TRIOBP-1 Protein Aggregation Exists in Both Major Depressive Disorder and Schizophrenia, and Can Occur through Two Distinct Regions of the Protein

The presence of proteinopathy, the accumulation of specific proteins as aggregates in neurons, is an emerging aspect of the pathology of schizophrenia and other major mental illnesses. Among the initial proteins implicated in forming such aggregates in these conditions is Trio and F-actin Binding Protein isoform 1 (TRIOBP-1), a ubiquitously expressed protein involved in the stabilization of the actin cytoskeleton. Here we investigate the insolubility of TRIOBP-1, as an indicator of aggregation, in brain samples from 25 schizophrenia patients, 25 major depressive disorder patients and 50 control individuals (anterior cingulate cortex, BA23). Strikingly, insoluble TRIOBP-1 is considerably more prevalent in both of these conditions than in controls, further implicating TRIOBP-1 aggregation in schizophrenia and indicating a role in major depressive disorder. These results were only seen using a high stringency insolubility assay (previously used to study DISC1 and other proteins), but not a lower stringency assay that would be expected to also detect functional, actin-bound TRIOBP-1. Previously, we have also determined that a region of 25 amino acids in the center of this protein is critical for its ability to form aggregates. Here we attempt to refine this further, through the expression of various truncated mutant TRIOBP-1 vectors in neuroblastoma cells and examining their aggregation. In this way, it was possible to narrow down the aggregation-critical region of TRIOBP-1 to just 8 amino acids (333–340 of the 652 amino acid-long TRIOBP-1). Surprisingly our results suggested that a second section of TRIOBP-1 is also capable of independently inducing aggregation: the optionally expressed 59 amino acids at the extreme N-terminus of the protein. As a result, the 597 amino acid long version of TRIOBP-1 (also referred to as “Tara” or “TAP68”) has reduced potential to form aggregates. The presence of insoluble TRIOBP-1 in brain samples from patients, combined with insight into the mechanism of aggregation of TRIOBP-1 and generation of an aggregation-resistant mutant TRIOBP-1 that lacks both these regions, will be of significant use in further investigating the mechanism and consequences of TRIOBP-1 aggregation in major mental illness.


Introduction
Schizophrenia, like many chronic mental illnesses, has a highly complex pathophysiology, epitomized by the large number of genetic variants that contribute to it [1], making it difficult to identify specific protein targets for diagnosis or therapy. Therefore, in a aggregation: a refined version of this central aggregation motif, but also the N-terminal untranslated region. The domain structure of TRIOBP-1, shown in a one-dimensional format, as determined previously [11,14,15,23]. In coiled-coil domains, predicted coils are shown in darker shades, while loops are shown in paler shades. Amino acid (AA) numbers refer to the 652 AA form of TRIOBP-1, which includes the optionally translated region. (B) A section of the primary structure of TRIOBP-1, with the amino acids previously implicated in protein aggregation [15] shown in blue. Below each image are shown the regions of this protein contained in each of the major TRIOBP-1 plasmids used in this paper. More details of these plasmids can be found in Supplementary Table S1.

Insoluble TRIOBP-1 Is Found in the Brains of Patients with Major Depressive Disorder and Schizophrenia, Indicating Its Aggregation in Both Conditions
In order to confirm whether aggregated TRIOBP is present in the brains of patients with major mental illness, we optimized two previously published protocols [2,25] for isolating the insoluble protein fraction of brain tissue (Supplementary Figures S1 and S2) which will be referred to as the high and low stringency purification techniques. The low stringency protocol, which was developed to enhance sensitivity, purifies both proteins known to form unfolded aggregates (as confirmed by looking at DISC1 and Tau in the brains of established transgenic rodent lines [26,27]) as well as actin (Supplementary Figure S2). Notably, actin is known to be found in the insoluble protein fraction of various assays as a result  [11,14,15,23]. In coiled-coil domains, predicted coils are shown in darker shades, while loops are shown in paler shades. Amino acid (AA) numbers refer to the 652 AA form of TRIOBP-1, which includes the optionally translated region. (B) A section of the primary structure of TRIOBP-1, with the amino acids previously implicated in protein aggregation [15] shown in blue. Below each image are shown the regions of this protein contained in each of the major TRIOBP-1 plasmids used in this paper. More details of these plasmids can be found in Supplementary Table S1.

