The prion-like CTD seems to underlie the aggregation propensity of TDP-43 and most mutations linked to familial ALS (fALS) are located within this domain [
38,
39]. However, scattered reports indicate that the NTD may play a pivotal role in regulating the assembly state of TDP-43. Single-molecule measurements coupled to bulk biophysical techniques indicated that TDP-43 converts into amyloid-like fibrils following two distinct pathways. In one pathway, the CTDs of different molecules assemble directly [
34]. In a second, parallel, pathway, fibrillation is induced by the NTD following a two-step process in which first the NTD domains dock, then the CTDs lock the molecules into an amyloid-like aggregate [
34]. This is consistent with a previous finding that the first 10 residues of the TDP-43 sequence are crucial for both the folding and misfolding of the protein [
31]. Indeed, by removing such a segment, TDP-43 appears to no longer be able to aggregate in the cytosol, even if the cytoplasmic accumulation of the protein is enhanced by removing the NLS sequence. However, other authors suggested that the NTD-driven homo-polymerization might exert a protective role against pathological aggregation. For example, it was reported that NTD dimerization or tetramerization via disulfide bridges may prevent aggregation of TDP-43 and is necessary for the splicing activity of the protein [
12]. Other authors reported that TDP-43 oligomerizes in the nucleus via the NTD [
11]. This may separate spatially the CTDs of different TDP-43 molecules, therefore antagonizing pathological aggregation. Alterations of this equilibrium lead to the accumulation of TDP-43 in the cytoplasm, where the protein aggregates irreversibly, via the CTD [
11].
Be it protective or detrimental, the NTD-controlled multimerization appears to play a major role in the aggregation properties of TDP-43 and, consequently, the identification of possible alternative conformational ensembles populated either transiently or permanently by this domain can help understand the behavior of the bulk protein. Based upon the entire body of experimental evidence presented in this manuscript, we can build a model underpinning the different steps of the folding pathway of NTD (
Figure 7). This model is able to describe a series of conformational states populated transiently by NTD during its folding. The unfolded state exists in a pre-equilibrium between molecules having all the proline residues in
trans (U
t) and a subset of molecules, possessing one or more specific X-Pro peptide bonds in the
cis configuration (U
c). Both these unfolded states convert, within the dead time of our stopped-flow experiments (6.1 ms), into an ensemble of collapsed states, which again can possess all proline residues in
trans (CS
t), or one or more X-Pro peptide bonds in
cis (CS
c). This process is associated with the
rate constant. We surmised this fast step from the difference between the fluorescence emission observed at the beginning of the refolding process and that of the unfolded state. Albeit not yet folded, the CS state possesses a buried indole moiety of Trp68. The CS can then convert into the folded state following two parallel pathways. In the first pathway, CS converts into an intermediate state (I), in a process associated with the
rate constant. This I state, again, can have all prolines in
trans (I
t) or one or more X-Pro peptide bond in
cis (I
c). The I
t state is on the pathway of folding and can convert into the folded state, in a process associated with the
rate constant. The on-pathway nature of the I state was supported by the observation that
was 0.62 ± 0.06 s
−1, a value smaller than
, equal to 3.28 ± 0.30 s
−1. This is compatible only with an on-pathway model [
40]. At very low urea concentrations, the I state is relatively stable and possesses energy similar to that of the CS ensemble. This is shown by the fact that the chevron plot for
and
can be analyzed with a two-state model and does not exhibit downward curvatures at low urea concentration (
Figure 5f). As urea concentration increases, I becomes more unstable than CS and this leads to a second, parallel, folding pathway, in which CS
t undergoes a direct transition that leads to the formation of the folded state (F), with no need of intermediate states accumulating during the process. The occurrence of this pathway was demonstrated by the trend observed in
Figure 5f between denaturant concentrations of 0.67 and 3.70 M, i.e., between the two
values for the chevron plots of
and
(with
). Within this range of concentrations,
was increasing upon increasing urea concentration, indicating denaturation of the I state, but
was instead detectable, because folding was still favorable. This means that, under these conditions, NTD underwent folding even though I was less stable than CS and, therefore, folding proceeded directly from CS
t to F
t, while I was no longer accumulated. It is also possible that I still forms under these conditions (CS
t → I
t →F
t), in a folding process where it has an energy value intermediate between that of CS
t and the transition state for folding, therefore remaining undetected.
In either pathway (CS
t →F
t or I
t →F
t), formation of the folded state can occur only if all X-Pro peptide bonds are in the correct
trans configuration. We observed that a small subset of protein molecules, possessing one or more specific X-Pro peptide bond(s) in the wrong
cis configuration, fold only after proline isomerism, associated with the
rate constant. Last, the F state is able to dimerize, with a rate constant (
) significantly lower than that of refolding (
), indicating that NTD dimerization follows its folding. Dimerization gives rise to a head-to-tail homodimer (F
tF
t), as assessed by our FRET measurements. The head-to-tail arrangement we identified for the dimer is not only in agreement with other models previously reported [
11,
13] but it is also compatible with the idea of a NTD-driven polymerization of TDP-43. Indeed, a head-to-tail dimer is inherently prone to further polymerization [
11], whereas a head-to-head (or tail-to-tail) dimer has no docking sites free and exposed to the solvent and needs to undergo at least one further misfolding step prior to becoming prone to aggregation. Indeed, we found that the folded state of NTD is able to self-assemble into higher molecular weight assemblies (
).
Of note, the folded dimer is prone to undergo structural rearrangements and populate a native-like state (F
t*), as confirmed by our finding that small amounts of urea can distort the twisting of the β-sheets without inducing full denaturation and that small modifications to the pH alter the chemical environment surrounding the tryptophan side chain. Interestingly, the F state of NTD can form tetramers via disulphide bridging, as found in our experiments and by other authors [
12]. When isolated via gel filtration, the tetramer appeared less stable than the dimer, as we found
values of 47.2 °C (
Figure S7) and 52.8 °C for tetramer and dimer, respectively. While it is presently difficult to explain the destabilization observed for the tetramer, these findings lend further support to the idea of a structurally susceptible folded state of NTD, whose oligomerization may trigger further assembly.