Cold agglutinins (CAs) are IgM autoantibodies characterized by their ability to agglutinate in vitro RBC at low temperatures. These autoantibodies cause hemolytic anemia in patients with CA disease. Many diverse Ags are recognized by CAs, most frequently those belonging to the I/i system. These are oligosaccharides composed of repeated units of N-acetyllactosamine, expressed on RBC. The three-dimensional structure of the Fab of KAU, a human monoclonal IgM CA with anti-I activity, was determined. The KAU combining site shows an extended cavity and a neighboring pocket. Residues from the hypervariable loops VHCDR3, VLCDR1, and VLCDR3 form the cavity, whereas the small pocket is defined essentially by residues from the hypervariable loops VHCDR1 and VHCDR2. This fact could explain the VH4-34 germline gene restriction among CA. The KAU combining site topography is consistent with one that binds a polysaccharide. The combining site overall dimensions are 15 Å wide and 24 Å long. Conservation of key binding site residues among anti-I/i CAs indicates that this is a common feature of this family of autoantibodies. We also describe the first high resolution structure of the human IgM CH1:CL domain. The structural analysis shows that the CH1-CL interface is mainly conserved during the isotype switch process from IgM to IgG1.

Cold agglutinins (CAs)3 are IgM autoantibodies characterized by their ability to agglutinate in vitro RBC at low temperatures (4–22°C) (1, 2). These autoantibodies cause hemolytic anemia in patients with CA disease (CAD). CAs appear in the context of monoclonal gammopathies secondary to B cell dyscrasias ranging from benign to malignant lymphoproliferation (3, 4, 5). They can also be detected in normal patients at low titers (6), though these titers increase with different infectious processes (7, 8, 9).

Many diverse Ags are recognized by CAs, most frequently those belonging to the I/i system. These are oligosaccharides composed of repeated units of N-acetyllactosamine, expressed in a linear form (i) on fetal RBC or in a branched form (I) on adult RBC.

Early studies revealed that the vast majority of Abs with CA activity reacted with a polyclonal antiserum (10) generated by an isolated CA molecule. This phenomenon is called “idiotypic cross-specificity” because these types of autoantibodies share an antigenic structure (or idiotope). The rat mAb 9G4 (11) served to identify the cross-reacting idiotope associated with the expression of a particular variable domain of the heavy chain (VH) region. Amino acid and nucleotide sequence analyses have confirmed this VH as derived from the VH4-34 (VH4-21) germline gene (12, 13). The presence of a VH chain derived from VH4-34 is necessary both for the CA property and for the idiotope that is recognized by 9G4 mAb. Abs derived from the VH4-34 gene also recognize different autoantigens as is the case of some rheumatoid factors (14), anti-DNA Abs (15), and the anti-D Abs (16).

Anti-I/i CAs show a great variability in VHCDR3, suggesting that this hypervariable region could not be directly involved in the specific recognition or that the mode of binding is different among CAs. Li et al. (17) postulated that the VHFR1 region is essential in the specific recognition of the anti-I/i system, whereas the VHCDR3 and variable domain of the light chain (VL) dictate the fine specificity and strength of binding. Most of the anti-I CAs VL domains are derived from the VκIII germline gene, although some are encoded by VκI or VκII. In contrast, CAs with anti-i activity make no preferential usage of L chains.

The IgM KAU CA was obtained from the serum of a patient suffering CAD. Its amino acid sequence has been determined by Leoni and coworkers (18). As all CAs that recognize the I/i system, the VH domain is derived from the VH4-34 germline gene, showing a single point mutation in the VHCDR1 (Gly31Asp). The VL domain is derived from the kv305 germline gene (VκIIIb). The Fab from IgM KAU was crystallized, and preliminary x-ray diffraction data was reported (19).

Here we present the three-dimensional structure of the Fab KAU. Its combining site shows an extended cavity, as expected of an anti-carbohydrate Ab. Conservation of key binding site residues among anti-I/i CAs indicates that this is a common feature of this family of autoantibodies. We also describe the first high-resolution structure of the human IgM first constant domain of the heavy chain:constant domain of the light chain (CH1:CL) domain and compare its features with those of other human and murine isotypes.

In a previous work (19) we described the preliminary x-ray diffraction of a single crystal of Fab KAU to a resolution of 2.8 Å. Using synchrotron light at the Laboratorio Nacional de Luz Sincrotron (LNLS, Campinas, Brazil) we obtained a data set comprising 31,333 unique reflections (96.8% complete between 13.0 and 2.8 Å resolution, 99.0% complete in the last resolution shell between 2.9 and 2.8 Å).

This crystal belongs to P3121 space group with cell dimension of a = b = 114, 23 Å, c = 172, 78 Å; α = β = 90°, γ = 120°. Estimation of solvent content in this crystal, using the Matthews coefficient (20), indicated that there are two Fabs in the asymmetric unit (Vm = 3.38 Å3Da−1 and solvent content = 63%). The crystal structure of KAU Fab was initially determined at 2.8 Å by the molecular replacement method using the program AMoRe (21). The search model consisted of the Fab of Ab 3d6 (22) (Protein Data Bank (PDB) code 1dfb). Several of the best rotation function solutions were used for the translation search. The seven best translation function solutions were subjected to rigid body refinement using reflections in the resolution range 13.8–3.5 Å. The two most significant solutions have correlation coefficients of 0.219 and 0.210, whereas all other solutions were below 0.145. Combining these two solutions resulted in an R factor = 0.497 and correlation coefficient = 0.344. In all of these calculations, the search model elbow angle was not changed.

