The present study builds on prior work to examine the sources of variability in the efficiency of prion-like seeding in mutant SOD1 mice. From the effort we describe here, prior work in our laboratories [15, 17], and data described by Bedhendi [18], it is clear that first passage G85R seeds efficiently seed G85R mice, and that first passage L126Z seeds efficiently seed L126Z mice (Fig. 9a). By contrast, mice expressing WT-, G37R-, or G93A-SOD1 variants were partially or fully resistant to isologous seeding (Fig. 9a). Notably, mice that express human G85R-SOD1 fused to YFP were found to be susceptible to seeding with a wide range of seeding preparations, including tissue homogenates from paralyzed G37R and G93A mice and preparations of recombinant SOD1 fibrilized in vitro (Fig. 9b). In general, heterologous seeding of G85R-SOD1:YFP mice was less efficient than isologous seeding, with seed preparations that contained the G37R or H46R variant being among the least efficient. We further observed that the seeds that arose in G85R-SOD1:YFP mice after seeding by preparations from paralyzed G37R mice appeared to acquire attributes that led to longer incubation times upon second passage. Collectively, these findings show that sequence variation and seed source have profound effects on the prion-like activity of misfolded mutant SOD1.
In assessing the significance of animals that failed to develop paralysis or pathology as a result of intraspinal or sciatic nerve seeding, we must consider the possibility of experimentation error. It is important to note that all of the injections performed in this study were conducted by only two operators that had extensively trained to perform the injections. Although we cannot rule out the possibility of occasional injection error, we have no indication that operator error would explain the data in cases where few or no animals in a cohort developed paralysis. Thus, we conclude that there is considerable variation in the susceptibility of different lines of mutant SOD1 mice to seeding, and that the sequence of the misfolded SOD1 in the seeds can influence prion-like transmissibility to SOD1 transgenic hosts.
Role of structural stability in the susceptibility of SOD1 variants to seeding
A key distinguishing feature of transgenic SOD1 mice that were vulnerable to seeding versus those that were more resistant could be related to the structural stability of the SOD1 variant expressed by the host. Mice that express WT, G37R, or G93A human SOD1 show high steady-state levels of soluble protein prior to the onset of paralysis, with the accumulation of misfolded SOD1 aggregates occurring as paralysis develops [41, 43, 44]. By contrast, the steady-state levels of soluble mutant SOD1 in G85R mice are relatively low in proportion to transgene mRNA levels [43]. Both the L126Z and QV103Z truncation variants are unstable and show very low steady state levels until they begin to aggregate as paralysis develops [20, 21]. Thus, for the G85R, L126Z, and QV103Z mutants, their susceptibility to isologous, or heterologous, seeding may be linked to poor structural stability.
In WT SOD1 a structurally important intramolecular disulfide bond links Cys 57 to Cys 146 [45]. In studies of SOD1 aggregation and fibrillization in vitro, the presence of the normal intramolecular disulfide bond appears to be a key factor in modulating aggregation [46]. In mutant SOD1 mice that develop paralysis due to over-expression, the misfolded mutant protein that accumulates as pathological inclusions develop lacks the normal intramolecular disulfide bond [43, 44]. Previous analysis of the same G85R mice used here for seeding studies found that all of the mutant protein present in spinal cord of pre-symptomatic mice lacks this critical structural stabilizer [43]. Similarly, the L126Z and QV103Z variants would not be able to produce a normal intramolecular disulfide bond. It is important to note, however, that in GurWT and Gur1G93A SOD1 mice, 8–14% of the over-expressed protein present in spinal cords of pre-symptomatic mice also lacks the normal disulfide bond [43, 44]. Thus, whether the G85R, L126Z, and QV103Z mice are susceptible to seeding solely because the expressed mutant lacks an intramolecular disulfide bond is unclear.
Another important post-translation modification in the structural stability of SOD1 is the binding of Cu and Zn [47]. Similar to WT SOD1, the G93A or G37R variants, are capable of binding Cu and Zn with high affinity [3, 23, 26]. The G85R variant has been shown to have a weak affinity for Cu and Zn [48, 49] and the enzyme appears to be inactive in vivo [50]. Crystal structures of the G85R SOD1 variant demonstrate disorder in the Zn binding and electrostatic loop elements of SOD1 (amino acids 50–83 and 121–142, respectively) [48]. The L126Z and experimental QV103Z truncation mutants have not been crystallized, but neither of these variants possess intact electrostatic loop elements. Whether the L126Z truncation mutant can bind Cu or Zn is unknown, but the experimental QV103Z variant that we studied was engineered to remove the critical Cu-binding sites in SOD1 [21]. It is notable, however, that the majority of over-expressed WT or G93A SOD1 present in the spinal cords of Gur1 G93A or Gur WT mice is inactive due to insufficient acquisition of Cu [43]. Thus, the ability of WT and G93A SOD1 to bind Cu does not seem to fully explain the low susceptibility of mice over-expressing these proteins to prion-like seeding if poor Cu binding were the only structural feature involved.
