Báo cáo khoa học: Influence of inflammation-related changes on conformational characteristics of HLA-B27 subtypes as detected by IR spectroscopy potx

Together with our previous studies of HLA-B27 subtypes complexed with the unmodified self-peptide RRKWRRWHL, these findings support the existence of subtype-specific conformational features

Influence of inflammation-related changes on

conformational characteristics of HLA-B27 subtypes as detected by IR spectroscopy

Heinz Fabian1, Bernhard Loll2, Hans Huser3, Dieter Naumann1, Barbara Uchanska-Ziegler3 and Andreas Ziegler3

1 Robert Koch-Institut, Berlin, Germany

2 Institut fu¨r Chemie und Biochemie, Abteilung Strukturbiochemie, Freie Universita¨t Berlin, Germany

3 Institut fu¨r Immungenetik, Charite´-Universita¨tsmedizin Berlin, Freie Universita¨t Berlin, Germany

Keywords

ankylosing spondylitis; citrullination;

conformational differences; HLA-B27

subtypes; IR spectroscopy

Correspondence

D Naumann, Robert Koch-Institut, P 25,

Nordufer 20,

D-13353 Berlin, Germany

Fax: +49 30 1875 42606

Tel: +49 30 1875 42259

E-mail: naumannd@rki.de

(Received 3 December 2010, revised 8

March 2011, accepted 11 March 2011)

doi:10.1111/j.1742-4658.2011.08097.x

Inflammatory processes are accompanied by the post-translational modifi-cation of certain arginine residues to yield citrulline, and a pH decrease in the affected tissue, which might influence the protonation of histidine resi-dues within proteins We employed isotope-edited IR spectroscopy to investigate whether conformational features of two human major histocom-patibility antigen class I subtypes, HLA-B*2705 and HLA-B*2709, are affected by these changes Both differ only in residue 116 (Asp vs His) within the peptide-binding grooves, but are differentially associated with inflammatory rheumatic disorders Our analyses of the two HLA-B27 sub-types in complex with a modified self-peptide containing a citrulline RRKWURWHL (U = citrulline) revealed that the heavy chain is more flexible in the HLA-B*2705 subtype than in the HLA-B*2709 subtype Together with our previous studies of HLA-B27 subtypes complexed with the unmodified self-peptide RRKWRRWHL, these findings support the existence of subtype-specific conformational features of the heavy chains under physiological conditions, which are undetectable by X-ray crystallog-raphy and exist irrespective of the sequence of the bound peptide and its binding mode They might thus influence antigenic properties of the respec-tive HLA-B27 subtype Furthermore, a decrease in the pH from 7.5 to 5.6 during the analyses had an influence only on HLA-B*2709 complexed with the unmodified self-peptide, where His116 is not contacted by any peptide side chain This permits us to conclude that histidines, and in particular His116, influence the stability of MHC:peptide complexes The conditions prevailing in inflammatory environments in vivo might thus also exert

an impact on selected conformational features of HLA-B27:peptide complexes

Structured digital abstract

l B*27 and VIPR bind by biophysical ( View interaction ).

Abbreviations

AS, ankylosing spondylitis; H⁄ D, hydrogen ⁄ deuterium; HC, heavy chain; HLA, human leukocyte antigen; b 2 m, b2-microglobulin; MHC, major histocompatibility complex; pLMP2, (RRRWRRLTV); pVIPR, (RRKWRRWHL); pVIPR-U5, (RRKWURWHL; U = citrulline);

TIS, (RRLPIFSRL).

Major histocompatibility complex (MHC) class I

mole-cules are cell-surface membrane glycoproteins that

con-sist of a highly polymorphic heavy chain (HC)

noncovalently associated with a light chain, b2

-micro-globulin (b2m), and a peptide derived from intracellular

proteins [1] Recognition of these peptide-loaded MHC

molecules by cellular ligands on effector cells triggers

immune responses [2] For human leukocyte antigen

(HLA) class I molecules, peptides derived from self- or

nonself proteins are usually 8–12 amino acids long and

are accommodated in a binding groove of the molecule

by means of HLA allele-characteristic ‘anchor’ amino

acids This selectivity of MHC molecules towards

cer-tain anchor residues of peptides provides the basis for

HLA subtype-specific immune responses and impacts

on disease associations as described, for example, for

the group of HLA-B27 alleles [3–6]

It is known that the citrullination of proteins, a

post-translational modification, influences immune

responses and inflammatory reactions [7,8] This

modi-fication involves arginine, which is strongly basic,

whereas the resulting citrulline is a neutral amino acid

[9] Citrullination is found within synovial tissue from

patients with reactive arthritis, an HLA-B27-associated

disorder [10] Moreover, fragments derived from

citrul-linated polypeptides are most likely also available for

presentation by MHC antigens in patients with

anky-losing spondylitis (AS), another spondylarthropathy

that is even more strongly associated with HLA-B27

than reactive arthritis [2,11]

Although distinct, both diseases are also

character-ized by inflammatory processes that are, by their very

nature, associated with a decrease in the pH value of

the affected tissues [12], leading to proton

concentra-tions that can be elevated greatly (100- to 200-fold)