Several cycles of model building with the program O (23) followed by refinement of atomic coordinates with Refmac (24) reduced the Rfree factor to 0.268, the R factor to 0.216, root mean square deviation (rmsd) bond lengths to 0.016 Å, and rmsd bond angles to 3.3°. Appropriate amino acid substitutions were made in the model structure using the KAU primary structure (18). Omit maps were calculated after the final cycle of refinement to check the integrity of the model.

At this point, data to 2.28 Å from another trigonal crystal form became available (space group P3121, cell dimensions a = b = 110.9 Å, c = 170.8; α = β = 90°, γ = 120°; Vm = 3.50 Å3Da−1, and solvent content 65%). This data set was obtained from a single crystal grown as described (19). Briefly, Fab KAU was concentrated to 5.5 mg/ml and crystallized using 17% PEG 8000 and 0.1 M HEPES, pH 7.5, as mother liquor, and the temperature was kept at 4°C. This crystal was subsequently cryocooled using 12% PEG 400 as cryoprotectant, and a data set was collected at −123°C using Synchrotron light at the Laboratorio Nacional de Luz Sincrotron. At this step, molecular replacement using the 2.8 Å structure as a probe was performed with this new data set. Refinement was initiated with a rigid body procedure using the program CNS (25) followed by ten cycles of manual model adjustment and addition of solvent water molecules using the program TURBO (26) and simulated annealing refinement using the program CNS, resulting in R factor = 0.18 and Rfree = 0.22. This model consists of two Fab molecules with a total of 6594 protein atoms; 614 solvent molecules have been included in the model structure (see Table I).

Table I.

Refinement statistics summary of the Fab fragment of the anti-I KAU Ab

Program CNS  
Resolution range (Å) 17.0–2.28 
Unique reflections (F > 0; F/ς > 1) 45,494 
Final Rcrysta 0.180 
Final Rfreeb 0.224 
rmsd bond length (Å) 0.023 
rmsd bond angle (°) 2.4 
Mean temperature factor (Å2 
Fab1 (main) 31.5 
Fab1 (side) 29.9 
Fab2 (main) 29.9 
Fab2 (side) 28.2 
Program CNS  
Resolution range (Å) 17.0–2.28 
Unique reflections (F > 0; F/ς > 1) 45,494 
Final Rcrysta 0.180 
Final Rfreeb 0.224 
rmsd bond length (Å) 0.023 
rmsd bond angle (°) 2.4 
Mean temperature factor (Å2 
Fab1 (main) 31.5 
Fab1 (side) 29.9 
Fab2 (main) 29.9 
Fab2 (side) 28.2 
a

Rcryst = Σhkl∥Fo| − |Fc∥/Σhkl|Fo|

b

Rfree is calculated in the same manner as Rcryst, but from 4599 reflections not used for refinement.

Results of the refinement of the KAU Fab crystal structure are summarized in Table I. In one of the Fabs present in the unit cell CH1 residues 135–140 and 199–201 have no associated electron density and have been removed from the model. Likewise, CH1 residues 135–142 and 199–201 in the second Fab have been removed due to lack of electron density. All residues in the VL, CL, and VH domains of both molecules are well resolved. Only two nonglycyl residues in each Fab, VL Ala52 and VH Asp31, are in disallowed regions of a Ramachandran plot (data not shown). Similar occurrences in tight turns have been observed in other Fabs (27, 28, 29).

Calculation of a Fo-Fc difference Fourier map using the final model coordinates reveals electron densities at the CH1 glycosylation site. This density is continuous with the side chain of CH1 Asn166, indicating covalent bonding to the Fab. However, this density is unresolved, with breaks in the glycosidic linkages, and is only approximately the length of a trisaccharide. Attempts to model and refine an oligosaccharide into this density resulted in significant (>2.5 ς) error peaks in a Fo-Fc Fourier.

Superposition between the two Fabs using all Cα atoms gave a rmsd of 0.63 Å (0.48 Å using only Cα atoms <2 rmsd differences). Superposition of the two Fvs gave values of 0.34 Å (0.30 Å); meanwhile, the superposition of the CH1:CL modules resulted in a rmsd of 0.50 Å (0.27 Å).

Solvent-excluded surface areas were computed with the probe radius of 1.7 and 1.4 Å using the program GRASP (30). Superposition alignments were made using the programs ALIGN (31) and Macromodel (32). The final coordinates have been deposited in the RCSB Protein Data Bank (RCSB ID: PDB 1DN0).

The Fab KAU shows the usual Ig fold of Ab molecules (33). The two Fabs in the asymmetric unit are positioned in opposite directions and have their interfaces at the C domains. The two Fabs can be superimposed by a rotation of ∼180°.

The final coordinates of the Fab A-B include 429 residues, 214 in the heavy chain and 215 in the light chain, whereas for the other Fab in the asymmetric unit, 212 amino acids of the heavy chain could be located. The quantitative agreement between the two Fabs gives a measure of the accuracy of the structural analysis (see Materials and Methods). The elbow angle made by the two pseudodyads of the variable and constant domains is 170° for both Fabs. This value falls within the range of those of other Fabs, from 127° to 227° (34). cis-P proline residues occur at position 8, 96, and 142 for the light chain and at position 155 for the heavy chain.

