A consistent model for the key complex in chronic beryllium disease

2.1 Model description

Our model for the key complex of CBD (1) is derived from a model of Clayton et al. (2) [12], which essentially is a ternary complex with the following components

1. DP2: a protein from the class II of the major histocompatibility complexes (MHC-II), the HLA-DP2 (human leucocyte antigen of type DP2), more specifically its β-chain subunit (that is sometimes also written as HLA-DP2β1) in specific HLA-DPB1 alleles. The (various) HLA-DPB1 alleles that have glutamic acid (E) residues at positions β26E, β68E and β69E.

2. M2: a mimiotype peptide called M2 of sequence FWIDLFETIG with aspartic acid (D) at position 4 (p4D) and another glutamic acid at position 7 (p7E)

3. AV22: a CD4⁺ T-helper cell antigen receptor protein (TCR-AV22)

4. a Be²⁺ and a Na⁺ cation

The refined single crystal X-ray structure 2 (PDB-ID: 4p4k) [12] consists (apart from solvent, water and co-crystallizing agents) mainly of V-shaped and C₂-symmetric dimeric units of complexes containing HLA-DP2β1, HLA-DP2α1, M2, AV22, Na⁺ and Be²⁺. Clayton et al. could confirm that the binding site (S, see Fig. 1a and b) of the metal cations agrees with what previously has been deduced from mutant and allele variant studies of both the DP2 [20], [21], [22], [23] and bound (self)-peptides [20], and it shows a close spatial proximity and possibly chemically bonding interactions between the metal cations and the five previously mentioned carboxylate moieties β26E, β68E, β69E in the DP2 complex and p4D, and p7E in the M2 peptide and in addition with the carbonyl group of the Leu residue at position 5 (p5L). A section of the structure model 2 from two perspectives is shown in Fig. 1a and b.

Fig. 1:

The beryllium binding site S in the key complex of CBD. (a) Binding site S in the X-ray structure model of 4p4k (2) [12], first perspective, with the three carboxylate moieties of DP2 arranged vertically on the left and the two carboxylate moieties of M2 arranged horizontally and in top view. In the center is the original Na⁺+Be²⁺ model. (b) Binding site S in the X-ray structure model of 4p4k (2) [12], second perspective. (c) Schematic illustration of the bis-meridional binding site S in the metal free coordination pocket of the DP2·M2 complex, roughly from the same perspective as in Fig. 1a. (d) Schematic illustration of the binding site S in our new model 1 containing the Be₄O⁶⁺ moiety, roughly from the same perspective as in Fig. 1a.

The five carboxylate groups in S are arranged in two roughly perpendicular arrays, the one consisting of those from β26E, β68E, β69E of the DP2 and the other of those of p4E, and p7E of the mimo-type peptide M2. A diagram that schematically illustrates this specific geometric arrangement is given in Fig. 1c which can be compared with the X-ray structure model shown in Fig. 1a.

Our new model for S [19] contains a [Be₄O]⁶⁺ complex ion as it is well known from the so called “basic beryllium carboxylate” compounds [3] as the coordination center (see Fig. 1d) binding the M2 to the DP2 protein. Its structure can be described as an oxide centered tetrahedron of Be²⁺ cations which is coordinated by six carboxylate ligands in an η ²-mode, i.e. all six edges of the Be₄ tetrahedron are each bridged by one carboxylate group, thus total electro-neutrality and at the same time the optimal tetrahedral arrangement of oxygen around Be²⁺ is achieved. We note that also one example of a non-homoleptic derivative of a basic beryllium carboxylate has been described in the literature [24], which is a product of the formal substitution [RCO₂]⁻→ [OH]⁻ of composition Be₄O(O₂CR)₅ (OH).

2.2 DFT structure optimizations

The structures of 3 (see Fig. 2) and Be₄O(OAc)₆ have been optimized using Turbomole [25], [26] (version 7.2) at the DFT level of theory using the PBE functional [27], [28], Grimmes D3 dispersion correction together with the Becke-Johnson damping [29], [30], [31] and the RI approximation for the Coulomb integrals [25]. The numerical quadrature grid M4 [32] for the exchange-correlation contribution and the def2-TZVP [33] basis set was used. Prior to the chemical shielding calculations with Turbomoles mpshift module [34], single point calculations with stricter convergence criteria (denconv 0.1d-07 and scfconv 8) have been performed.

Fig. 2:

DFT (PBE0+D3/def2-TZVP) optimized structure of the minimal [Be₄O(EGGGEE)(DLFE)]⁺ model (3 ⁺) for the coordination site S in 1, hydrogen atoms omitted for clarity and labels corresponding to the respective positions in 1, see also Fig. 1d).