Insoluble TRIOBP-1 Is Found in the Brains of Patients with Major Depressive Disorder and Schizophrenia, Indicating Its Aggregation in Both Conditions
In order to confirm whether aggregated TRIOBP is present in the brains of patients with major mental illness, we optimized two previously published protocols [2,25] for isolating the insoluble protein fraction of brain tissue (Supplementary Figures S1 and S2) which will be referred to as the high and low stringency purification techniques. The low stringency protocol, which was developed to enhance sensitivity, purifies both proteins known to form unfolded aggregates (as confirmed by looking at DISC1 and Tau in the brains of established transgenic rodent lines [26,27]) as well as actin (Supplementary Figure S2). Notably, actin is known to be found in the insoluble protein fraction of various assays as a result of its existence as highly multimeric fibrous F-actin, thus being an example of a protein that can be characterized as being "insoluble" despite existing in a correctly folded state.
In contrast, the high stringency protocol, which was developed for enhanced specificity, seemingly detects only unfolded insoluble protein, detecting aggregated DISC1 and to an extent aggregated Tau, but not actin (Supplementary Figure S2). These protocols were used to purify the high and low stringency insoluble pellets of post mortem brain samples (anterior cingulate cortex, BA23) from 25 schizophrenia patients, 25 patients with major depressive disorder and from 50 control individuals (see Supplementary Table S1 for demographic information). Western blotting of the ensuing fractions, in a blinded manner, was then used to determine the presence of insoluble TRIOBP in these samples (Figure 2A).
The high stringency insoluble protein fraction revealed TRIOBP signal in a significant proportion of samples; however, the strength of the signal was too low to accurately quantify ( Figure 2A). Each sample was therefore rated as either containing or not containing a distinguishable 72 kDa TRIOBP band, corresponding to full length TRIOBP-1 which is implicated as aggregating in schizophrenia [9]. This band was present in the high stringency insoluble fraction of 12/25 schizophrenia samples and 15/25 major depression samples (48% and 60% respectively), but only 2/50 control samples (4%, Figure 2B), implying a substantial enrichment of aggregated TRIOBP-1 in the brains of patients with either of these conditions.
In the low stringency insoluble protein pellets, four major TRIOBP bands could be easily detected (Figure 2A). The longest of these was over 140 kDa in size and is consistent with the long TRIOBP-5 or 6 splice variants, which incorporate almost the entire length of TRIOBP-1 [12][13][14]. The 72 kDa full length TRIOBP-1 band was also seen in addition to two smaller species of 60 kDa and 55 kDa, which would either correspond to breakdown products or shorter splice variants of TRIOBP. Levels of the TRIOBP-1 species in the low stringency insoluble fraction were decreased (by approximately 50%) in the brain samples from both schizophrenia and major depression patients, when compared to the control group ( Figure 2C). The TRIOBP-5/6 band did not consistently vary with diagnosis. There was no correlation between TRIOBP-1 signal and post mortem interval or age of the patient at death, and only a weak correlation with the pH of the brain sample (Pearson correlation coefficient 0.249, p = 0.012, Supplementary Figure S3A). Ethnicity may also have an effect; however, low sample numbers make it difficult to be certain (Supplementary Figure S3B).
The result that low stringency-derived insoluble TRIOBP-1 is decreased in mental illness contrasts with the increase in high stringency-derived aggregated TRIOBP-1 in the same samples. Indeed, brain samples displaying TRIOBP-1 in their high stringency insoluble pellet, regardless of diagnosis, showed on average less than 50% of the amount of the same protein band in their low stringency insoluble fraction, with an inverse correlation existing between the two (point-biserial correlation coefficient: −0.349, p = 3.7 × 10 4 , Figure 2D). Therefore, while TRIOBP generally exists assembled into a high molecular weight complex, presumably associated with F-actin as part of its physiological role in actin modulation [11,14,15], it appears that in some individuals with mental illness, a fraction of this TRIOBP-1 is converted to an unfolded aggregated state (detectable using the high stringency protocol).
There was no association between the level of any TRIOBP species detectable in the total brain homogenate with mental illness, nor between levels of total and insoluble TRIOBP, as detected using either protocol (Supplementary Figure S4A-D). Notably, however, the main 72 kDa form of TRIOBP-1 was seen as a doublet in the brain homogenate, despite only one such band being seen in the insoluble fractions. The ratio between the level of expression of these two species was significantly associated with major depression (Supplementary Figure S4E). A potential explanation for this is that it represents TRIOBP-1 protein that either has or is lacking the 59 AA optional N-terminal domain. of its existence as highly multimeric fibrous F-actin, thus being an example of a protein that can be characterized as being "insoluble" despite existing in a correctly folded state.  Comparison of the total level of 72 kDa TRIOBP-1 in low stringency insoluble protein fractions, between brain samples that either did or did not contain 72 kDa TRIOBP-1 in their high stringency insoluble fraction. * p < 0.05, ** p < 0.01, n.s.: not significant.