The binding of Zn by nascent SOD1 appears to be critical to achieving a native conformation [51, 52]. Analysis of partially purified homodimeric SOD1 from spinal cords of the GurWT, G37R-Line 29, and Gur1G93A mice found the over-expressed protein appeared have no deficiency in the level of bound Zn; measured to be between 2.5 and 3 atoms per dimer [53]. Thus, it is possible that these variants resist seeding because bound Zn limits prion-like propagation.
Overall, our data indicate that WT SOD1 and WT-like variants are less likely to exist in conformations that are susceptible to prion-like conformational templating as compared to mutants that are more destabilized. Whether any specific structure-stabilizing feature is critical in mediating susceptibility to seeding is unclear. What is clear is that mice expressing the more destabilized variants, G85R, L126Z, and QV103Z were far more susceptible to prion-like seeding.
Do ALS-associated mutations in SOD1 imprint strain-like attributes that influence prion-like activity?
Based on our observation that we can induce and monitor the spread of G85R-SOD1:YFP aggregation by seeding into the sciatic nerve of this model, we had begun to link the progressive spread of weakness in SOD1-ALS to the prion-like spread of misfolded conformations [16]. The question that arises is whether there may be conformational attributes that are enciphered by the primary sequence of mutant SOD1 that, by some manner, modulate the ability of misfolded SOD1 to spread within the CNS. It is important to note that in the transgenic mouse models that over-express mutant SOD1, the mutation-specific differences in disease duration that is seen in humans are not recapitulated. For example, mice that express the G93A and G37R variants at similar levels exhibit disease onset and duration that is comparable; 4–6 months to onset and 3–4 weeks duration [23, 26]. Presumably the high-level expression of mutant SOD1 throughout the CNS that is required to produce disease within the lifespan of the animal masks the mutation-specific attribute that causes very slow progression in persons with the G37R mutation. Here, we have asked whether the G37R, or H46R, mutations could impart some specific conformational information to the misfolded SOD1 that accumulates in these animals that would lead to distinct behavior in prion-like transmission studies.
Our newborn injection paradigm is one way to assess the seeding competency of different SOD1 variants. In heterologous seeding in G85R-SOD1:YFP mice by spinal homogenates from paralyzed G93A or L126Z mice (avg. durations of 2.4 and 3.8 years in humans, respectively [35]), the efficiency of seeding at first passage was variable. Second passage of G93A seeds through G85R-SOD1:YFP mice, however, produced a consistently earlier onset of paralysis (avg. 2.8 months post-injection). Our study of L126Z seeds in G85R-SOD1:YFP mice is more limited, but we did similarly observe consistent acceleration of paralysis by second-passage L126Z seeds in a small cohort of animals (Additional file 2: Fig. S15; data from [17]). Thus, for two mutants associated with rapidly progressing disease we observe relative ease of first passage with accelerated and highly efficient second passage seeding.
For the H46R and G37R variants, associated with slowly progressing ALS in humans, we observed poor seeding activity in first passage, and with G37R-derived seeds we osbserved variable and protracted incubation periods. These data are consistent with the idea that mutations associated with slowly progressing disease may be less efficient in prion-like seeding.
An important consideration in interpreting these results is the potential impact of the host G85R SOD1 variant on any conformational change propagated by exposure to seeds from mice expressing any other variant. In PrP prions, sequence differences between the primary sequence of PrP in seeds and the sequence of PrP in the host can modify strain attributes [54]. Moreover, single amino acid differences of the PrP sequence in the seed and host can create a species barrier that lowers transmission efficiency [55]. To date, G85R-SOD1 mice have shown efficient seeding with spinal cord preparations containing misfolded A4V, G85R, D90A, G93A, and L126 or G127 truncation mutants [15, 17,18,19]. We observed that fibrilization treatments of 7 different recombinant SOD1 proteins could produce seeds that induce early paralysis in G85R-SOD1:YFP mice. These data indicate that G85R-SOD1 is broadly susceptible to heterologous conformational templating. Indeed, seeds from G37R mice were more effective in G85R SOD1 mice than G37R SOD1 mice. Thus, the poor seeding of G85R-SOD1:YFP mice by spinal tissues from H46R mice and the long incubation periods produced by G37R seeds may indicate that these variants produce strains of misfolded SOD1 that are less effective as seeds to propagate misfolded conformations.
Is prion-like spreading a common feature in SOD1 ALS?
