwa, Y., unpublished data). These hydrodynamic parameters enabled us to estimate the molecular weight of RHP, RHP-R30 and R30 to be 77, 127, and 40 kDa, respectively. The discrepancies between theoretical and apparent molecular weights for proteins containing RHP reflect the unfolded nature of this peptide.
In contrast, the consistency between theoretical and apparent molecular weights for R30 illustrates the folded nature of R30. Importantly, PrDOS-based analysis of full length 4.1R 135 predicted the 30 kDa domain to be the only region in the whole protein to adopt an ordered (folded) structure. The crystal structure of 4.1R 30 kDa domain is reminiscent of the shape of a cloverleaf or of a propeller, with three clearly distinct lobes (PDB: 1GG3) [25]. First, the N-lobe, corresponding to the first 78 amino acids and which includes the band 3 binding motif L 37 EEDY, consists of 4 double-stranded β-strands. Second, the α-lobe, corresponding to the following 90 amino acids and which includes the GPC binding site, consists of 4 α-helices. Third, the COOH-terminal lobe (C-lobe), which contains the p55 binding surface, is made of seven β-strands, and ends with an α-helix ( Figure 4). Although many membrane skeletal proteins contain intrinsically disordered (unfolded) regions, there are very few reports describing the function(s) of these intrinsically disordered region [30][31][32][33][34][35]. Our findings will contribute not only to a better understanding of the structure of membrane skeletal proteins but also of the function of intrinsically disordered proteins.   [12,38]). At higher Ca 2+ concentrations, CaM-binding affinity for the Ca 2+ -dependent site, located in peptide 9 of the 30 kDa domain, increases. This results in a conformational and/or electric surface change which alters 4.1R binding sites, 4.1R interacting consequently with lower affinity with its binding partners p55, GPC, and spectrin/actin. This model implies that CaM regulates protein 4.1R binding to transmembrane proteins through Ca 2+ -independent and Ca 2+ -dependent binding sites. observed for 4.1R 80 . These findings imply that the presence or absence of HP in 4.1R isoforms modulates their binding affinity for GPC but not for band 3 [29].

Expression of 4.1R
In order to obtain independent confirmation of the binding affinities of 4.1R 135 to band 3cyt and GPCcyt, we used the IAsys system based on the resonant mirror detection method [29]. In agreement with the binding data using IOVs described above, there was a dramatic difference in the binding affinity of 4.1R 135 to band 3cyt and GPCcyt, the binding affinity being much higher for band 3cyt (23 ± 2 nM) than for GPCcyt (1327 ± 103 nM). In marked contrast, K (D) values for binding of 4.1R 80 to both band 3cyt and GPCcyt were very similar, in the submicromolar range. This confirmed an important role for HP in regulating 4.1R affinities for its two major transmembrane binding partners. In contrast to the marked differences in the binding affinities of 4.1R 135 and 4.1R 80 to band 3cyt and GPCcyt, the two isoforms bound to p55 with very similar affinities, in the submicromolar range.
As expected from the data obtained with 4.1R 80 and 4.1R 135 isoforms, the addition of RHP to R30 (RHP-R30) results in a profound change in the ability of R30 to bind to band 3cyt and GPCcyt. Thus, the binding affinity of RHP-R30 for band 3cyt is 35-fold higher than for GPCcyt. Together, these findings highlight an important role for RHP in modulating the interaction of R30 with its two membranebinding partners.