Refining the Aggregation Domain of TRIOBP-1, Using C-Terminal Constructs
It was previously shown that a 25 amino acid stretch from AA 324-348 is required for aggregation in cultured cells [15]. This stretch consists of a presumed unfolded loop within the CCC domain of TRIOBP-1 and is rich in charged amino acids. To elucidate the exact amino acids responsible for its aggregation propensity, plasmids encoding Nterminally truncated human TRIOBP-1 ( Figure 3A) were expressed in SH-SY5Y cells. This neuroblastoma cell line was selected for all immunofluorescent microscopy experiments, as it can be transfected with an efficiency that is plausible for these experiments (≈10%, based on immunofluorescence microscopy), but also displays neuron-like characteristics. For confirmation of plasmids by Western blotting, HEK293 cells were instead used, due to their significantly higher transfection rate and therefore higher level of protein expression.
This previously described aggregation-critical region consists of two clusters of charged amino acids, flanking a central uncharged cluster ( Figure 1B). Novel constructs were therefore generated, encoding the C-terminal half of TRIOBP-1, but only part of this 25 amino acid stretch. TRIOBP-1 333-652 was seen to readily aggregate and was indistinguishable from TRIOBP-1 324-652, indicating that the first cluster of charged amino acids is not required for aggregation ( Figure 3D). This fragment was unstable, however, and expressed poorly in cells. In contrast, TRIOBP-1 341-652 and 343-652 both showed visibly reduced levels of aggregation, with aggregates visible in some cells but not others ( Figure 3D). This indicates that both the uncharged region from AA 333-340 and the cluster of amino acids beginning at AA 341 are likely to be involved in the aggregation process.
To confirm this, a blinded experiment was performed in which SH-SY5Y cells were transfected with TRIOBP-1 324-652, 341-652, 343-652 and 349-652, and their aggregation status characterized and quantified. Transfected cells were then classified as either "clearly aggregating" (all visible TRIOBP-1 is present in clusters of 0.5 nm or larger), "clearly not aggregating" (TRIOBP-1 is diffuse in the cytoplasm) or "potentially aggregating" (either TRIOBP-1 is forming clusters that may or may not be aggregates, or only a small proportion of the TRIOBP-1 present resembles aggregates). TRIOBP-1 333-652 was not quantified due to a low expression level (not affected by proteasomal inhibition, Supplementary Figure S5) and very low number of surviving transfected cells, which together are indicative of protein toxicity. There was a significant effect of transfection status on aggregation (ANOVA, F > F crit , p < 0.001 for each classification). Compared to TRIOBP-1 324-652, in which obvious TRIOBP-1 aggregation occurred in approximately 70% of cells, there were essentially no fully aggregating cells seen with either TRIOBP-1 341-652 or 343-652 ( Figure 3E, p tukey < 0.001 for each). There was, however, a significant reduction in cells showing partial aggregation between TRIOBP-1 341-652 (approximately 35%) and TRIOBP-1 349-652 (less than 3%, p tukey = 0.001). Given that the expression patterns of TRIOBP-1 324-652 and TRIOBP-1 333-652 were identical, this indicates that the uncharged region from AA 333-340 is highly involved in aggregation, while the charged AA from 341-345 is also likely to be involved. amino acids responsible for its aggregation propensity, plasmids encoding N-terminally truncated human TRIOBP-1 ( Figure 3A) were expressed in SH-SY5Y cells. This neuroblastoma cell line was selected for all immunofluorescent microscopy experiments, as it can be transfected with an efficiency that is plausible for these experiments (≈10%, based on immunofluorescence microscopy), but also displays neuron-like characteristics. For confirmation of plasmids by Western blotting, HEK293 cells were instead used, due to their significantly higher transfection rate and therefore higher level of protein expression.

Aggregation of TRIOBP-1 Can Arise through Its Isoform-Specific N-Terminal Unstructured Region
To confirm the importance of these residues for aggregation, constructs were generated encoding full-length TRIOBP-1, but with deletion of AA 333-340, 333-343, 341-345 or 344-345. Unexpectedly, all of these still formed aggregates when expressed in neuroblastoma cells ( Figure 4A,B, with aggregation seen in the majority of transfected cells; quantified results in Supplementary Figure S6). This suggests that, while AA 333-340 are important for aggregation, they are not the only region of the protein involved. This was surprising, as previously constructs beginning immediately after the PH domain were shown not to aggregate once AA 324-348 were deleted [15]. Therefore, either the PH domain is involved in aggregation, or the N-terminal optionally translated region, which arises from two alternative start codons in TRIOBP-1 is responsible for it [15]. As previous data has suggested that the PH domain is unlikely to be involved in aggregation [9], we instead tried removing the 59 optionally translated amino acids from TRIOBP-1 (AA 1-59), in addition to deleting AA 333-340. Strikingly, this almost completely abolished aggregation compared to the full-length protein ( Figure 4C), as could be confirmed in a blinded quantitative assay ( Figure 4D). Specifically, introducing these deletions caused the number of cells with clearly visible aggregates to drop from around 50% to zero (p < 10 −3 , Welch's t-test), while the number of cells with no sign of aggregation rose from less than 5% to around 80% (p < 10 −8 , Welch's t-test). Other cells displayed TRIOBP1-1 structures that may or may not be aggregates.