The pH decrease is expected to affect primarily

histi-dine residues, due to their sensitivity to relatively small

pH shifts at physiologically relevant values (the pKaof

histidine in proteins is 6.6) [13] It is well established

that protonation of ionizable groups in folded proteins

may contribute to their conformational stability [14]

Among the HLA-B27 subtypes, HLA-B*2705 (in

short, B*2705) is strongly associated with AS, whereas

another, HLA-B*2709 (in short, B*2709) is not [2,3,11]

The proteins encoded by these two alleles differ only by

a micropolymorphism (Asp116 in B*2705 and His116

in B*2709) within the peptide-binding groove formed

by each of the HC [1,3] Detailed functional, structural

and thermodynamic studies of these very closely related

subtypes have been carried out to shed light on the

molecular mechanisms underlying their differential

association with AS [15–26] High-resolution crystal structures of these subtypes complexed with peptides constituting the HLA-B27 repertoire reveal that some peptides, such as the self-peptides TIS (RRLPIFSRL)

or pCatA (IRAAPPPLF), are displayed very similarly

by the two HLA-B27 subtypes [21,26], whereas the viral peptide pLMP2 (RRRWRRLTV) [22] and the self-pep-tide pVIPR (RRKWRRWHL) [18] exhibit drastically different conformations In addition, pVIPR is bound

in a canonical single conformation by B*2709, but in an exceptional dual conformation by the AS-associated B*2705 subtype One of the conformations observed in B*2705:pVIPR is identical to that seen in B*2709, whereas in the other, peptide Arg5 (pArg5) forms a salt bridge to HC residue Asp116, resulting in a noncanoni-cal binding mode of the ligand The dual conformation

of this ligand in B*2705 has also been linked to differ-ential T-cell responses among the two HLA-B27 sub-types [15,18]

Recently, an isotope-edited IR spectroscopic study

of B*2709:pVIPR and B*2705:pVIPR demonstrated that the HC is more flexible in the B*2705 subtype than in the B*2709 subtype at physiological tempera-tures [27] Furthermore, similar conformational differ-ences between the HC of the two subtypes were also found in complexes with the viral peptide pLMP2 and the self-peptide TIS [28] Collectively, these findings reveal the existence of subtype-specific, but peptide sequence- and conformation-independent conforma-tional differences between the two HLA-B27 HC at physiological temperatures, which to date have not been detectable using X-ray crystallography

To approach the situation prevailing in inflamed tis-sue, we have now extended these IR spectroscopic studies to a citrullinated self-peptide, pVIPR-U5 (RRKWURWHL; U = citrulline) and performed experiments at a lower pH value This peptide is pre-sented by the two HLA-B27 molecules in binding modes that differ drastically not only from each other [29], but also from the conformations exhibited by the noncitrullinated version of the peptide [18] Specifi-cally, pVIPR-U5 is displayed by B*2705 in a canonical conformation (Fig 1A), but exhibits a noncanonical binding mode in the B*2709 subtype, where the side chain of citrulline at peptide position 5 (pU5) is embedded within the binding groove and forms a hydrogen bond to His116 of the HC (Fig 1B) The comparative IR spectroscopic analyses described here address the question of whether the previously observed difference in flexibility between the B*2705 and B*2709 HC is also found when a peptide assumes,

because of citrullination, a noncanonical binding mode

in the B*2709 subtype Furthermore, we investigated

whether the micropolymorphism that distinguishes the

two HLA-B27 subtypes exerts an influence on the

sta-bility of the complexes when the pH is lowered to a

value representative of an inflammatory environment

Results

Infrared absorbance spectra of HLA-B27:pVIPR-U5

complexes

The IR spectroscopic behaviour of the B*2709⁄

13C-b2m:pVIPR-U5 and B*2705⁄13C-b2m:pVIPR-U5

complexes was initially studied at pH 7.5 after transfer into D2O-buffer (Fig 2A) The use of13C-labelled b2m for reconstitution with separately expressed unlabelled

HC in the presence of the corresponding peptide greatly reduced overlapping of its amide I band with that of the HC in the spectroscopic analyses The spec-tra of the two HLA-B27:pVIPR-U5 complexes were very similar to those previously determined for com-plexes with the unmodified self-peptides pVIPR [27], TIS [28] or the viral peptide pLMP2 [28], in a given sub-type A feature at 1594 cm)1due to13C-labelled b2m and a dominant broad absorbance corresponding to the

HC centred at 1640 cm)1were observed The under-lying HC-specific band components between 1620 and

1700 cm)1 had previously been assigned to different secondary structure elements of the HC Bands at

 1650 and 1643 cm)1 are primarily due to helical structures which change as a consequence of hydro-gen⁄ deuterium (H ⁄ D) exchange, whereas a strong band component at  1624 cm)1 and weaker band compo-nents between 1693 and 1681 cm)1 are due to the

b sheets of the HC [27]

