6.5 Discussion
10.4.3 Residue-specific ssNMR analysis
Two-dimensional (13C,13C) ssNMR spectra of D2Q15K2, GK2Q38K2, and GK2Q54K2 fib-rillar aggregates reveal striking similarities indicative of a common structural organization (Fig. 10.4). Most prominently, in all constructs and across different types of spectra se-lecting for more rigid parts of the sample, two major populations of glutamine residues are apparent whose resonance signals fall into distinct spectral regions (in the following denoted as populations 1 and 2). A third glutamine population is seen in (13C,13C) corre-lation spectra with longer mixing times, likely due to enhanced mobility in these residues.
A rather high spectral linewidth of around 2 ppm in 13C and 3 ppm in 15N dimensions indicates some structural heterogeneity in all samples. Apparently, glutamine-rich se-quences can assume slightly different molecular conformations within fibrillar aggregates, as also suggested by the distribution of fibril diameters observed in EM for GK2Q38K2
and GK2Q54K2 constructs. The observations of rather large spectral linewidths and of distinct sets of NMR signals from residues of the same type are consistent with recent ss-NMR data on fibrils from asparagine- and glutamine-rich peptides and proteins [307, 315].
Notably, crosspeaks corresponding to population 1 and 2 glutamines are also found in spectra of lyophilized polyglutamine samples recorded after addition of water or buffer solution, even though only oligomers can be detected in such samples by EM (Appendix Fig. F.2). This might indicate that the local organization leading to the occurrence of two distinct glutamine populations forms already in early stages of aggregation.
The two main populations of glutamine residues appear in ssNMR spectra with ap-proximately equal intensities in all constructs, suggesting a distribution close to 1:1. Reso-nances of all nuclei of population 1 glutamines are consistently shifted to higher frequencies with respect to population 2 (Table 10.1). This effect has also been observed for a glu-tamine residue in fibrils formed by a asparagine- and gluglu-tamine-rich peptide from the yeast prion Sup35 [315]. Compared to average Cα and Cβ chemical shift values for glutamine, both populations exhibit a strongly negative secondary chemical shift of about -5 ppm (Table 10.1; see ref. [158] and http://www.bmrb.wisc.edu/). This clearly indicates that both groups of glutamine residues participate in β-sheet secondary structure [153] and
10.4 Results 143
Figure 10.4: (a, b) (13C,13C) spin diffusion correlation spectra (150 ms mixing time) and (c, d) intraresidue NCACB correlation spectra of D2Q15K2 (a, c) and GK2Q54K2 (b, d). Bracketed numbers in assignments refer to the different glutamine populations (see text). Corresponding carbonyl regions are depicted in Appendix Figure F.3.
Gln population or
D2Q15K2 residue N CO CA CB CG CD CE NE2
1 118.0 175.9 55.9 34.2 33.9 178.5 107.9
2 115.4 174.0 54.0 31.6 29.8 177.4 104.4
3 174.0 55.7 29.9 33.9 180.2
D2 177.6 54.6 40.9 180.0
K18 118.6 174.8 55.1 33.5 24.7 29.5 42.1
Table 10.1:List of average chemical shifts (in ppm) observed for the three glutamine populations in D2Q15K2, GK2Q38K2, and GK2Q54K2 samples. Populations 1 and 2 constituteβ-sheets, pop-ulation 3 corresponds to residues in turn regions or unstructured molecules. For completeness, assignments for residues Asp2 and Lys18 of D2Q15K2are also given.
144 10 |Structural characterization of polyglutamine fibrils that polyglutamine fibrillar aggregates are largely composed of β-sheets, consistent with their cross-β X-ray diffraction pattern.
On the other hand, Cα and Cβ chemical shifts of glutamine population 3 visible in long mixing time (13C,13C) spectra are close to random-coil values. These residues could constitute turns or unstructured regions within the fibrils, or they could occur in unstruc-tured monomers with enhanced, but limited mobility. The appearance of loose protein in some electron micrographs is consistent with this assumption. However, the presence of highly mobile individual monomers as seen in fibril preparations ofα-synuclein [20] can be excluded for D2Q15K2, where INEPT-based spectra probing molecules with solution state-like dynamics [113] only reveal minuscule lysine sidechain signals (Appendix Fig. F.4). In contrast, clear signal in INEPT spectra can be observed for GK2Q38K2 and GK2Q54K2. However, compared to cross-polarization (CP) and direct excitation spectra, lysine signals are strongly enhanced with respect to glutamine signals, indicating that the INEPT signal predominantly arises from mobile termini of molecules that are otherwise rigid. These data suggest that, in the longer constructs, the lysine-containing N- and C-termini are not part of the fibril core, but protrude from it.
