1

2005-01-00847B

Supplementary Information

Supplementary Methods

Protein expression and purification: SH3 from the -subunit of bovine phosphatidylinositol-3´-kinase (PI3-SH3) was expressed in Escherichia coli (strain HMS174 or BL21) as a fusion protein with glutathione S-transferase, purified, and freeze-dried as described previously1,2. Uniformly 15N-labelled PI3-SH3 was produced by growing cells in M9 minimal media containing 15NH4Cl (Martek Biosciences Corporation) as the sole nitrogen source. The recombinant protein used in this study consists of 84 residues from the PI3-SH3 sequence, plus a two-residue (GS) extension at the N-terminus.SDS/PAGE and MS showed that the protein was pure andthe molecular weight was that expected from the sequence(experimental 9638.7 ± 0.1 Da; theoretical 9638.6 Da).

Preparation of PI3-SH3 amyloid fibrils: Amyloid fibrils were first generated by incubating 0.5 mM of freshly prepared PI3-SH3 in H2O at pH 2.0 for 21 days at 35 ºC. Fibrils for the experiments described in this paper were then produced by incubating 0.5 mM of freshly prepared PI3-SH3 in H2O at pH 2.0 following addition of 2% (v/v) of a solution of pre-formed fibrils for 7 days at 35 ºC. Immediately after addition of the pre-formed fibrils, the solution was sonicated for 2 min to fragment the added fibrils and create additional polymerization surfaces. Second derivative analysis of the infrared amide I band corresponding to seeded and unseeded fibril preparations shows an overall similarity in the number and positioning of the various vibrational modes in both samples (Fig. S1). A detailed examination of the FTIR spectra, however, reveals some important differences. The component at 1660 cm-1, assigned to disordered regions, turns and loops, is substantially decreased in the seeded when compared to the unseeded fibrils, suggesting a decrease in the overall population of these conformational elements, probably contributed by less ordered assemblies. Further, the characteristic bands corresponding to -structure present in amyloid fibrils (1610-1630 cm-1) aremore prominent and better defined in seeded fibril preparations. Taken together, these observations suggest a decrease in the heterogeneity of the seeded sample versus the unseeded fibril preparation, consistent with the disappearance of less structured assemblies or non-fibrillar aggregates. Thus, seeded growth was used throughout this investigation to ensure an amyloid fibril population as homogeneous as possible.

H/D exchange of PI3-SH3 amyloid fibrils: Amyloid fibrils were ultracentrifuged using an Optima TLX ultracentrifuge (Beckman) using either a TLA 120.2 or a MLA 130 rotor (Beckman) at 90,000 rpm for 1 h and 30 min at 25 ºC. The supernatant was removed and the pellet was subsequently diluted into deuterated buffer at pH* 1.63, mixed by means of a pipette tip, and incubated at 28 ºC. The volume in which the amyloid pellet was resuspended was the same as the volume of supernatant removed to ensure that the concentration of fibrils was kept constant. During the time the fibrils were allowed to exchange, some condensation occurred within the sample container. To minimise this effect, approximately every five days the fibril suspension was centrifuged (Eppendorf, Centrifuge 5415D) for 10 s and then homogenised by pipetting up and down ca. 30 times. In those experiments in which additional shearing was applied during the exchange reaction, the fibril suspension was pipetted ca. 100 times (B) or sonicated for 2 min (C). The same treatment was applied three times on the 1st, 2nd, and 5th days and once on the 12th, 19th, 20th, 28th, 33rd, 42nd, and 49th days of exposure to exchange conditions. For all the H/D experiments, 100 µL aliquots (for NMR analysis) or 20 µL aliquots(for MS analysis)were removed at specific times.Aliquots for NMR were withdrawn after 4 and 15 days while aliquots for MS were taken at a variety of lengths of time ranging from 0 to 49 days. Each of the aliquots was then subjected to ultracentrifugation using either a TLA 100 or a TLA 120.1 rotor (Beckman) at 90,000 rpm for 1h and 30 min at 4 ºC. The supernatant was immediately removedand the pellet was freeze-dried and kept at -80 °C until analysis.

Amyloid fibril dissolution:Amyloid fibrils are highly insoluble in water so analysis of their deuterium content cannot be carried out directly by NMR and ESI-MS techniques. Thus, amyloid fibrils were first diluted into 95% DMSO-d6/5% H2O,a process that results in complete dissolution of the fibrils into monomers within seconds.No significant differences in mass or in NMR chemical shifts (Fig. S2) were observed between freshly prepared protein samples and samples of protein recovered from the fibrils after being solubilised in 95% DMSO/5% H2O.