Because 38% and 23% of the protein backbone of the HC is formed by b sheet and a-helical structures, respectively, spectral components attributed to these structures dominate the IR spectrum The spectral con-tributions of the pVIPR-U5 nonapeptide are ‘buried’ under those of the 276 HC residues The same holds true for the spectral features associated with the His fi Asp exchange in the HC The amino acid side chain of Asp gives rise to a relatively strong absorp-tion band between 1550 and 1585 cm)1, which over-laps with an absorption band due to Glu of similar intensity [30,31] Taking into account the 34 Asp and Glu residues in B*2709, the spectral contribution of one additional Asp in B*2705 is practically negligible This can be expected even more for His, because its side-chain absorption band around 1600 cm)1 is very weak [31] The remaining IR intensity at  1545 cm)1 (amide N–H deformation vibration) 1 h after transfer into D2O buffer, together with the presence of a band

at 3286 cm)1(N–H stretching vibration), shows that

a number of amide NH groups of the HC in the two complexes are protected from H⁄ D exchange The band at 3286 cm)1(amide A) is the best indicator of residual nonexchanged N-H groups, owing to the lack

of other protein absorption in the range 3200–

3400 cm)1 [30] Unfortunately, the very strong water absorption band at  3400 cm)1 (O–H stretching vibration) prevents one from obtaining the amide A band of the HLA-B27:peptide complex in H2O buffer, even when using IR transmission cells of only a few

lm pathlength Thus, the amount of nonexchanged

Fig 1 Structure of the pVIPR-U5 peptide in complex with B*2705

and B*2709 The peptide is depicted from the side of the a2 helix

(not shown) The floor of the peptide-binding groove and the

a1 helix are shown in grey ribbon representation, the

subtype-spe-cific residue 116 (Asp116 or His116) is indicated in green (A) The

pVIPR-U5 peptide is drawn as a purple stick model bound to

B*2705 (B) The pVIPR-U5 peptide is drawn as a yellow stick

model anchored to B*2709 by a hydrogen bond connecting pU5 O7

and His116NE2, as indicated by a dashed red line Oxygen atoms

are shown in red, nitrogen atoms in blue B*2705 binds the peptide

in canonical conformation, while it is presented by B*2709 in a

noncanonical binding mode.

amide protons for the partially exchanged state of the

complex at the beginning of the experiment in D2O

was approximated (Fig S1) by setting the difference in

peak intensity of amide II at 1550 cm)1 between the

spectra of the sample in H2O buffer (0% exchange)

and after thermal denaturation of the complex at

90C (fully deuterated state) to 100% This approach

is only an approximation because: (a) the amide II

bands of different conformations of proteins have

different peak maxima, and (b) IR bands due to amino acid side-chain absorptions (overlapping bands of Glu and Asp) between 1550 and 1585 cm)1 may overlap with residual amide II band features and may change

as a function of conformational changes Keeping this

in mind, the IR data suggest that  50% of the amide protons of the B*2709⁄13C-b2m:pVIPR-U5 complex remained unexchanged 1 h after transfer into D2O buffer

100 mA

20 mA

Wavenumber (cm –1 )

3314 3306

B C

D

E F

Fig 2 HLA-B27 complexes measured in D2O buffer at pH 7.5 (Lower) (A) IR absorbance spectra of B*2709 ⁄ 13 C-b2m:pVIPR-U5 (black trace) and of B*2705 ⁄ 13 C-b2m:pVIPR-U5 (red trace), both measured at 15 C 1 h after transfer into D 2 O buffer The spectra of the two sam-ples were normalized using the tyrosine absorption band at 1514 cm)1as an internal intensity standard (Middle) Differences of IR spectra

of the HLA-B27:peptide complexes, all measured at 15 C 1 h after transfer into D 2 O-buffer (B) B*2709⁄ 13 C-b2m:pVIPR-U5 ) B*2705 ⁄ 13

C-b2m:pVIPR-U5 and (C) B*2709 ⁄ 13 C-b2m:pVIPR ) B*2705 ⁄ 13 C-b2m:pVIPR The IR data of B*2709 ⁄ 13 C-b2m:pVIPR and B*2705⁄ 13

C-b 2 m:pVIPR are from previous work by our group [27] Note that the absorbance scale for the difference spectra (B, C) was expanded by a factor of five compared with the scale of the absorbance spectra (A) (Upper) (D) Second derivatives of the IR spectra of B*2709 ⁄ 13

C-b2m:pVIPR-U5 (black trace) and B*2705 ⁄ 13 C-b2m:pVIPR-U5 (red trace) at 15 C (E) IR difference spectra (B*2709⁄

13 C-b 2 m:pVIPR-U5 ) B*2705 ⁄ 13 C-b 2 m:pVIPR-U5) of the second derivatives at 15 C (black trace) and at 90 C (red trace) (F) IR difference spectra of the second derivatives of experiments with two independent preparations of each HLA-B27:pVIPR-U5 complex (black trace; B*2709; red trace, B*2705), demonstrating the high reproducibility of the experimental data.