The quaternary structure of this region is similar to the previously described human and mouse Abs. The Fab KAU VL (VκIIIb) and VH (VH4-34) domains form a compact module like other Fvs with an average buried surface area of 1587 Å2 (788 Å2 for VH and 799 Å2 for VL). The VL-VH interaction is mediated by 220 atomic contacts made mainly by VH residues Leu45, Trp47, Tyr106, and Trp110 and VL Pro44, Tyr92, Leu97, and Phe99. Residues Trp47H (VHFR2), Tyr92L, and Leu97L (VLCDR3) are also part of the combining site (see below).

The conformations of the hypervariable loops of the Fab KAU are all in defined canonical groups (35, 36). The complementarity-determining regions (CDRs) H1, H2, L2, and L3 all belong to their respective class 1 canonical groups. KAU L1 could belong to canonical group 6, or a subset thereof, due to its length and similar sequence with the L1 loop of Fab 1f7 (PDB file 1fig). Fab 1f7 is at present the only structure assigned that exhibits the canonical length and conformation of L1 group 6 (37). However, the rmsd of main chain atoms between KAU L1 and 1f7 L1 (1.70 Å) is quite high (Fig. 1 B). Fab 409.5.3 (PDB file 1aif), like KAU and 1f7, also contains 8 residues in L1. Likewise, the rmsd of the main chain atoms with those of KAU (1.57 Å) is also too high for consideration of canonical class. For the 1f7 structure, the difference in conformation of L1, relative to KAU L1, is dominated by two changes in the orientation of carbonyl groups (residues 26 and 29). For the Fab 409.5.3 structure, the difference in conformation relative to KAU appears to be by packing of residues 25 (Ala in KAU, Val in 409.5.3) and 29 (Val in KAU, Ile in 409.5.3) with the extra bulk of the Val-Ile interaction in 409.5.3, resulting in a higher loop conformation at residue 29 and a subsequent change in the positions of residues 30–31. Moreover, Tyr 71 in 409.5.3 makes a hydrogen bond to the amide of residue 31. This hydrogen bond is missing in the KAU structure as a result of the Phe framework residue. As such, the conformational differences between KAU L1 and 409.5.3 L1 bear a striking similarity to the A and B subclasses of an L1 class 2 loop with the subclasses defined by interactions with residue 71. The resolution at which the 1f7 (3.0Å) and Fab 409.5.3 (2.9 Å) structures were solved, precluded their analysis of canonical structure by Al-Lazikani et al. (38). As such, the characteristics of L1 canonical group 6 are still open to further developments.

FIGURE 1.

A, Stereo view of the omit map of L1 KAU (residues 24–35) contoured at a significance level of 2.5 ς. B, Stereo view of the superposition of L1 KAU (thick trace) and L1 1f7 (thin trace).

FIGURE 1.

A, Stereo view of the omit map of L1 KAU (residues 24–35) contoured at a significance level of 2.5 ς. B, Stereo view of the superposition of L1 KAU (thick trace) and L1 1f7 (thin trace).

Close modal

The KAU combining site shows an extended cavity (39, 40) and a neighboring pocket (see Fig. 2 A). Residues from the hypervariable loops VHCDR3, VLCDR1, and VLCDR3 form the cavity whereas the small pocket is defined essentially by residues from the hypervariable loops VHCDR1 and VHCDR2.

FIGURE 2.

Analysis of the KAU combining site. A, Combining site surface colored by electrostatic potential (GRASP). Residues that constitute the extended groove are labeled. B, Conservation of combining site residues among 32 sequenced anti-I/i CAs (as described in references and BLAST database). The figure is color coded based on the percentage of CAs containing the same residue at this particular position. Purple, 66–100%; red, 33–66%; yellow, 0–33%.

FIGURE 2.

Analysis of the KAU combining site. A, Combining site surface colored by electrostatic potential (GRASP). Residues that constitute the extended groove are labeled. B, Conservation of combining site residues among 32 sequenced anti-I/i CAs (as described in references and BLAST database). The figure is color coded based on the percentage of CAs containing the same residue at this particular position. Purple, 66–100%; red, 33–66%; yellow, 0–33%.

Close modal

The extended cavity floor is composed by residues Tyr92L, Leu97L (VLCDR3), Pro98H (VHCDR3), and Glu50H (VHFR2). The cavity is surrounded by residues Gln58H (VHFR3), Trp47H (VHFR2), Ser94L, Ser93L, Gly92L (VLCDR3), Ser30L, Ser31L, and Tyr33L (VLCDR1).

The cavity floor communicates with the pocket floor by means of Pro98H (VHCDR3). The pocket floor is continued by Tyr33H and some main chain atoms of residues Asp31H (VHCDR1), Asp101H, and Thr102H (VHCDR3). The pocket is completely surrounded by polar and charged residues like Thr102H, Asp101H, (VHCDR3), Asp31H (VHCDR1), His53H, Asn52H, Ser54H, and Ser56H (VHCDR2).

The first CH domain of the human IgM (Cμ1) displays the basic Ig fold as do other constant regions from different isotypes (41). The Cμ1 and CL domains are covalently joined by a disulfide bond between the residues Cys134H and Cys215L. Both domains face each other, related by a pseudodyad angle of 167.0°.

The interaction between Cμ1 and CLk results in a buried surface of 2211 Å2 (1158 Å2 for Cμ1 and 1053 Å2 for CL). The Van der Waals contacts, hydrogen bonds, the disulfide bond, and the salt bridge between Cμ1 and CL are listed in Table II. The salt bridge occurs between Lys219H (NZ) and Glu124L (OE2) with a distance of 2.5 Å and forming an angle of 104.5°. It is worth noting that Glu124L also participates in an equivalent salt bridge with Lys221H of CH1 from mouse IgG2aκ isotype (41) (as shown in Abs 17/9 and 33F12, PDB entries 1HIL, Ref. 42 and 1AXT, Ref. 43 , respectively) and mouse IgG1κ (Ab NMC-4, PDB entry 1OAK) (44).