The presence of distinct glutamine populations with different chemical shifts raises the question whether these populations occur within a fibril monomer unit. Sequential NCOCA (Ni,Ci−1) correlation spectra on the polyglutamine constructs exhibit spectral intensity predominantly at the same positions as seen in intraresidue NCACB (Ni,Ci) correlation spectra (Fig. 10.5 a-c), suggesting that glutamine residues of the two major β-sheet populations are mostly flanked by residues of the same population within the amino acid sequence. However, some peaks do occur that correspond to correlations between the two populations in the longer constructs. Sequential (13C,13C) correlation spectroscopy [125] with inherently greater resolution and sensitivity confirms the existence of sequential correlations between the two mainβ-sheet populations 1 and 2 in GK2Q38K2
(Fig. 10.5 d) and GK2Q54K2. This means that, at least in the longer polyglutamine constructs, glutamine residues of different chemical shift populations can occur within the same monomer. They appear to be clustered, however, in stretches of several residues of one population, as seen from sequential (Ni,Ci−1) correlation spectra.
Spectral resolution, uniform isotope labeling, and the dominant presence of only one residue type preclude residue-specific sequential resonance assignments in the longer poly-glutamine constructs. More insight into the distribution of poly-glutamine residues with differ-ent shifts can be expected from the selectively labeled D2Q15K2 construct (Section 10.3.1)
10.4 Results 145
Figure 10.5: Interresidue correlation spectra.(a)–(c)Interresidue (Ni,Ci−1) NCOCA correlation spectra of (a) D2Q15K2, (b) GK2Q38K2(recorded by Henrike Heise), and (c) GK2Q54K2. Grey lines in (a) indicate chemical shifts of Lys18 N and Asp2 Cαimportant for assignment; red labels mark their most likely correlation partners. Red assignments in (b, c) indicate sequential correlations between different glutamine populations. (d) Section representing Cα–Cβ correlations of a se-quential (13C,13C) spin diffusion spectrum of GK2Q38K2recorded under weak coupling conditions [125] with 150 ms mixing time. Sequential correlations between different glutamine populations are indicated.
146 10 |Structural characterization of polyglutamine fibrils in which sequential correlations from only three residue pairs can be expected. The labeled residues Asp2 and Lys18 can be identified unambiguously in ssNMR spectra based on their chemical shifts, although Asp2 resonances are usually weak (Fig. 10.4 b). The main inter-residue cross-correlations in an (Ni,Ci−1) correlation spectrum of D2Q15K2 (Fig. 10.5 a) then indicate that Gln17 belongs to glutamine population 1 and Gln3, Gln9 and Gln10 to population 2. However, shoulders of the main peaks within the observed range of Lys18 amide nitrogen and Asp2 Cα chemical shifts (grey lines in Fig. 10.5 a) show that the re-spective other possibilities cannot be ruled out completely. In fact, the NCOCA spectrum in Figure 10.5 a could in principle even be explained by two groups of D2Q15K2 molecules composed entirely of one population of glutamines, since the chemical shifts of the labeled residues cannot be correlated across gaps in the labeling pattern. These data indicate that glutamine populations 1 and 2 do not necessarily correspond to specific positions within the peptide sequence, although preferences at different residue positions seem to exist.
It is important to note that the interresidue correlation corresponding to neighboring glutamines of population 2 in the D2Q15K2 NCOCA spectrum (labeled Q(2)N-CA in Fig. 10.5 a) is outside of the chemical shift ranges of both the Lys18 amide nitrogen and the Asp2 Cα nucleus. Thus, to account for this peak, the Gln9-Gln10 residue pair has to be invoked. This means that, at least in a substantial part of the sample, two consecutive central glutamines exhibit chemical shifts corresponding to an extended β-sheet conformation (see also below),i.e. they cannot form a bend or turn.