Detection of HLA-B27 subtype-dependent

conformational properties

The IR difference spectrum obtained by subtracting

the spectrum of B*2705⁄13C-b2m:pVIPR-U5 from that

of B*2709⁄13C-b2m:pVIPR-U5 at 15C (Fig 2B) is

characterized by spectral features in the amide A,

ami-de I¢, and amiami-de II ⁄ II¢ region The positive IR

differ-ence band at  3292 cm)1 demonstrates the presence

of less H⁄ D-exchanged amide groups in the HC of

B*2709 compared with the B*2705 subtype, which is

supported by the broad positive difference feature at

 1560 cm)1 Because less H⁄ D exchange means less

flexibility of the proteins’ core regions to make them

accessible to the solvent, the IR data demonstrate that

the B*2705 HC is more flexible than the B*2709 HC

The difference feature at  3292 cm)1 (Fig 2B)

accounts for only  2% of the total area under the

amide A band of the B*2709⁄13C-b2m:pVIPR-U5

com-plex (black trace in Fig 2A) relative to a baseline

between 3180 and 3410 cm)1 Taking the estimated

50% of nonexchanged amide protons of the

B*2709⁄13C-b2m:pVIPR-U5 sample as reference, the

IR data indicate that the observed differences in

flexi-bility between the two subtypes might be restricted to

only some residues of their HC

The difference features around 1688 and 1624 cm)1

are due to spectral characteristics associated with

ami-de I¢ band components attributed to the b sheets of

the HC, suggesting fine differences in the

hydrogen-bonding pattern of the b-type structures of the HC in

the two HLA-B27 samples By contrast, no positive or

negative features are observed around 1594 cm)1

(Fig 2B), indicating identical peak positions of the

b-sheet band of 13C-labelled b2m in the two samples In

turn, these characteristics suggest a very similar degree

of H⁄ D exchange of the amide protons of b2m in

B*2705⁄13C-b2m:pVIPR-U5 when compared with that

in B*2709 Remarkably, the spectral differences

observed herein for the two complexes with pVIPR-U5

(Fig 2B) revealed difference bands very similar to

those found previously by us in case of the

pVIPR-complexed B*2709 and B*2705 subtypes (Fig 2C),

including a positive difference feature in the amide A

region and positive and negative features in the

amide I¢ ⁄ II region of the spectrum Moreover, the

IR difference spectroscopic features (B*2709⁄13

C-b2m:pVIPR-U5) B*2705 ⁄13C-b2m:pVIPR-U5) did not

change appreciably between 15 and 55C (data not

shown), indicating that the conformational differences

between the two HLA-B27 complexes persist over this

temperature range, as observed formerly for the two

HLA-B27:pVIPR complexes [27]

To estimate the number and position of individual components under the broad amide I⁄ I¢ band con-tours, we also employed derivative spectroscopy This method allows to visualize fine differences in the posi-tion, intensity and shape of band components in greater detail than by making simple comparisons of the original spectra [31,32] Differences in peak posi-tion and⁄ or intensities of the amide I¢ bands attributed

to the b sheets of the HC are clearly visible by com-paring the second derivatives of the spectra of the two HLA-B27:pVIPR-U5 complexes collected at 15C and

pH 7.5 (black and red traces in Fig 2D) Striking is the difference in peak position of the amide I¢ band due to the high-frequency b-sheet band component at

1691⁄ 93 cm)1 which gives rise to obvious positive and negative features by subtracting the second derivative

IR spectrum of B*2705:pVIPR-U5 from that of B*2709:pVIPR-U5 (Fig 2E) These differences, together with the minor spectral differences of the low-frequency b-sheet component at  1624 cm)1, indicate the presence of fine differences in the relative orienta-tion of the b strands of the HC in the two HLA-B27 samples at physiological temperatures The features between 1635 and 1655 cm)1 (Fig 2E) suggest a more intense and⁄ or slightly shifted amide I¢ band compo-nent in the spectrum of B*2709 compared with that of B*2705 Subtle structural differences between the HC

of the two HLA-B27 subtypes, which cannot be speci-fied at present, are also indicated by the amide A band components at 3306 and 3314 cm)1 of B*2709 and B*2705, respectively (Fig 2D,E)

The high-temperature IR difference spectrum is fea-tureless (red trace in Fig 2E), indicating the loss of all conformational differences between the two HLA-B27 subtypes Moreover, the subtype-dependent spectral differences (black trace in Fig 2E) were much more pronounced than the spectral differences between two independent preparations of the corresponding com-plexes (Fig 2F) This provides evidence that the observed spectral differences at low temperatures are really significant, and demonstrates the high quality of the experimental data (also see [28])

Subtype-dependent conformational properties as deduced from IR spectroscopy in water