Table II.

Atomic contacts between the constant domains of the light and heavy chains of Fab KAUa

F 129P 130L 131V 132S 133C 134A 145L 149Q 151R 174G 175F 176P 177V 179L 180R 181T 188Q 190L 192K 219Totals
F 117                   11 
I 118                   
F 119   16 10               40 
P 120                    
S 122                   
E 124                   5⧫ 13 
Q 125 19        3*            22 
S 132        3*            
V 134                    
L 136                    
N 138          6*         16 
N 139          6*           
Q 161              11     18 
E 162                    
S 163            10**       17 
V 164                    
T 165                  
S 175          4*          10 
L 176                    
S 177            10     3*    13 
T 179                    
T 181                    
C 215      1•               
Totals 32 21 11 13 14 16 32 15 17 16 223 
F 129P 130L 131V 132S 133C 134A 145L 149Q 151R 174G 175F 176P 177V 179L 180R 181T 188Q 190L 192K 219Totals
F 117                   11 
I 118                   
F 119   16 10               40 
P 120                    
S 122                   
E 124                   5⧫ 13 
Q 125 19        3*            22 
S 132        3*            
V 134                    
L 136                    
N 138          6*         16 
N 139          6*           
Q 161              11     18 
E 162                    
S 163            10**       17 
V 164                    
T 165                  
S 175          4*          10 
L 176                    
S 177            10     3*    13 
T 179                    
T 181                    
C 215      1•               
Totals 32 21 11 13 14 16 32 15 17 16 223 
a

The heavy chain residues are across the page and the light chain residues are down the page. The numbers represent van der Waals contacts otherwise indicated as follows: * represent a hydrogen bond, • represents a disulphide bond, and ⧫ represents a salt bridge.

In total, Cμ1 and CL establish 223 atomic contacts in Fab KAU A-B (247 contacts in Fab C-D). Light chain residues Phe119L, which makes 40 contacts, and Gln125L, which makes 22 contacts, account for 22% of contacts with Cμ1. On Cμ1, residues Phe129H, which makes 32 contacts, and Phe176H, which makes 32 contacts, represent 27% of Cμ1 contacts with CL. Residues Gln151, Arg174, Pro177, and Thr188 of the heavy chain make eight hydrogen bonds with CL (see Table II).

Electron density corresponding to a trisaccharide moiety (see Materials and Methods) can be visualized in the Cμ1 region, attached to Asn166H by means of a N-glycan bond, but structural disorder prohibits placing the carbohydrate in the final model due to breaks in electron density at the glycosidic linkages.

This is the first high-resolution structural analysis of the Fab from a human IgM, allowing us to compare the module Cμ1-CL with the corresponding human and murine domains from other isotypes.

A previous Fabμ structure was solved at a 3.2 Å resolution from a human rheumatoid factor IgM (RF-AN, PDB 1ADQ) (45). In that case, the structure of the constant segment is less well defined due to disorder in this region. This mobility is less pronounced at the Cμ1/CLλ interface, allowing us to compare the contact residues at this interface. Eighteen of the twenty residues of Cμ1 KAU that contact CLκ make equivalent contacts on Cμ1 RF-AN with CLλ. This analysis shows a remarkable conservation in this interface regardless of the isotype of the partner light chain.

The constant domains CH1 from KAU (human IgM), HIL (human IgG1) (46), D44.1 (murine IgG1) (47) and Bv04–01 (murine IgG2b) (48) were aligned by a least-squares superposition to compare the folding conservation among these isotypes. The α-carbon structure of the IgM KAU CH1 domain is very similar to all other CH1 domains analyzed (Fig. 3). The rmsd obtained for the Cα positions for the 73 spatially corresponding residues of KAU vs other isotypes are 0.83Å for human IgG1, 0.94Å for murine IgG1, and 1.03Å for IgG2b. The Cμ1 domain superimposes well overall with the exception of two segments, 163–168 and 194–204.

FIGURE 3.

Superposition of the Cα backbone of: human Cμ1 (yellow), human Cγ1 (blue), murine Cγ1 (magenta), and murine Cγ2b (red). A, Front stereo view of the CH1 face contacting CL. Cysteine residues of human IgM (yellow sticks) and murine IgG1 (light blue sticks) are highlighted to show their different location at the interface. B, Side view of the same domains, the conservation of the interface is clearly shown at the right side of the figure.

FIGURE 3.

Superposition of the Cα backbone of: human Cμ1 (yellow), human Cγ1 (blue), murine Cγ1 (magenta), and murine Cγ2b (red). A, Front stereo view of the CH1 face contacting CL. Cysteine residues of human IgM (yellow sticks) and murine IgG1 (light blue sticks) are highlighted to show their different location at the interface. B, Side view of the same domains, the conservation of the interface is clearly shown at the right side of the figure.

Close modal

In the region 163–168, the IgM CH1 domain deviates substantially from the other isotypes conformation (Fig. 3,B), a fact that can be explained, at least in part, by the presence of a covalently attached carbohydrate at residue Asn166 of the IgM. Human IgM conformation at residues 163–168 has a similar orientation to that of murine IgG2b, where this loop is pointing toward the hinge region, in clear contrast with the conformation observed in human IgG1, murine IgG1, (Fig. 3 B) IgG2a and IgG3 CH1 domains (49).