Our observation that mice expressing the G93A variant are not particularly receptive to seeding would appear to run counter to the notion that prion-like mechanisms of spreading apply to all fALS variants. As indicated above, the G93A mutation is associated with relatively rapidly progressing disease [35, 56], and thus the expectation would be that mice expressing G93A-SOD1 should be highly permissive to seeding, particularly isologous seeding. We have previously reported that the commonly used Gur1-G93A mice did not exhibit accelerated paralysis after intraspinal seeding [15]. To assess whether the level of G93A expression may have been a factor, we tested two additional lines of G93A mice. The Thy-1 G93A mice were unresponsive to seeding; however, with the VLE-G93A mice we began to see evidence that seeding could accelerate the appearance of paralysis albeit at a lower frequency than we would have expected for isologous seeding. As noted above, in the Gur1-G93A mice, a substantial portion of the over-expressed protein lacks Cu and a normal disulfide bond [43]. The absence of these modifications appears to be related to over-expression; increasing Cu loading in Gur1G93A mice dramatically mitigates its misfolding and toxicity [57]. The Cu loading and disulfide status of G93A SOD1 in spinal cords of VLE-G93A mice has not been examined, but the lower expression of mutant protein would be expected to result in lower levels of incompletely modified mutant SOD1.
The absence of seeding efficacy in the Thy-1 G93A mice could potentially be related to the pattern of transgene expression for this model. It is also possible that the level of G93A SOD1 expression in Thy-1 G93A mice is just below threshold to sustain the propagation of misfolded conformations. Additional studies of seeding efficacy in these two lines of G93A mice could provide insight into the basis of vulnerability. At the time of writing, we are not able to explain why the Gur1-G93A or Thy-1 G93A mice did not respond to seeding or why the VLE-G93A showed a modest response. If vulnerability to seeding is by some means related to the folding state of the SOD1 protein in the host, we would have expected the Gur1-G93A mice to be more susceptible to seeding due to a higher level of incompletely modified mutant protein. At face value, the poor seeding of G93A mice could foster skepticism over the role of prion-like spread in the pathogenesis of SOD1-linked fALS; however, there are mitigating factors that merit discussion.
A key variable in our experiments is the level of bioactive misfolded SOD1 seeds in the spinal homogenates we have used for seeding. Previous studies of mutant SOD1 mice that develop paralysis have demonstrated that the levels of misfolded mutant SOD1 in the spinal cords rise steadily and reach maximum when mice are paralyzed [20, 41, 44]. We originally had assumed that at endstage, regardless of age, the levels of misfolded SOD1 in any given animal for a given mutant would be at maximal levels. For all of our mouse to mouse propagation studies, we used 10% homogenates of spinal cord that had been clarified by low speed centrifugation. In our isologous seeding studies of G85R-SOD1:YFP, G85R, and L126Z mice, the levels of the seeds in homogenates from paralyzed transgenic mice were clearly sufficient to induce accelerated paralysis with high efficiency. For the WT-SOD1:YFP, WT-, G93A-, and G37R-SOD1 expressing mice, which did not respond to isologous seeding, it is possible that injection of a higher amount of seeds would have induced disease. Notably, our G93A and G37R seed preparations were effective, to varying degrees, in heterologous seeding of G85R-SOD1:YFP mice. Moreover, recombinant fibrils of WT human SOD1 were highly effective in seeding G85R-SOD1:YFP mice, with no activity in WT or WT-SOD1:YFP mice. Thus, it is clear that the lines of mice expressing WT and WT-like SOD1 mutants (G37R and G93A) are less susceptible to prion-like seeding with preparations that effectively induce early disease in mice expressing the unstable G85R mutant. Whether the lower susceptibility of WT and WT-like variants to seeding could be overcome by injection of higher amounts of misfolded SOD1 seeds will require additional investigation.
An additional consideration in assessing the susceptibility of WT and WT-like SOD1 mutants to prion-like propagation is the timing of exposure to the seed. We have relied heavily on a paradigm in which the seeds were injected into newborn mice, primarily as a means to facilitate wide-spread dissemination of the seeds. The earliest age to paralysis in mice expressing high levels of mutant SOD1 that has been reported is 3–4 months of age [23, 26, 29], suggesting that early in life the toxicity of the mutant SOD1 is suppressed by yet to be defined protective factors. It is possible that mice expressing WT-, G93A-, or G37R-SOD1 do not respond to seeding in the newborn injection paradigm because a combination of protective factors and the natural propensity of these mutants to fold into a more native-like enzyme, suppresses the establishment of a self-sustaining propagative process. It is also possible that the capacity of WT or WT-like SOD1 variants to fold and mature could change with aging as proteostatic protective factors decline. In humans, where disease usually occurs late in life, the WT-like variants of SOD1 could become more prone to adopt conformations that are susceptible to prion-like seeding.