IR measurements in D2O provide valuable information

on both the structure and flexibility (H⁄ D exchange)

of a protein Moreover, the different kinetics of H⁄ D exchange may assist in the assignment of absorption bands arising from different secondary structure classes [27,30–33] By contrast, IR experiments in D2O can also complicate the interpretation in the amide I¢

region, because spectral differences in the amide I¢

band contour due to H⁄ D exchange and changes in

secondary structure may overlap The only way to

overcome this complication is to fully exchange the

protein before monitoring conformational changes in

D2O medium Complete H⁄ D exchange is also

achiev-able in the case of the HLA-B27:peptide complexes by

keeping the sample solutions close to the denaturation

temperature before cooling them to low temperature,

but this is always accompanied by irreversible

aggrega-tion of the sample, thus rendering it impossible to

obtain the IR spectrum of a completely exchanged

native HLA-B27:peptide complex (data not shown, see

also [27]) The interpretation of the IR spectra

obtained in H2O medium is not complicated by the

above-mentioned spectral effects due to H⁄ D

exchange We therefore also measured the IR spectra

of the native HLA-B27:peptide complexes in H2O

buf-fer, despite the fact that it is more difficult to obtain

IR spectra in this solvent than in D2O-containing

buf-fer because of the interbuf-fering intense water

deforma-tion band at around 1640 cm)1[30]

For a direct comparison with the measurements in

D2O buffer described previously (Fig 2), the second

derivatives of the absorbance spectra of the four

complexes and their corresponding differences were

calculated Interestingly, the IR spectra obtained in

H2O buffer revealed subtype-specific spectral features

(Fig 3) More importantly, many of these features in

the amide I region resembled those described

previ-ously for the corresponding IR spectra in D2O buffer

(compare Fig 2D,E with Fig 3A,C) Differences in

peak position of the amide I bands at 1692⁄ 94 cm)1

assigned to the high-frequency b-sheet components,

which also give rise to clear positive and negative

fea-tures by subtracting the second derivative IR spectrum

of B*2705:pVIPR-U5 from that of B*2709:pVIPR-U5

(Fig 3C), together with the spectral differences of the

low-frequency b-sheet component at 1625 cm)1 were

observed This corroborates the conclusions derived

from analyses of the spectra in D2O buffer (Fig 2),

that fine differences in the relative orientation of the

strands in the b-sheet structures of the HC in the two

HLA-B27 samples do exist at low temperatures The

components at  1650 cm)1 (Fig 3C) suggest a more

intense and⁄ or slightly shifted amide I band in the

spectrum of B*2709 compared with that of B*2705

Again, the differences between the second derivative

spectra of B*2705:pVIPR-U5 and B*2709:pVIPR-U5

turned out to be very similar to those found for the

two HLA-B27:pVIPR complexes (compare Fig 3C

with Fig 3D) This holds true for the different peak

positions of the high-frequency b-sheet band around

1690 cm)1and the spectral differences between the fea-tures at  1650 and  1626 cm)1 as well Moreover, the subtype-dependent spectral differences were much more pronounced than the spectral differences between two independent preparations of the corresponding complexes (Fig 3E,F), as demonstrated previously for the corresponding IR spectra in D2O buffer (Fig 2F) [27] Altogether, the high degree of similarity between the corresponding IR spectra in D2O and in H2O buffer permits us to conclude that the spectroscopic

1694 1692

Wavenumber (cm –1 )

F E

D C

B

A

Fig 3 HLA-B27 complexes measured in H2O buffer at pH 7.5 (Lower) IR spectra (second derivatives) of B*2709 ⁄ 13 C-b 2 m and B*2705 ⁄ 13

C-b 2 m (black and red traces in each panel, respectively), complexed with (A) pVIPR-U5 or (B) pVIPR at 15 C (Middle) Differ-ences between the second derivative IR spectra of the HLA-B27:peptide complexes all measured at 15 C (C) B*2709 ⁄ 13 C-b2m:pVIPR-U5 ) B*2705 ⁄ 13 C-b2m:pVIPR-U5 and (D) B*2709 ⁄ 13 C-b2m:pVIPR ) B*2705 ⁄ 13 C-b2m:pVIPR (Upper) IR-difference spec-tra of the second derivatives of experiments with two independent preparations of each HLA-B27:peptide complex at 15C (E) com-plexed with pVIPR-U5 (black trace, B*2709; red trace, B*2705) and (F) complexed with pVIPR (black trace, B*2709; red trace, B*2705) The spectra of the samples were normalized by use of the tyrosine absorption band at 1516 cm)1as an internal intensity standard.

features in the amide I⁄ I¢ regions associated with the

polypeptide backbone both indicate subtle

subtype-spe-cific structural differences, rather than being the

conse-quence of minor subtype-specific differences in H⁄ D

exchange of the amide protons in the corresponding

HLA-B27:peptide complexes 1 h after transfer into

D2O buffer These subtype-specific structural

differ-ences might influence temporary local or global

unfolding of the HC, and thus its flexibility

pH-dependent thermal stabilities of

peptide-complexed HLA-B27 subtypes

Having established that HLA-B27 subtype-specific, but

peptide sequence-independent, conformational

differ-ences between the two HC do exist in solution, we next

investigated whether the presence of a hydrogen bond

between the pU5 side chain and His116 of the HC

might impact on the thermal denaturation behaviour

of the HLA-B27:pVIPR-U5 complexes at physiological

pH Following our previous approach [27], we

employed the aromatic ring-stretching vibration

of the tyrosine band at 1514 cm)1 to follow

tempera-ture-induced conformational changes in the HC

The decrease in absorbance at 1592 cm)1with increas-ing temperature was used to monitor denaturation of the secondary structure of b2m in the complexes The various frequency⁄ temperature plots obtained