A denaturation experiment was carried out to further assess the homogeneity of the fibril preparation. We exposed the protein pellets resulting from centrifugation of the fibril samples to increasing concentrations of DMSO and determined the concentration of 14N PI3-SH3 by adding the resulting solution to a known quantity of lyophilised15N PI3-SH3 and analysing the data using a Q-tof mass spectrometer. This experiment is in fact technically rather complex since there is a balance between both thermodynamic and kinetic factors that underlie the denaturation process. In order to minimise effects associated with slow kinetics of fibril dissolution that we have explored previously4, we exposed the aggregates to DMSO/H2O denaturing buffer for 12 hours. In accord with a homogeneous preparation of fibrils, at DMSO concentrations higher than 70% we observed that virtually complete dissolution of the aggregates occurred. In addition, despite the complexities of slow dissolution kinetics, at DMSO concentrations equal to, or lower than, 50%, the majority of the sample remained as an insoluble pellet.

Preservation of H/D exchange information:The use of 95% DMSO-d6/5% D2O as the solvent to solubilise the fibrils is based on the fact that it preserves the deuterium content of the protein molecules as amide hydrogen exchange with solvent is very slow5. Only the very labile side chain and chain termini hydrogens will exchange within the time of the NMR and MS experiments. The following control experiments were carried out (i) a 1H-15N HSQC spectrumof fully protonated SH3 fibrils6was recorded immediately after dissolution in the solubilising buffer at pH* 5.3 (1.5 min) and showed that all side chain protons exchange within the time of measurement (5.5 min to optimise the NMR parameters plus 22.3 min to record the 1H-15N HSQC spectrum). In addition, the volumes of most of the backbone amide cross-peaks remain unchanged, confirming that hydrogen exchange in the solubilising solvent is very slow, as reported for other proteins5. However, a significant decrease in volume was observed for the resonances of Ser 2, Asp 15, Asp 30, Gly 37 and Asp 70; data for these residues were therefore not included in the present analysis. (ii) The ESI-MS spectra of fully protonated SH3 fibrils acquired 5 min after dissolution at pH* 4.2 show that the mass of the protein had increased by 65 ± 3.4 Da; this number is close to the theoretical number of protons expected to exchange rapidly, 72 (i.e., from side chains and the chain termini)7.

NMR spectroscopy and assignment: Spectra of PI3-SH3 were recorded at 25 °C on a 500 MHz (Bruker DRX 500) or an 800 MHz spectrometer (Bruker Avance 800) both equipped with triple-axis gradient triple-resonance probes (1H/13C/15N). The sample conditions were 1 mM protein in 95% DMSO-d6 (Euriso-top), 5% H2O at pH* 4.9. The pH was adjusted with dichloroacetic acid-d3 (Aldrich) as described previously5. The field-frequency lock was referenced to the signal from DMSO-d6. Triple resonance experiments, CBCANH, CBCA(CO)NH were recorded for assignment of the PI3-SH3 resonances in DMSO. The assignments were confirmed by analysing the 3D 15N TOCSY-HSQC (m= 65 ms) and 3D 15N NOESY-HSQC (m= 200 ms). 1H chemical shifts were referenced with respect to the 2.5 ppm methyl signal of DMSO-d6. 15N and 13C chemical shifts were indirectly referenced using the frequency ratio N/H=0.10132905 and C/H =0.25144952, respectively. Data were processed and analysed using the programs NMRPipe8 and NMRView9.

H/D exchange analysed by NMR: Amyloid fibril samples exchanged and treated as described above were dissolved in 450 µL of 95% DMSO-d6, 5% D2O (Euriso-top) at pH* 5.3 under an IR lamp to prevent moisture being taken up by the hygroscopic DMSO. After dissolution and mixing for 1.5 min, the sample solutions were immediately transferred into an NMR tube and 1H-15N HSQC spectra recorded at 25 ºC. The same times for NMR parameter adjustment (5.5 min) and spectrum acquisition (22.3 min) were used for all measurements.

Although 95% DMSO-d6/5% D2O at pH*5.3 was used as the solvent to solubilise the fibrils because it preserves the deuterium content acquired during the exchange reaction, some exchange of amide protons does occur. In order to eliminate the effect of this behaviour on the experimental measurement of peak volumes, we use the relative peak volumes VD2O/VH2O; VD2O and VH2O correspond to the peak volumesin the 1H-15N HSQC spectra of the fibril samples incubated and not incubated in D2O, respectively. The ratio VD2O/VH2O after a given time in the solubilising solvent can be expressed as:

VD2O(t)/VH2O(t) = (VD2O(0) exp(-kex-st))/(VH2O(0) exp(-kex-st)) = VD2O(0)/VH2O(0)

where kex-s is the exchange rate in the solubilising solvent and VD2O(0) and VH2O(0) are the volumes at zero time when no exchange with the solubilising solvent has occurred. Since the same mixing times and dead times for parameter adjustment and spectrum acquisition were used, VD2O(t)/VH2O(t) is equal to VD2O(0)/VH2O(0) where exchange with the DMSO based buffer has not taken place. The relative total protein concentration of the solution was determined from integration of the non-exchangeable hydrogens in the methyl region of a 1D proton NMR spectrum recorded directly prior to each individual 1H-15N HSQC spectrum.