The backbone conformation of the IgM CH1 domain also differs significantly from other isotypes structure in the segment 194–204. This loop is three amino acids longer in human IgM and displays a more extended conformation, even though residues 199–201 were not modeled due to poor electron density.

The cis-Pro155 in Fab KAU aligns with its counterpart cis in the isotypes analyzed, as well as the intrachain S-S bond formed by Cys148-Cys208 in KAU. A very mobile segment formed by residues 135–140 in Fab KAU corresponds to a mobile segment in the rest of the isotypes at similar position.

The CH1 half-cystine forming the interchain disulfide bond with the light chain is located at the N-terminal end of the domain in most of the isotypes. IgG1 is an exception, where the cystine donated by the heavy chain is found at the carboxyl end of CH1 (50). As shown in the Fig. 3 A, in human IgM the half-cystine is in the loop between the strands 4-1 (A) and 4-2 (B) of the four-stranded β sheet, whereas in murine IgG1 is located in the C-terminal loop arising from the strand 3-3 (G), pertaining to the three-stranded β sheet. Consequently, the half-cystine in both isotypes do not superimpose, but their counterpart in the light chain (located also in a loop) can accommodate itself to interact to form the interchain disulfide bond.

The structural alignment also shows that the CH1-CL interface is mainly conserved during the isotype switch process from IgM to IgG isotypes. In effect, Fig. 3 shows the close correspondence of the α-carbon trace at the interface. This result is in agreement with the sequence alignment analysis of IgM against IgG1 isotypes made by Padlan and coworkers (51).

The usage of the VH4-34 gene is very common among autoantibodies, such us CAs, rheumatoid factors, anti-Rh, and anti-DNA Abs. This restricted gene usage is striking in the case of anti-I/i CAs where all the analyzed Abs belong to this family. Therefore, the structural description of VH KAU, encoded by this gene, is important to understand the molecular basis of anti-I/i autoimmune recognition.

The mAb 9G4 recognizes a public idiotope present in all CAs encoded by VH4-34 (11). As was described before, the area reacting with 9G4 is localized in framework region 1 (FR1; residues 23–25). Accordingly, previous studies (52, 53) showed that binding of anti-idiotypic Ab (anti-Id) 9G4 to the FR1 region of VH4-34-encoded Abs blocked the hemagglutination activity of CAs with I/i activity. The KAU three-dimensional structure shows that the 9G4 cross-reactive idiotope (residues 23–25 of FR1) and the combining site do not overlap. The inhibitory effect of 9G4 anti-Id could be explained by an induced conformational change on FR1 upon binding of the anti-Id or by steric hindrance because of the bulky nature of the anti-Id.

The topography of the combining site (Fig. 2 A) shows a pocket and a cavity formed by hydrophobic and aromatic residues, surrounded by polar and charged amino acids such as Asp31H, Asp101H, and Glu50H. Comparison with the binding site of the soybean lectin bound to the Ag I of RBC (SBA, PDB files 2sba, 1sbd, 1sbe) (54, 55), shows that both have a similar topography and overall residue composition. The soybean agglutinin has a pocket formed by hydrophobic residues (Phe128, Ala87, Ile216, Leu214) and make hydrogen bonds with the I Ag by means of surrounding residues, which typically participate in such interactions with carbohydrate moieties (Asp88, Asp215, Asn130, His104). That pocket resembles the Fab KAU combining site top-ography, consequently, the way KAU interacts with the I Ag could be similar.

The topography of the KAU combining site is consistent with the binding activity and genetic origin of human CAs (56, 57, 58, 59). Functional assays show that RBC agglutination produced by KAU is abrogated by treatment of the cells with β-endogalactosidase and inhibited by the linear (i) and branched (I) Ags (A. C., unpublished observation), suggesting that KAU recognizes a large carbohydrate Ag.

To gain insight about the common structural features between anti-I/i autoantibodies, all known sequences of anti-I/i CAs (see Fig. 2 B) were analyzed in terms of the Fab KAU CDRs and combining site. KAU H1 loop (residues 26–33 of the VH region) folds with the group 1 canonical structure. All the important residues for this conformation (60) are conserved in all the analyzed anti-I/i CAs, suggesting that the packing and conformation of KAU H1 loop is similar in all human anti-I/i. Most of sequenced anti-I/i CAs have the same residues in the H2 loop (residues 52–56 of the VH region). As in KAU, they should have a canonical group 1 H2 structure.

It was also postulated that the VH4-34-encoded region is mainly responsible for Ag I/i specificity, while VHCDR3 and VL modulate the affinity (17). Fig. 2 B shows the KAU combining site with its residues colored according to the degree of conservation of those amino acids among anti-I/i CAs. It is clear that most of VHCDR1 and all VHCDR2 residues are conserved among anti I/i CAs. These two loops form the pocket’s wall. FR2 residues Trp 47H, Glu 50H and FR3 residue Asn 58H, that form the first segment of the external perimeter, and Ser93L and Ser94L (both of VLCDR3), that extend this outer limit of the cavity, are remarkably conserved. Therefore, all of these highly conserved residues constitute most of the external limit of the whole combining site.