by monitoring the tyrosine ring vibration (Fig 4A) revealed very similar thermostabilities of the HC in the two complexes at pH 7.5 (Table 1) The thermal dena-turation temperatures estimated from the 13C-labelled

b2m band of the two HLA-B27:pVIPR-U5 complexes were almost identical (Fig 4C), with a weak tendency

to reach higher Tm values than those determined for the HC-specific tyrosine band (Table 1) Moreover, the transition temperature of b2m in the two complexes ( 64 C) was very similar to that of free 13C-labelled

b2m ( 65 C) (Table 1)

By contrast to this finding, distinct thermal denatur-ation temperatures were observed for the two HLA-B27 subtypes complexed with the unmodified peptide pVIPR (Fig 4B,D) The B*2709⁄13C-b2m:pVIPR complex was less thermostable than the B*2705⁄

13C-b2m:pVIPR complex by 4–5C Moreover, a dif-ference in peak position of the tyrosine band between the spectra of B*2709⁄13C-b2m:pVIPR and B*2705⁄

13C-b2m:pVIPR in the temperature range from 15 to

–1 )

–1 )

10 20 30 40 50 60 70 80 90

10 20 30 40 50 60 70 80 90

10 20 30 40 50 60 70 80 90

Temperature (°C)

10 20 30 40 50 60 70 80 90 Temperature (°C)

1514.8 1514.6 1514.4 1514.2 1514.0

1514.8 1514.6 1514.4 1514.2 1514.0

Fig 4 Thermostabilities of HLA-B27:peptide complexes at pH 7.5 All measurements were carried out in D 2 O buffer I and monitored by IR spectroscopy For each plot, the first data point (at 15 C) was obtained 1 h after transfer of the corresponding sample into D 2 O buffer The temperature dependence of the position of the HC-specific tyrosine band at 1514 cm)1is shown for (A) B*2709 ⁄ 13 C-b2m:pVIPR-U5 (d) and B*2705⁄ 13 C-b 2 m:pVIPR-U5 (s), as well as for (B) B*2709 ⁄ 13 C-b 2 m:pVIPR (d) and B*2705 ⁄ 13 C-b 2 m:pVIPR (s) The other panels depict the temperature dependence of the peak intensity of the b2m-specific b-sheet band at 1592 cm)1for (C) B*2709 ⁄ 13 C-b2m:pVIPR-U5 (d) and B*2705 ⁄ 13 C-b2m:pVIPR-U5 (s) as well as for (D) B*2709 ⁄ 13 C-b2m:pVIPR (d) and B*2705 ⁄ 13 C-b2m:pVIPR (s).

50C was observed (Fig 4B), indicating that the

microenvironment of at least some Tyr residues must

differ between the two subtypes [27] In addition, the

gain in stability of B*2709:pVIPR-U5 compared with

B*2709:pVIPR (Table 1) is likely to be a consequence

of an additional peptide–HC interaction (a hydrogen

bond connecting pU5O7and His116NE2) that is present

only in B*2709:pVIPR-U5 This suspected involvement

of His116 in stabilizing the B*2709 subtype complexed

with pVIPR-U5, but not pVIPR, prompted us to study

the effect of lowering the pH such that it approached

that in an inflamed tissue [12]

To this end, we analysed the influence of a pH value of 5.6 on the thermal stability of all four HLA-B27:peptide complexes (Fig 5) The data reveal that both HLA-B27:pVIPR-U5 complexes (Fig 5A,C) exhibited comparable and high thermostabilities with Tmvalues of

 63 C at pH 5.6 (Table 1) By contrast, and as suspected, a strong impact on the thermostability upon lowering the pH to 5.6 was observed for B*2709:pVIPR ( 10 C), but not for B*2705:pVIPR (Table 1) The lack of a comparable pH-induced decrease in thermal stability in case of B*2709:pVIPR-U5 allows to conclude that: (a) it is likely that the hydrogen bond between the

Table 1 Determination of the transition temperatures of HLA-B27:peptide complexes The transition temperatures (T m values in C) were calculated either from the intensity ⁄ temperature plot of the b-sheet band of b 2 m at 1592 cm)1or from the frequency ⁄ temperature changes

of the tyrosine ring vibration of the HC at 1514 cm)1 of the IR spectra of B*2705 ⁄ 13 C-b2m:pVIPR–U5, B*2709 ⁄ 13 C-b2m:pVIPR-U5, B*2709 ⁄ 13 C-b2m:pVIPR and B*2705 ⁄ 13 C-b2m:pVIPR 1 h after transfer into D2O buffer Measurements were performed at pH 7.5 or 5.6 (see Materials and Methods for experimental details) For comparison, the transition temperatures as estimated from IR experiments with free 13 C-labelled b2m are also shown The Tm values are the average of experiments with two or three independent preparations of each sample, with standard deviations of 0.5–1 C.