H/D exchange analysed by MS: Mass spectrometry was carried out on a Platform II mass spectrometer (Micromass) fitted with a nano-ESI source. Samples were electrosprayed from gold-coated glass capillaries prepared in-house as described previously10, with an applied capillary voltage of 1.7 kV and a cone voltage of 70 V. Amyloid fibril samples exposed to H/D exchange and processed as described above were dissolved in 35 µL of 95% DMSO-d6/5% D2O at pH* 4.2 (0.25 mM)under an IR lamp. After dissolution, the sample solution was immediately introduced into the mass spectrometer. The same dead time for parameter adjustment (5 min) was used for all measurements. Mass spectra of the samples in the DMSO/D2O buffer were also recorded after approximately 5 h and after approximately 10 days. Over time the peak corresponding to the Ppd species converted fully into the one corresponding to the Pfd species. The spectra were analysed using MassLynx 3.1 (Micromass). The peak widths in the fibril exchange samples are comparable to those in spectra of solutions in which the protonated monomeric PI3-SH3 protein exchanged directly with 95% DMSO-d6/5% D2O solvent, indicating that the peaks do not result from overlap of multiple peaks of species with different degrees of exchange.

The theoretical number of exchangeable hydrogens in the PI3-SH3 monomer was calculated by taking into consideration all labile hydrogens in the side chains and in the main chain. These hydrogens are those that are covalently bound to nitrogen, oxygen or sulphur atoms but not to carbon atoms. All of the side chain atoms are regarded as being uncharged, regardless of the pH, as the net charge of the molecule is accounted for in the calculations of the mass during data analysis. Consequently, for the PI3-SH3 monomer, there are 85 labile hydrogens from the main chain amide groups (allowing for the three proline residues and the additional hydrogens at the chain termini). Addition of the 72 hydrogens from the side chains, gives the total number of exchangeable hydrogens as 157.

The increase in mass of the Pfd species is smaller than that corresponding to the total number of exchangeable sites (157) because the most labile deuterons (i.e., those from side chains and the chain termini) exchange rapidly with the hydrogens of atmospheric water as a result of the hygroscopic nature of DMSO and/or during the ionization process itself. In support of this conclusion, we observe no increase in mass of the Pfd species when the protein is left for extended periods of time in the 95% DMSO-d6/5% D2O buffer.

H/D exchange of monomeric PI3-SH3: When protonated PI3-SH3 molecules are left to exchange in deuterated buffer at pH*1.6 and 22 °C, the overall exchange rate measured by ESI-MS is very similar to the theoretical exchange rate for a random coil calculated using HXPRO and based on the pH, temperature, residue and N-side neighbouring residue (available as SPHERE at Under these conditions, PI3-SH3 molecules exchange all of their labile hydrogens within just a few minutes of being dissolved.

Determination of the protein concentration in the supernatant: The concentration of protein in the supernatant was estimated by diluting aliquots 11 times with H2O at pH 2.0. The resulting solutions were mixed with a solution of 15N labelled PI3-SH3 of known concentration and analysed using a Q-tof mass spectrometer (Micromass). The molecular weights of the two isotopically distinct species are sufficiently different for their peaks to be fully resolved in the mass spectra. The experiments were performed in positive ion mode with a capillary voltage of 2.5 kV and a cone voltage of 50 V. The samples were infused with a 50 µL syringe at a flow rate of 10 µL/min. The source and desolvation temperatures were maintained at 80 °C and 150 °C, respectively. The flow rates of the nebulising gas and drying gas were set at 40 L/h and 450 L/h, respectively. 100 spectra with mass range from 100 to 1600 Da were acquired in each case. Minimal smoothing was applied to the combined spectra for analysis using MassLynx 3.1 (Micromass). Since the labelled and unlabelled species produce peaks of similar width, the concentration of the protein in the supernatant was estimated by measuring the relative peak heights of the +8, +9, and +10 charge states of the 14N and 15N peaks (Fig. S3). Theresults show that although there is some variation in the concentration that ranges from16 ± 8 µM to 40 ± 10 µM in the different samples, there is always a significant quantity of soluble protein in the solution containing the fibrils.