The small pocket at the KAU combining site, formed essentially by VH4-34 encoded residues (VH CDR1, FR2, and CDR2) could explain the VH4-34 restriction among CAs. Differences in composition at VHCDR3, VLCDR3, and VLCDR1 would explain the diversity in the fine specificity of this family of autoantibodies. Further three-dimensional structural studies of other anti-I/i CAs, free and complexed with its carbohydrate Ags are needed to ascertain the structural basis of this autoimmune recognition.

We thank Dr. Roberto J. Poljak for the critical reading of the manuscript, and Dr. Ricardo A. Margni and Dr. José M. Delfino for collaboration and encouragement.

1

This work was supported by grants and fellowships from the Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), Fundación Antorchas, and the University of Buenos Aires (to A.C., J.L., and F.A.G.). B.C.B. was supported by National Institutes of Health Grant 1R15AI44790-01 and National Aeronautics and Space Administration Grant NCC5-232 (Model Institutes for Excellence). J.G.C., R.A., and I.P. were supported by Fundaçao de Amparo à Pesquisa do Estado de Sao Paulo (via Grant 1996/2285-5) and Conselho Nacional de Pesquisas (Brazil).

3

Abbreviations used in this paper: CA, cold agglutinin; CAD, CA disease; VH, variable domain of the heavy chain; VL, variable domain of the light chain; CDR, complementarity-determining region; FR, framework region; CH1, first constant domain of the heavy chain; CL, constant domain of the light chain; Cμ1, first constant domain of the IgM μ heavy chain; anti-Id, anti-idiotypic Ab; rmsd, root mean square deviation; PDB, Protein Data Bank.