10 20 30 40 50 60 70 80 90 1513.8

1514.0 1514.2 1514.4 1514.6 1514.8

10 20 30 40 50 60 70 80 90 1513.6

1513.8 1514.0 1514.2 1514.4 1514.6 1514.8

–1 )

–1 )

10 20 30 40 50 60 70 80 90

10 20 30 40 50 60 70 80 90

C

A

D

B

Fig 5 Thermostabilities of the HLA-B27:peptide complexes at pH 5.6 All measurements were carried out in D2O buffer II and monitored

by IR spectroscopy For each plot, the first data point (at 15 C) was obtained 1 h after transfer of the corresponding sample into D 2 O buffer The temperature dependence of the position of the HC-specific tyrosine band at 1514 cm)1is shown for (A) B*2709 ⁄ 13 C-b2m:pVIPR-U5 (d) and B*2705 ⁄ 13 C-b2m:pVIPR-U5 (s), as well as for (B) B*2709 ⁄ 13 C-b2m:pVIPR (d) and B*2705 ⁄ 13 C-b2m:pVIPR (s) The other panels depict the temperature dependence of the peak intensity of the b 2 m-specific b-sheet band at 1594 cm)1for (C) B*2709 ⁄ 13

C-b 2 m:pVIPR-U5 (d) and B*2705 ⁄ 13 C-b2m:pVIPR-U5 (s), as well as for (D) B*2709 ⁄ 13 C-b2m:pVIPR (d) and B*2705 ⁄ 13 C-b2m:pVIPR (s).

pU5 side chain and His116 also exists in solution at

physiological pH; and (b) the specific peptide–protein

interaction involving His116 contributes to the

confor-mational stability of the HLA-B27 complex

Correlation of the results from IR spectroscopy

with X-ray crystallographic data

In an attempt to obtain a structure-based

interpreta-tion for the observed subtype-specific IR spectroscopic

findings, we performed a detailed inspection of the

X-ray structures of the two HLA-B27:pVIPR-U5

sub-types and analysed, in particular, the crystallographic

temperature factors (B factors) of the different

struc-tural domains Because both HLA-B27:pVIPR-U5

subtypes crystallized isomorphously, we can exclude

the possibility that differences between the two

struc-tures are due to different packing of the protein chains

in crystallo A comparison of the X-ray structures

revealed, however, that the structures of the HC and

of b2m of the two subtypes appear nearly

indistin-guishable (Ca root mean square deviation £ 0.6 A˚)

[29] An earlier assessment of the B factors for the

different structural units of B*2709:pVIPR and

B*2705:pVIPR revealed that the peptide-binding

groove exhibits the lowest flexibility, whereas major

parts of the a3 domain and of b2m are more flexible

[27] At the same time, this comparison failed to

pro-vide indications for differences between B*2709 and

B*2705, which might serve to explain the observed

dis-similarities in amide protection between the HC of the

two subtypes In the case of complexes with the

unmodified pVIPR peptide, this might have been due

to the different resolutions at which the two structures

were solved (B*2705 at 1.47 A˚, B*2709 at 2.2 A˚) [18]

Such uncertainties do not exist for the

HLA-B27:pVIPR-U5 subtypes, whose structures had been

determined at high and comparable resolutions of

 1.8 A˚ [29] An inspection of the binding grooves of

B*2705:pVIPR-U5 and B*2709:pVIPR-U5,

colour-coded according to the binding groove flexibility in the

crystalline state at 100 K, revealed no clear indications

for subtype-specific differences In summary, neither

the comparison of the crystallographic temperature

factors nor the detailed comparison of structural

fea-tures provide hints which could help to understand the

IR spectroscopic findings observed in solution at

phys-iological temperatures

Discussion

This study addresses questions that are relevant for

understanding how an inflammatory environment,

such as that observed in reactive arthritis or AS, might affect MHC molecules: (a) Is the HC flexibility

of two minimally distinct HLA-B27 subtypes affected

by citrullination of peptides? and (b) Can the stabil-ity of HLA-B27:peptide complexes be impaired by lowering the pH to levels which prevail in inflamed tissues?

The isotope-edited IR spectroscopic results described here corroborate the previously observed differential flexibility of the two HLA-B27 subtypes To date, we had analysed only peptides (pVIPR, TIS, pLMP2) that are bound to B*2709 in the canonical binding mode, with the middle of the peptide bulging out of the bind-ing groove [18,21,22,27,28] This, however, is not the case with pVIPR-U5, because the pArg5fi pU5 exchange leads to a reorientation of this peptide in the binding groove of the B*2709 subtype, accompanied

by the creation of a novel hydrogen-mediated contact between citrulline and His116 (Fig 1) [29] Therefore, the increased conformational flexibility of the B*2705