Electron Microscopy: Samples were applied to Fomvar-coated copper grids (200 and 400 mesh), negatively stained with 2% uranyl acetate (w/v), and viewed in a JEOL JEM1010 or a Philips CM100 transmission electron microscope, both operating at 80 kV. Before EM analysis, each sample was diluted 20- or 40-fold.

Supplementary Discussion

Models of H/D exchange data: To simulate the process of hydrogen exchange from molecules in the fibrils, we assimilated generic models of amyloid fibril formation/fragmentation11 and protein H/D exchange12. Eqns. S1a and S1b show the basic features of the model used for simulating amyloid fibril formation and fragmentation. Eqn. S1a describes the model for fibril growth which consists of a bi-molecular association event governed by a second order rate constant kon(i), and a uni-molecular dissociation event governed by a first order rate constant koff(i) (Ci refers to the molar concentration of species of degree i). Eqn. S1b describes the model for the amyloid fragmentation event in which a fibril is broken into several resultant fibrils of shorter lengths (i = j+k), the process being governed by a first order rate constant ks(i).

kon(i)

Ci + C1 ⇌ Ci+1 [S1a]

koff(i)

ks(i)

Ci  Cj + Ck [S1b]

Eqns. S2a andS2b describe the model for H/D exchange. Eqn. S2a depicts the process of proton exchange for a molecule of type j13 existing within an amyloid fibril of degree i. The molecule of type j undergoes a reversible transition between a non-exchangeable state, (Ci)(j,H), and an exchangeable state, (Ci)(j*,H), governed by first order rate constants k (j) and k(j*), respectively. Once in the exchangeable form, the molecule of class j exchanges all14 of its amide protons to form a fully deuterated species, (Ci)(j,D), governed by a single first order rate constant kex(j). The observed rate constant for proton exchange for a molecule of type j within the fibril, kobs(j), is related to the observed rate constant for the process of proton exchange for the monomer, kobs(M), derived on the basis of the steady state approximation by a unit less protection factor P(j) that is specific for a particular type of molecule within the fibril (Eqn. S2b). The concentration of fully deuterated molecules of a particular class j at any time is then given by Eqn. S2c.

k(j) kex(j)

(Ci)(j,H) ⇌ (Ci)(j*,H)  (Ci)(j,D)[S2a]

k(j*)

kobs(j) = kobs (M) / P(j)[S2b]

(Ci)(j,D) (t) = (Ci)(j) tot [1  exp(kobs(j) t)][S2c]

As an ensemble of amyloid fibrils exist as a distribution of differently sized species, we assume two possible types of behaviourinvolving either (i) a static distribution in which amyloid fibril formation proceeds irreversibly or (ii) a dynamic distribution in which amyloid fibril formation proceeds reversibly. In modelling the data we sought to explain two observations (i) that the time course for H/D exchange (Fig. 2b) required,as a minimum, the sum of two exponential rate equations (Eqn. S2c) for an adequate description of the kinetics15, and (ii) that only two distinct isotopic populations of PI3-SH3, the partially deuterated, Ppd, and the fully deuterated, Pfd, are observed experimentally in the ESI-MS spectrum, implying that exchange within an individual molecule is essentially a co-operative two-state process.

Static Distribution: If amyloid fibril formation proceeds irreversibly, then the end product will be static with regard to time. In this case, for bi-exponential kinetics to be observed for hydrogen exchange there have to be at least two classes of molecules within the polymer distribution that exchange at distinct rates kobs(1) and kobs (2). If each fibril exists as a symmetric bundle of protofilaments, as EM experiments suggest16,then the required difference in exchange behaviour could result from a situation where protein molecules located at the ends of an amyloid fibril exchange at a different rate from protein molecules locatedin the interior of the fibril (Fig. S4a). Our data suggest that this model is unlikely for two reasons. Firstly, the mass of Pfd is such that the conformation of molecules at the fibril ends or, in any other accessible location, should allow complete isotope exchange while the mass of Ppd is such that the conformation of the rest of the molecules should generate almost complete protection from exchange. Secondly, after 49 days at least 40% of the molecules in each of the samples studied herewere found to be deuterated, requiring the deuterated solvent to penetrate along the fibril and cause complete deuteration of molecules in almost half the fibril length, while leaving the other molecules completely unaffected.Alternatively, fibrils formed from PI3-SH3 could exist as a distribution of structures with different widths as well as lengths17.In this case, diversity in the exchange behaviour could arise from the existence of fibrils with different lateral morphologies such that distinct environments exist for different molecules as a result of the packing of protofilaments (Fig. S4b). This model is also highly unlikely to be the origin of the differential exchange observed here as the conformation of the molecules within one protofilament unit would have to be such that complete exchange could occur whilst in any other alternative lateral morphology virtually no exchange of any of the amide hydrogensin a molecule could take place.