1
Roelcke, D..
1989
. Cold agglutination.
Transfus. Med. Rev. III.
2
:
140
2
Ciejka, J. Z., E. B. Cook, D. Lawler, J. Martin, R. D. Woodson, F. Graziano.
1999
. Severe cold agglutinin disease and cryoglobulinemia secondary to a monoclonal anti-Pr2 IgM lambda cryoagglutinin.
Clin. Exp. Rheumatol.
17
:
227
3
Ramos, R. R., B. R. Curtis, C. S. Eby, G. A. Ratkin, H. Chaplin.
1994
. Fatal outcome in a patient with autoinmune haemolytic anemia associated with an IgM bithermic anti-ITP.
Transfusion
34
:
427
4
Sandhaus, L. M., K. Raska, Jr, H. V. Wu.
1986
. Diffuse large cell lymphoma with monoclonal IgMκ and cold agglutinin.
Am. J. Clin. Pathol.
86
:
120
5
Berentsen, S..
1995
. Chronic cold agglutinin disease.
Tidsskr. Nor. Laegeforen.
115
:
473
6
Jefferies, L. C., C. M. Carchidi, L. E. Silberstein.
1993
. Naturally occurring anti-i/I cold agglutinins may be encoded by different VH3 genes as well as the VH4-21 gene segment.
J. Clin. Invest.
92
:
2821
7
Janney, F. A., L. T. Lee, C. Howe.
1978
. Cold hemagglutinin cross-reactivity with Mycoplasmapneumoniae.
Infect. Immun.
22
:
29
8
Rosenfield, R., P. J. Schmidt, R. C. Calvo, M. H. McGinniss.
1965
. Anti-i, a frequent cold agglutinin in infectious mononucleosis.
Vox Sang.
10
:
631
9
Pruzanski, W., D. Roelcke, E. Donnelly.
1986
. Persistent cold agglutinins in AIDS and related disorders.
Acta Haematol.
75
:
171
10
Silverman, G., F. Goni, J. Fernandez, P. P. Chen, B. Frangione, D. A. Carson.
1988
. Distinct patterns of heavy chain variable region subgroup use by human monoclonal autoantibodies of different specificity.
J. Exp. Med.
168
:
2361
11
Potter, K. N., Y. Li, V. Pascual, R. C. Williams, Jr, L. C. Byres, M. Spellerberg, F. K. Stevenson, J. D. Capra.
1993
. Molecular characterization of a cross-reactive idiotope on human immunoglobulins utilizing the VH4-21 gene segment.
J. Exp. Med.
178
:
1419
12
Pascual, V., K. Victor, D. Lelsz, M. B. Spellerberg, T. J. Hamblin, K. M. Thompson, I. Randen, J. Natvig, J. D. Capra, F. K. Stevenson.
1991
. Nucleotide sequence analysis of the V regions of two IgM cold agglutinins: evidence that the VH4-21 gene segment is responsible for the major cross reactive idiotype.
J. Immunol.
146
:
4385
13
Silberstein, L. E., L. C. Jefferies, J. Goldman, D. Friedman, J. S. Moore, P. C. Nowell, D. Roelcke, W. Pruzanski, J. Roudier, G. J. Silverman.
1991
. Variable region gene analysis of pathologic human autoantibodies to the related i and I red blood cell antigens.
Blood
78
:
2372
14
Kraj, P., D. F. Friedman, F. Stevenson, L. E. Silberstein.
1995
. Evidence for the overexpression of the VH4-34 (VH4.21) Ig gene segment in the normal adult human peripheral blood B cell repertoire.
J. Immunol.
154
:
6406
15
Stevenson, F. K., C. Longhurst, C. J. Chapman, M. Ehrenstein, M. B. Spellerberg, T. J. Hamblin, C. T. Ravirajan, D. Latchman, D. Isenberg.
1993
. Utilization of the VH4-21 gene segment by anti-DNA antibodies from patients with systemic lupus erythematosus.
J. Autoimmun.
6
:
809
16
Borretzen, M., C. Chapman, F. K. Stevenson, J. B. Natvig, K. M. Thompson.
1995
. Structural analysis of VH4-21 encoded human IgM allo- and autoantibodies against red blood cells.
Scand. J. Immunol.
42
:
90
17
Li, Y., M. B. Spellerberg, F. K. Stevenson, J. D. Capra, K. N. Potter.
1996
. The I binding specificity of human VH4-34 (VH4-21) encoded antibodies is determined by both VH framework region 1 and complementary determining region 3.
J. Mol. Biol.
256
:
577
18
Leoni, J., J. F. Ghiso, F. Goni, B. Frangione.
1991
. The primary structure of the Fab fragment of protein KAU, a monoclonal immunoglobulin M cold agglutinin.
J. Biol. Chem.
266
:
2836
19
Cauerhff, A., I. Polikarpov, I. Mathov, L. Plotkin, C. Abatangelo, F. Goldbaum, J. Leoni.
1998
. Crystallization and preliminary diffraction studies of a human Fabμ with anti-I activity.
Prot. Pept. Letters
5
:
177
20
Matthews, B. W..
1968
. Solvent content of protein crystals.
J. Mol. Biol.
33
:
491
21
Navaza, J..
1994
. AMoRe: an automated package for molecular replacement.
Acta Crystallogr. A
50
:
157
22
He, X. M., F. Ruker, E. Casale, D. Carter.
1992
. Structure of a human monoclonal antibody Fab fragment against gp41 of human immunodeficiency virus type 1.
Proc. Natl. Acad. Sci. USA
89
:
7154
23
Jones, T. A., M. Kjeldgaard.
1993
.
Refmac, a New Refinement Program: O Version 5.9, The Manual
Uppsala University, Uppsala, Sweden.
24
Murshudov, G. N., A. A. Vagin, E. J. Dodson.
1997
.
Acta Crystallogr. D
53
:
240
25
Brunger, A. T., P. D. Adams, G. M. Clore, W. L. Delano, P. Gros, R. W. Grosse-Kunstleve, J. S. Jiang, J. Kuszewski, M. Nilges, N. S. Pannu, et al
1998
. Crystallography & NMR system: a new software system for macromolecular structure determination.
Acta Crystallogr. D
54
:
905
26
Roussel, A., and A. G. Inisan. 1992. TURBO-FRODO. Technopole de Chateau-Gombert, Europarc Bat. C. Marseille, France.
27
Arevalo, J. H., E. A. Stura, M. J. Taussig, I. A. Wilson.
1993
. Three-dimensional structure of an anti-steroid Fab and progesterone complex.
J. Mol. Biol.
231
:
103
28
Schulze-Gahmen, U., J. M. Rini, I. A. Wilson.
1993
. Detailed analysis of the free and bound conformations of an antibody: x-ray structures of Fab 17/9 and three different Fab-peptide complexes.
J. Mol. Biol.
234
:
1098
29
Braden, B. C., H. Souchon, J.-L. Eisele, G. A. Bentley, T. N. Bhat, J. Navaza, R. J. Poljak.
1994
. Three-dimensional structures of the free and the antigen-complexed Fab from monoclonal anti-lysozyme antibody D44.1.
J. Mol. Biol.
243
:
767
30
Nicholls, A., K. A. Sharp, B. Honig.
1991
. Protein folding and association-insights from the interfacial and thermodynamic properties of hydrocarbons.
Proteins Struct. Funct. Genet.
11
:
281
31
Satow, Y., G. H. Cohen, E. A. Padlan, D. R. Davies.