Differences in CaM Binding to 4.1R Isoforms.
We have previously documented that 4.1R 80 binds to CaM with a K (D) in the submicromolar range, both in the presence and absence of Ca 2+ implying that this interaction is Ca 2+independent [38]. We have also examined the nature of the interaction between 4.1R 135 and CaM. Kinetic analysis of 4.1R 135 interaction with CaM using the IAsys system identified a very strong interaction with a K (D) of 51 ± 5 nM in the presence of Ca 2+ . In the absence of Ca 2+ , the binding affinity decreased by over 100-fold. Thus, in contrast to 4.1R 80 , the interaction of 4.1R 135 with CaM is strongly Ca 2+ dependent. Probing of the HP region alone confirms a Ca 2+dependent interaction with CaM, implying that this region harbors the CaM-binding site [27,28]. Our observations are in accordance with Leclerc and Vetter's study that identifies the S 76 RGLSRLFSSFLKRPKS peptide as the Ca 2+ -dependent . Human erythroblasts were cultured in the presence or absence of 1 mM EGTA and immunostained with a rabbit antibody to RHP as previously described [29].
CaM binding sequence in RHP [27,28] ( Figure 5). its binding to GPC and p55 has implications for the function of this 4.1R isoform in early erythroblasts. Indeed, while low levels of Ca 2+ in early erythroblasts will lead to membrane association through high-affinity interaction with band 3, increasing levels of Ca 2+ during erythroid differentiation will lead to the displacement of the protein from the membrane and to a possible degradation and loss of this isoform from erythroblasts. Our findings that, in early erythroblasts, a fraction of 4.1R 135 is actually associated with the membrane lends support to this hypothesis [29] (Figure 6). Strikingly, in human erythroblasts cultured for 7 days and treated with 1mM EGTA, 4.1R 135 is more clearly distributed at or near the plasma membrane than in nontreated cells ( Figure 7). Precise quantitative measurements of Ca 2+ levels in erythroblasts at different stages of maturation need to be performed to validate further this hypothesis.

PART II: 4.1R 135 and 4.1G in Erythroblasts
4.1R 135 and 4.1G are simultaneously expressed in erythroblasts and in nonerythroid cells, such as epithelial cells [42,43]. Computer analysis of the 3D structure of the 30 kDa domain of 4.1G has demonstrated that its folded clover-like structure is very similar to that of 4.1R [41] (Figure 8). This observation validates the structural basis for 4.1G binding to previously defined 4.1R binding partners through its 30 kDa domain. As observed for the 30 kDa domain of 4.1R, 4.1G could also interact with CaM in a Ca 2+ -independent manner.
Using a combination of computational calculations (aimed at calculating the disorder probability based on PrDOS software analysis), SDS-PAGE analysis and size exclusion chromatography, we established that, like the HP region of 4.1R, the HP region of 4.1G adopts an unstructured state [41]. As expected from their similar structure, R30 and G30 are both folded polypeptides, this 30 kDa region representing the only structured (folded) domain for both proteins [41].  [41]. These findings demonstrate that 4.1G can bind to transmembrane proteins of the erythrocyte membrane through its 30 kDa domain.

Expression of 4.1G and 4.1R
4.1G interacts in vitro with band 3cyt and GPCcyt with K (D) s in the ∼200 nM range. Importantly, the binding affinities of 4.1G for band 3cyt and GPCcyt are different from those of 4.1R 135 despite the presence of an HP region in both proteins. Thus, 4.1G interacts with band 3cyt with a much lower affinity than 4.1R 135 , the reverse being observed for GPCcyt. These differences result mainly from differences in the association rate constant k a . In contrast, both 4.1G and 4.1R 135 interact with p55 with similar affinities [44].
Binding affinities of full length 4.1G and of its 30 kDa domain (G30) for the membrane proteins described above are very similar, suggesting that 4.1G interacts with its binding partners primarily through G30, the headpiece GHP having a negligible effect on these interactions. This is in marked contrast to the interactions of the 30 kDa domain of 4.1R (R30) which are significantly affected by RHP [29].
International Journal of Cell Biology Interestingly, recombinant chimera proteins consisting of either RHP and G30 (RHP-G30) or GHP and R30 (GHP-R30) showed similar binding affinities as G30 and R30. This implied significant differences in the structure and function of RHP and GHP. It should be noted that neither GHP nor RHP binds to any of these membrane proteins.
We showed an important role for the HP region in regulating 4.1R 135 30 kDa domain binding to membrane proteins. Thus, the HP region improves accessibility of the Nlobe to band 3, but impairs accessibility of the α-lobe to GPC whereas it does not have a significant effect on the C-lobe [29]. 4.1G HP does not appear to modulate the accessibility of the three lobes in G30 to their respective binding partners, the binding profile of 4.1G being similar to that of G30.
We demonstrated that 4.1G binds to various previously characterized 4.1R binding partners, including transmembrane proteins band 3, GPC, and p55, through its 30 kDa domain. The HP domain does not affect these interactions. However, Ca 2+ -dependent CaM binding to the HP region has a profound effect on the interaction of 4.1G with its binding partners. The documented binding profiles for 4.1G are markedly different from those previously reported for 4.1R 135 [29]. Since the primary structure of the 30 kDa domain of 4.1G and 4.1R is highly conserved (71% sequence similarity), the differences in binding profiles are likely to arise primarily from the nonconserved HP region.  [17,44], we examined the effect of CaM binding to G30 on its binding properties to membrane proteins using a RHP-G30 chimera protein. Binding affinities of RHP-G30 for band 3cyt, GPCcyt, p55, and CD44cyt were measured in the presence or absence of Ca 2+ and CaM. The K (D) s obtained for each binding partner in the absence of CaM were similar to those obtained with full-length 4.1G. In contrast, binding assays performed with RHP-G30 preincubated with Ca 2+ /CaM showed a major decrease in binding affinity (7-10 fold in K (D) ) of RHP-G30 for band 3cyt, GPCcyt, and p55. These results indicate that although CaM can bind to G30 independently of Ca 2+ , G30 interactions with membrane proteins can be regulated by CaM in a Ca 2+ -dependent manner. These results also indicate that the regulation of 30 kDa domain binding properties by unfolded HP domain has unique features in the case of 4.1R 135 and 4.1G.