The Hessian and vibrational frequencies were calculated with Turbomoles aoforce module [35], in order to confirm the structure as a minimum on the potential energy surface.

2.3 ⁹Be quadrupolar coupling and linewidth

In this section the quantum chemistry details of nuclear quadrupole coupling (NQC) are provided, followed by a discussion of half width of the ⁹Be NMR signals.

As a starting point for estimating nuclear quadrupole coupling (NQC), we consider the quadrupole coupling constants (C _Q) and asymmetry parameters (η):

where the absolute electron charge is denoted e, the nuclear quadrupole moment Q (for ⁹Be), the Planck constant h and the eigenvalues of the electric field gradient tensor V _aa, aa=xx, yy, zz, respectively. We provide NQC for the models 3, the basic beryllium formiate: Be₄O(HCO₂)₆ (4) and 4 where one formiate is replaced by formaldehyde as a minimal model for the carbonyl group of the p7L residue of the self peptide (M2): [Be₄O(HCO₂)₅OCH₂]⁺ (5) in Table 2.

3.1 A.

The binding site contains five singly negatively charged carboxylate groups in close spatial proximity. As was previously noted [12] this accumulation of negative charges poses a problem for a model containing a mononuclear Be-complex, since the Be²⁺-cation has a formal electronic charge of only 2+, while the [Be₄O]⁶⁺ coordination center gives an almost perfect match.

If the carbonyl moiety of the p7L residue is taken as an additional donor into consideration (see also Fig. 1c and d and Fig. 3d in ref. [12]), topologically, both the M2– and DP2–ligands provide a meridional arrangement around the coordination site. This bis-meridional arrangement circumscribes (thus requires) an octahedral coordination center, which [Be₄O]⁶⁺ indeed is.

The size of the coordination site in 2 can be estimated from the distance between trans-positioned carbon atoms of the six donor functions, i.e. C ^{p 4 D} –C ^{p 7 E} =5.98 Å, C ^{p 5 L ( C = O )}–C ^{β 68 E} =6.58 Å, and C ^{β 26 E} –C ^{β 69 E} =9.06 Å, respectively. A Comparison with the values from a DFT optimization for a simplified model 3 (vide infra) which yields 5.98 Å, 6.88 Å and 6.17 Å, respectively, shows that the steric requirements of the [Be₄O]⁶⁺ moiety are very well met by the coordination site S. The larger mismatch of the value for C ^{β 26 E} –C ^{β 69 E} is discussed under D.

The [Be₄O]⁶⁺ moiety is composed of four doubly charged cations but still matches the relatively small region of the coordination site S. This can be explained by the short average distance of about 1.6 Å between Be and the central oxygen atom. Thus effectively, the [Be₄O]⁶⁺ coordination center is only about the 1.5-fold size of a methane molecule (CH 4 ) and the [Be₄O]⁶⁺ moiety meets almost perfectly the electrostatic, topologic and spatial requirements of the coordination site S.

3.2 B.

To the best of our knowledge, the only other metal (M) except beryllium, that is known to form stable, isolatable compounds of composition M₄O(RCO₂)₆ is zinc.^[1] The two most significant differences between Zn₄O(RCO₂)₆ and Be₄O(RCO₂)₆ type compounds, are firstly, that the former are soluble in water while the latter are insoluble in any aqueous media including strong acids and bases. In aqueous media Be₄O(RCO₂)₆ species are not accessible, they in fact are prepared usually by solid state reactions and sublimation or at least under phase-transfer conditions [19]. The reason for this markedly different behavior of Zn and Be can be explained by the inability of the small Be²⁺ cation to extend its coordination number to numbers larger than four, unlike the larger and electronically more flexible Zn²⁺. Hence a hydrolysis of the Be species is kinetically hindered, or in fact impossible. Secondly, the Zn–O distance in Zn₄O(RCO₂)₆ is with about 1.9 Å, significantly larger than the corresponding Be–O distance of about 1.6 Å. For that it seems likely that the formation of a [Zn₄O]⁶⁺ at the coordination site S is unfavorable or least kinetically strongly hindered. However, if such a complex eventually is formed it is likely to undergo hydrolysis, dissociation or be present only in low concentration in chemical equilibrium, while the formation of the [Be₄O]⁶⁺/M2/DP2 complex is more likely to be practically irreversible. The same arguments as for zinc apply for any other metal cation of charge 2+, like those of Mg, Mn, Co, Fe or Ni. In this sense the M₄O(RCO₂)₆ type compounds are both by structure and reactivity unique for beryllium among all other elements.