Although our studies, and others [18, 19], demonstrate that mutant SOD1 has the potential to acquire prion-like capabilities, there are aspects of the mouse models that provided sources of misfolded SOD1 seeds that should be considered. In all the mouse or rat models used to produce inoculum, mutant SOD1 is highly expressed throughout the CNS. In this scenario, the development of disease would not necessarily depend on the prion-like spread and therefore there would be little selective pressure towards the accumulation of misfolded SOD1 proteins that can efficiently propagate between cells. It is possible that the absence of such selective pressure leads to the accumulation of mixtures of misfolded SOD1 strains in the transgenic mice that were used to produce the seeds. Any given individual animal could potentially harbor a unique mixture of strains and as such could account for some of the variability we observe in first-passage seeding experiments.
Intriguingly, we observed multiple examples in which seeded mice reached relatively old ages without developing symptoms despite significant inclusion pathology burden. The most striking example was in the cohort of G85R-SOD1:YFP mice seeded with spinal homogenates from paralyzed H46R rats. We also observed high levels of inclusion pathology in aged bigenic mice created by crossing GurWT mice with WT-SOD1:YFP mice. These intriguing findings raise the possibility that some conformations of misfolded SOD1 may be non-toxic. Additional studies are required to understand whether strain variants of misfolded mutant SOD1 may exist that propagate between cells more slowly, or produce less toxic aggregates, and whether such strain variations explain aspects of human SOD1 linked ALS.
Role of WT-SOD1 in sporadic ALS
The role of prion-like propagation in sporadic ALS, and the identity of the misfolded propagon, remains to be determined. Multiple studies have examined whether a misfolded form of WT-SOD1 could be propagating throughout the neuraxis in sporadic ALS patients (reviewed in [58]). Recombinant WT-SOD1 that has been fibrilized in vitro clearly possesses high seeding activity when injected into G85R-SOD1:YFP mice [17], but these same preparations showed no activity in the WT-SOD1:YFP or GurWT mice. From our experiments in which mutant SOD1 was co-expressed with WT-SOD1:YFP, it was clear that with sustained exposure it is possible to induce WT-SOD1:YFP to produce inclusions. Collectively, our data indicate that although it is possible to create a seeding-competent conformer of WT-SOD1, it does not appear to transmit to WT substrates efficiently in short-term seeding paradigms. It is possible that propagation of misfolded conformations by WT-SOD1 is inhibited by the high propensity of WT-SOD1 to acquire a stable native conformation. Whether other types of conformational changes in WT-SOD1 that could contribute to ALS, and possibly propagate in a prion-like fashion as suggested by other studies [10, 58, 59], has not been ruled out by our study. However, we have not observed WT SOD1 to be able to support propagation of a disease-causing misfolded SOD1 conformation thus far.
Potential role of amyloidogenic segments in prion-like propagation of misfolded SOD1
Although the G85R-SOD1:YFP, or untagged G85R, mice appear to be highly susceptible to seeding, we observed that seeds prepared from mice that over-express murine SOD1 with the G85R mutation were not effective in either host. Incompatibility between the misfolded seed and the host was also noted when QV103Z mice were seeded with spinal homogenates from L126Z mice, but not vice versa. These seemingly disparate outcomes may provide clues to initial sites of interaction between the propagating seed and the naïve host protein. Ivanova and colleagues identified four amino acid segments in SOD1 that were highly prone to aggregation [60]. The four segments were distributed across the protein at residues 14–21, 30–38, 101–107, and 147–153. The ability of misfolded QV103Z to propagate to itself or L126Z-SOD1 indicates that if these segments are involved in prion-like propagation, then the segments 14–21 and 30–38 must be involved. For the L126Z mutant, segment 101–107 could also be important because although L126Z could self-seed, it could not cross seed to QV103Z. It is also noteworthy that mice expressing the G37R variant are largely resistant to seeding, and similarly, a synthetic peptide of the amyloidogenic segment 30–38 with the G37R mutation was much slower to aggregate in vitro [60]. Segment 30–38 is also of interest because of significant sequence divergence between mouse and human SOD1 within this segment; the human sequence is KVWGSIKGL and the mouse sequence is VLSGQITGL. Our studies clearly show that misfolded mouse G85R-SOD1 seeds are not transmissible to mice expressing human G85R-SOD1. In segments 14–21 and 101–107 there is only one amino acid difference between mouse and human SOD1, and in segment 147–153 there are no differences. One of the differences between human and mouse sequence in segment 30–38 is a W to S change at position 32, which has been implicated in prion-like propagation of misfolded SOD1 [14, 36, 61, 62]. Collectively, these data suggest the amyloidogenic segment between residues 30–38 could be an important modulator of prion-like propagation of misfolded SOD1.