Similarities and Differences of CaM
The Ca 2+ concentration dependence of the CaMmodulated interaction of 4.1G with band 3cyt and GPCcyt has been demonstrated [38,44]. The half maximal binding of 4.1R 135 and 4.1G to band 3cyt and GPCcyt occurs at Ca 2+ concentrations in the submicromolar range [39,40], supporting the potential biological relevance of our biochemical findings [29]. Ca 2+ /CaM-dependent modulations of 4.1R 135 and 4.1G binding to membrane proteins may be triggered upon signal transduction during erythroid development. Indeed, it has been documented that, at early stages of erythropoiesis, intracellular calcium levels increased from a basal level of 55 ± 5 nM to 259 ± 49 nM following binding of erythropoietin to its receptor [39]. Such an increase in intracellular calcium levels would be sufficient to modulate the interaction of 4.1R 135 and 4.1G with its binding partners in erythroid cells. Our findings further suggest that 4.1G offers a unique opportunity to explore divergence of protein structure and function during evolution and development. In erythroblasts, we showed that, consistent with earlier reports [42,43], 4.1G and 4.1R 135 are both expressed during terminal erythroid differentiation and that both proteins can interact with common transmembrane proteins, such as band 3, GPC, and p55. Different binding affinities and Ca 2+ /CaM-dependent modulation of interaction with band 3 and GPC suggest that these 4.1 proteins may play specific roles in membrane biogenesis during terminal erythroid differentiation ( Figure 9).
Thus, the unstructured HP domains of 4.1R and 4.1G seem to play a unique role in regulating the membranebinding properties of those proteins. Understanding the structural basis for differences and similarities in 4.1 binding properties will help us unveil novel biological functions for various 4.1 gene products. To that end, we are currently carrying out a structural analysis of the HP-Ca 2+ /CaM complex using NMR and small-angle X-ray scattering (SAXS). These biophysical analyses should help us further understand the structural basis for the regulatory role of the unstructured HP domain.

Conclusion
During erythropoiesis, the HP domain acts as a regulator of 4.1R and 4.1G interaction with the plasma membrane. We hypothesize that these regulatory properties are in part the result of the unstructured conformation of the HP region. We also show that these regulatory properties depend on intracellular calcium concentrations, with these concentrations varying during erythropoiesis. Thus, the function of the HP domain may evolve depending on the structure of the 4.1 protein isoforms expressed at each stage of erythropoiesis.

Future Studies on 4.1R 135 and 4.1G
This paper focuses on the structure and function of the N-terminal intrinsically disordered region (HP) and membrane-binding FERM domain of 4.1R 135 and 4.1G and on the role of Ca 2+ in regulating binding to membrane proteins through CaM. Our findings are based on in vitro binding assays. Direct evidence for these interactions and their regulations in living cells remains to be established. Although it is known that the RHP contains phosphorylation sites [28,45], the relationship between Ca 2+ /CaM regulation and phosphorylation remains to be investigated. 4.1G binds to spectrin/actin [46,47] and receptors through its Cterminal region [48,49]. Does Ca 2+ /CaM binding to HP also regulate these interactions? Answering such mechanistic questions will help us define the biological significances of 4.1R 135 and 4.1G in the late stage of erythropoiesis.