HC in comparison to that of B*2709 found also in case of pVIPR-U5 (see IR results, Fig 2) must be regarded as an intrinsic property of the HC of the two subtypes and not as a peptide sequence- or binding mode-related characteristic, particularly because this modified peptide is bound in a single, canonical con-formation by B*2705 Ultimately, the polymorphic HC residue 116 must be responsible for the observed reori-entation of pVIPR-U5 in comparison with the pVIPR peptide (Fig 6) We have already proposed a struc-ture-based explanation to account for effects observed

in conjunction with the Asp116His exchange, suggest-ing that a repositionsuggest-ing of water molecules is responsi-ble for the altered flexibility of the two opposing helical segments of the binding groove [28] IR spec-troscopy cannot be used to localize the regions where the two HC differ, but molecular dynamics simulations

of complexes of HLA-B27 subtypes with pVIPR have suggested an increased flexibility of two opposing heli-cal segments (residues 75–60 and 137–150) of the B*2705 binding groove in comparison with that of B*2709 [27] Corresponding MD simulations of the two HLA-B27 subtypes with the modified peptide pVIPR-U5 have not been carried out, but the high degree of similarity of the subtype-specific spectral dif-ferences for B*2705⁄ B*2709 either with pVIPR or pVIPR-U5 as observed in this study (Figs 2 and 3) argues for a comparable nature of the underlying structural differences

The observed subtype-specific differences in the IR b-sheet spectral features of the HC could not be explained on the basis of their X-ray structures, because all b strands of the B*2705 and B*2709 HC

complexed with pVIPR-U5 overlay perfectly Moreover,

a comparative analysis of the B factors for the

differ-ent structural domains provided no clear indications

for differences between the HC of the two subtypes,

suggesting that these conformational characteristics

are only detectable in solution and thus inaccessible

through X-ray data collection at cryogenic

tempera-tures This might be because of the very limited

infor-mation on protein dynamics that can be obtained at

100 K, or because the X-ray technique is not sensitive

enough to resolve these differences As discussed

pre-viously [28], it seems plausible to assume that changes

in the location of the bound water molecules near the

helical regions may also impact on the conformation

of the b sheet of the peptide-binding groove, which in

turn may cause the spectral changes described Subtle

spectral differences associated with b strands in

proteins, which cannot easily be explained by X-ray

crystallography, are not uncommon For example, we

have shown before that a comparative IR

spectro-scopic analysis of the protein ribonuclease T1 and

some of its variants can also reveal such alterations

[34] The only difference observed by X-ray

crystal-lography in these structures is a string of water

mole-cules between the a-helix and the major b-sheet that

are distinctly located in the wild-type protein and in

the variants As pointed out previously [28], it is

con-ceivable that peptides with a C-terminal basic residue

might lead to an altered binding groove flexibility in the B*2705 subtype because of the formation of salt bridges to Asp116 within the molecule’s F pocket [17,35,36] This indicates that the conclusion reached with regard to the enhanced flexibility of the B*2705

HC may not be valid for those complexes that dis-play a peptide with Arg or Lys at the C-terminus (see also [19,25,37] for further discussions) These considerations are irrelevant for the B*2709 subtype, however, because peptides with basic C-termini bind only very rarely to this subtype in vivo [4,16]

How distinct dynamic characteristics of the two HLA-B27 subtypes impact on their function is cur-rently unknown We have previously argued that the interaction of these molecules with receptors on effec-tor cells might be altered in dependence on HC flexibil-ity [28] However, in the absence of thorough analyses

of dynamic properties concerning entire assemblies of MHC class I molecules and receptors on T cells or natural killer cells, it is currently not clear whether interactions of the binding partners are indeed influ-enced Analysis of a T-cell receptor footprint on an HLA-A2:peptide complex by NMR spectroscopy [38]

is a first step towards understanding this intricate issue, although the flexibility of the MHC molecule was not investigated in this study

In addition to the general subtype-specific conforma-tional features discussed above, our experimental

find-Fig 6 Conserved interactions between the displayed peptide and amino acid residues residing on the a1- or a2 helix All structures are superimposed and the view is towards the carboxylate of the C-terminal p9 residue Only the peptide segments from p5 to p9 are shown, and side chains are omitted with the exception of pArg5 (pVIPR) and pU5 (pVIPR-U5) HC residues Asp77 and Trp147, which are involved in conserved interactions with peptide residues, are shown as grey sticks Hydrogen-bonding interactions of the peptide residues p8 (main chain carbonyl) to Trp147 (indole NE atom) and of p9 (main chain amide) to Asp77 (carboxylate OD1 atom) are depicted with red dashed lines (A) B*2705:pVIPR in canonical conformation (green) and in noncanonical conformation (orange) Only the latter conformation allows the peptide to anchor to the HC by a salt bridge from pArg5 to Asp116 The conformation of pVIPR in B*2709 (brown) is indistinguishable from the canonical binding mode of this peptide in B*2705 (B) B*2705 in complex with pVIPR-U5 is shown in purple and B*2709 with pVIPR-U5 in yellow In the latter complex, a hydrogen bond is formed between His116 and pU5 The pU5 side chain points to different directions in the two subtypes Despite differences in peptide sequences and conformations, the formation of highly conserved hydrogen bonds from the a1- and a2 helices to the peptide main chain atoms is still permitted in all four structures.