3.3 C.

The composition of a [Be₄O]⁶⁺/M2/DP2 complex (1) suggests that its formation under physiological conditions should be a rather slow, rarely occurring process, since four Be²⁺ cations have to accumulate in the small coordination site S and also since an oxide dianion has to be formed. This very general expectation would meet the fact that CDB has long, and partially very long latency times.

On the other hand its not implausible, that eventually such complexes will form: We suggest that the mechanism could involve a step-wise transport of one single or two Be²⁺ cations via the peptide inside the lysosome to the MHC-II. The M2 peptide with its three donor functions is a plausible chelating ligand for Be²⁺ cations at the position between p4D and the p5L carbonyl oxygen, (or between p7E and one of the former). The Be²⁺-loaded self-peptide can then act as “vector” transporting Be²⁺ to the β26E-, β68E-, and β69E-sites of the DP2. The higher negative charge accumulation at the DP2 (3− as compared to 2− in M2) can lead to a transfer of Be²⁺ from the M2 to the DP2. If more Be²⁺ is available to further peptide molecules inside the lysosome this process can take place again and can lead to an accumulation of possibly up to four Be²⁺ cations at coordination site S.

However, the ireversibility (vide supra) of its formation implies severe consequences for the immunological mechanisms involved. The self peptide, that is distorted by the coordination of [Be₄O]⁶⁺ cannot be removed by CD4-T cells, which could trigger the emergence of CD4-T cells that are adapted to distorted self-peptides, explaining the auto-immuno symptomatics of CBD. Our model in addition explains the enhanced interleukin-2 production of T-cell hybridomas when plexin self-peptides are presented (together with Be), that contain three acidic residues (D, D, E) [23]. In this way the ideal number of six carboxylate ligands is yielded in the complex at site S. This trend is also reflected in the high thermal stability of the Be/plexin A4/DP2/AV22 complex [12].

3.4 D.

However plausible 1 is as a model, the structure factors of 4p4k are inconsistent with it [41], [42]. In particular the proposed conformation of the β69E carboxylate group is in contradiction with the steric requirements for the [Be₄O]⁶⁺ coordination center. This is also reflected in the C ^{β 26 E} –C ^{β 69 E} distance of 9.06 Å, which is about 3 Å larger than expected for the new model (while the other two C–C distances are matching very well). It is known, that the β69E carboxylate group can be engaged in binding the AV22 protein of the CD4 T-cell. Hence, this carboxylate group is pointing towards the AV22 site in 2.

Due to the predictive power of the [Be₄O]⁶⁺ hypotheses, we are inclined to conclude that the in-vitro preparation [12] did not lead to the key complex of CBD, but possibly to some intermediate product. This was already considered by the authors of the study:

“While our data support the conclusion that the cation captured in this crystal is Na ⁺ , our structure here does not rule out the possibility that, in vivo or tissue culture, cations other than Na ⁺ might occupy this position and may even complete the TCR ligand.” [12]

A mere reorientation of the carboxylate group β69E towards the binding site S could lead to a shortening of the C ^{β 26 E} –C ^{β 69 E} distance from 9.06 to 6 Å as it is required for the new model 1.

The presence of beryllium in the complex in principle opens another source of experimental data, namely ⁹Be NMR spectroscopy for a chemical characterization. ⁹Be NMR data on CBD related beryllium-protein complexes are scarce [20], nevertheless the method is an ideal complement to single crystal X-ray diffractometry, since it allows to monitor in-vitro reaction progress in solution. For that we have calculated the isotropic chemical shielding values for the model complex 3 (relative to Be₄O(OAc)₆ in CDCl₃ with δ 9 B e = 0.9 ppm), yielding 0.6, 0.7, 1.4 and 1.0 ppm, respectively. If these four signals can be resolved depends critically on their half-width. The ⁹Be NMR signal half width of Be₄O(OAc)₆ is very small (W _1/2=4 Hz at 20°C) [43], which is unusual for a nucleus with a non-zero nuclear quadrupole moment. The small peak width is explained by the high local symmetry of the ligand spheres around the Be nuclei, which eventually leads to a very small electric field gradient at the nucleus such that the fast quadrupole relaxation mechanism becomes prohibited.

We provide in Section 2.3 details on the calculation of the quadrupole coupling constants for models 3–5 and introduce additional calculations needed to characterize such species in a relaxation study.