1986
. Phosphocoline binding immunoglobulin Fab McPC603: an X-ray diffraction study at 2.7 Å.
J. Mol. Biol.
190
:
593
32
Mohamadi, F., N. G. Richards, W. C. Guida, R. Liskamp, M. Lipton, C. Caufield, G. Chang, T. Hendrickson, C. Still.
1990
. Macromodel: an integrated software system for modelling organic and bioorganic molecules using molecular mechanics.
J. Comp. Chem.
11
:
440
33
Poljak, R. J., L. M. Amzel, B. L. Chen, R. P. Phizackerly, F. Saul.
1973
. Three-dimensional structure of the Fab′ fragment of a human immunoglobulin at 2,8-A resolution.
Proc. Natl. Acad. Sci. USA
70
:
3305
34
Wilson, I. A., R. L. Stanfield.
1993
. Antibody-antigen interactions.
Curr. Opin. Struct. Biol.
3
:
113
35
Chothia, C., A. M. Lesk.
1987
. Canonical structures for the hypervariable regions of immunoglobulins.
J. Mol. Biol.
196
:
901
36
Chothia, C., A. M. Lesk, A. Tramontano, M. Levitt, S. J. Smith-Gill, G. Air, S. Sheriff, E. A. Padlan, D. Davies, W. R. Tulip, et al
1989
. Conformations of immunoglobulin hypervariable regions.
Nature
342
:
877
37
Tomlinson, I. M., J. P. L. Cox, E. Gherardi, A. M. Lesk, C. Chothia.
1995
. The structural repertoire of the Vκ domain.
EMBO J.
14
:
4628
38
Al-Lazikani, B., A. M. Lesk, C. Chothia.
1997
. Standard conformations for the canonical structures of immunoglobulins.
J. Mol. Biol.
273
:
927
39
Webster, D., A. H. Henry, A. R. Rees.
1994
. Antibody-antigen interactions.
Curr. Opin. Struct. Biol.
4
:
123
40
Padlan, E. A., E. Kabat.
1991
. Modelling of antibody combining sites.
Methods Enzymol.
203
:
3
41
Padlan, E. A..
1994
. Anatomy of the antibody molecule.
Mol. Immunol.
31
:
169
42
Rini, J. M., U. Schulze-Gahmen, I. A. Wilson.
1992
. Structural evidence for induced fit as a mechanism for antibody-antigen recognition.
Science
255
:
959
43
Barbas, C. F., A. 3rd, G. Heine, T. Zhong, S. Hoffmann, R. Gramatikova, B. Bjornestedt, J. List, E. A. Anderson, I. A. Stura, I. A. Wilson, R. A. Lerner.
1997
. Immune versus natural selection: antibody aldolases with enzymatic rates but broader scope.
Science
278
:
2085
44
Celikel, R., K. I. Varughese, A. Madhusudan, J. Ware Yoshioka, Z. M. Ruggeri.
1998
. Crystal structure of the von Willebrand factor A1 domain in complex with the function blocking NMC-4 Fab.
Nat. Struct. Biol.
5
:
189
45
Corper, A. L., M. K. Sohi, V. R. Bonagura, M. Steinitz, R. Jefferies, A. Feinstein, D. Beale, M. J. Taussig, B. J. Sutton.
1997
. Structure of human IgM rheumatoid factor Fab bound to its autoantigen IgG Fc reveals a novel topology of antibody-antigen interaction.
Nat. Struct. Biol.
4
:
374
46
F. A. Saul, R. J. Poljak. Fab Fragment From Human Immunoglobulin Igg1 (Hil, PDB entry 8FAB).
47
Braden, B. C., H. Souchon, J-L. Eisele, G. A. Bentley, T. N. Bhat, J. Navaza, R. J. Poljak.
1994
. Three-dimensional structures of the free and the antigen complexed Fab from monoclonal anti-lisozyme antibody D44.1.
J. Mol. Biol.
243
:
767
48
Herron, J. N., X. M. He, D. W. Ballard, P. R. Blier, P. E. Pace, A. L. Bothwell, E. W. Voss, Jr, A. B. Edmundson.
1991
. An autoantibody to single-stranded DNA: comparison of the three-dimensional structures of the unliganded Fab and a deoxynucleotide-Fab complex.
Proteins
11
:
159
49
Sheriff, S., P. D. Jeffrey, J. Bajorath.
1996
. Comparison of CH1 domains in different classes of murine antibodies.
J. Mol. Biol.
263
:
385
50
Poljak, R. J..
1975
. X-ray diffraction studies of immunoglobulins.
Adv. Immunol.
21
:
1
51
Padlan, E., G. H. Cohen, D. R. Davies.
1986
. Antibody Fab assembly: the interface residues between CH1 and CL.
Mol. Immunol.
23
:
951
52
Stevenson, F. K., G. J. Smith, J. North, T. J. Hamblin, M. J. Glennie.
1989
. Identification of normal B-cell counterparts of neoplastic cells which secrete cold agglutinins of anti-I and anti-i specificity.
Br. J. Haematol.
72
:
9
53
Thompson, K. M., J. Sutherland, G. Barden, M. D. Melamed, I. Randen, J. Natvig, V. Pascual, J. D. Capra, F. K. Stevenson.
1991
. Human monoclonal antibodies against red blood antigens preferentially express a VH4–21 variable region gene-associated epitope.
Scand. J. Immunol.
34
:
509
54
Dessen, A., D. Gupta, S. Sabesan, C. F. Brewer, J. C. Sacchettini.
1994
. X-ray crystal structure of the soybean agglutinin cross-linked with biantennary analog of the blood group I carbohydrate antigen.
Biochemistry
34
:
4933
55
Olsen, L. R., A. Dessen, D. Gupta, S. Sabesan, J. C. Sacchettini, C. F. Brewer.
1997
. X-ray crystallographic studies of unique cross-linked lattices between four isomeric biantennary oligosaccharides and soybean agglutinin.
Biochemistry
36
:
15073
56
Pascual, V., K. Victor, M. Spellerberg, T. J. Hamblin, F. K. Stevenson, J. D. Capra.
1992
. VH restriction among human cold agglutinins: the VH4-21 gene segment is required to encode anti-I and anti-i specificities.
J. Immunol.
149
:
2337
57
Schutte, M. E. M., J. H. van Es, L. E. Silberstein, T. Logtenberg.
1993
. VH4-21-encoded natural autoantibodies with anti-i specificity mirror those associated with cold hemagglutinin disease.
J. Immunol.
151
:
6569
58
Riboldi, P., G. Gaidano, E. W. Schettino, T. G. Steger, D. M. Knowles, R. Dalla-Favera, P. Casali.
1994
. Two acquired immunodeficiency syndrome-associated Burkitt’s lymphomas produce specific anti-i IgM cold agglutinins using somatically mutated VH4-21 segments.
Blood
83
:
2952
59
Chapman, C. J., M. B. Spelleberg, G. A. Smith, S. J. Carter, T. J. Hamblin, F. K. Stevenson.
1993
. Autoanti-red cell antibodies synthesized by patients with infectious mononucleosis utilize the VH4-34 gene segment.
J. Immunol.
151
:
1051
60
Chothia, C., A. M. Lesk, E. Gherardi, I. M. Tomlinson, G. Walter, J. D. Marks, M. B. Llewelyn, G. Winter.
1992
. Structural repertoire of the human VH segments.
J. Mol. Biol.
227
:
799