Aggregation of TRIOBP-1 Can Occur through Either Its Optionally Translated N-Terminus, or a Loop near the Center of the Protein
To confirm whether the AA 333-340 of TRIOBP-1 are required for aggregation even in the absence of the optionally translated N-terminal domain, we cloned a vector encoding AA 60-652 of TRIOBP-1, corresponding to the 593 AA TRIOBP-1/Tara species. This showed signs of aggregation when expressed in neuroblastoma cells, but seemingly less than the 652 AA version ( Figure 5A). In a blinded quantitative assay, it was confirmed that deletion of either AA 1-59 or AA 333-340 led to a reduction in the number of cells that showed clear TRIOBP-1 aggregation ( Figure 5B, p tukey < 0.001 and p tukey = 0.001 respectively, ANOVA, F > F crit , p < 0.001), with deletion of AA 1-59 having the larger effect, and also showing a reduction in the number of cells that showed clear absence of TRIOBP-1 aggregation (p tukey = 0.012, ANOVA, F > F crit , p < 0.001). Deletion of both of these regions did not significantly affect the number of clearly aggregating cells, but did increase the number of cells that showed a clear absence of aggregation (p tukey = 0.008 and p tukey < 0.001 when compared to deletion of AA 1-59 or 333-340 alone). Therefore, it appears that each of these regions alone can affect aggregation of TRIOBP-1, but it is the combination of the two that has the most significant effect, as summarized in Figure 6.       [15]. (C) Deletion of AA 324-348 in a vector lacking the optional N-terminal region and PH domain also blocks aggregation [15]. (D) Further N-terminally truncated vectors narrow the critical region to AA 333-340 (see Figure 3). (E) Deletions of the aggregation critical region in otherwise full-length TRIOBP-1 does not block aggregation (see Figure 4). (F) Double deletion of the N-terminal optionally translated region and the central aggregation critical region abolishes aggregation (see Figure 5).

Discussion
The study of protein aggregation in mental illness is a relatively novel approach to uncovering the pathophysiology of these conditions, and one with the potential to act as an integration point for both genetic and environmental risk factors for them [3]. To date, several proteins have been published that may form insoluble aggregates exclusively in the brains of patients with one or more of these conditions, including TRIOBP-1, which was implicated through its insolubility in post mortem brain tissue from patients with schizophrenia [9].  [15]. (C) Deletion of AA 324-348 in a vector lacking the optional N-terminal region and PH domain also blocks aggregation [15]. (D) Further N-terminally truncated vectors narrow the critical region to AA 333-340 (see Figure 3). (E) Deletions of the aggregation critical region in otherwise full-length TRIOBP-1 does not block aggregation (see Figure 4). (F) Double deletion of the N-terminal optionally translated region and the central aggregation critical region abolishes aggregation (see Figure 5).

Discussion
The study of protein aggregation in mental illness is a relatively novel approach to uncovering the pathophysiology of these conditions, and one with the potential to act as an integration point for both genetic and environmental risk factors for them [3]. To date, several proteins have been published that may form insoluble aggregates exclusively in the brains of patients with one or more of these conditions, including TRIOBP-1, which was implicated through its insolubility in post mortem brain tissue from patients with schizophrenia [9].
We analyzed TRIOBP-1 aggregation using two insoluble protein purification protocols (Supplementary Figure S1). Of these, the low stringency protocol detects not only aggregated unfolded proteins, but also actin which is insoluble as a result of its forming high molecular weight, but still correctly folded, F-actin fibers. This protocol was originally developed as a more sensitive manner of detecting insoluble proteins, and has previously been used to identify a number of actin-related proteins as experiencing proteostatic changes in rats with cognitive impairment [25]. In comparison to the "actin-positive" low stringency protocol, the high stringency protocol is "actin-negative", detecting aggregating transgenic DISC1, but not actin. This is the protocol, or close variations of it, that has been used in the majority of studies to date regarding protein aggregation in mental illness. Curiously, while DISC1 was present strongly in the high stringency insoluble fraction, Tau was present at a more reduced level. This is consistent with the fact that formic acid extraction, an established method for purifying insoluble Tau [28], also leads to insoluble fractions which contain actin [29], and are therefore also "actin-positive". The combination of these two protocols, one optimized for high stringency detection of only the most insoluble aggregated proteins in the brain, and the other optimized for more sensitive detection of a wider array of insoluble protein complexes (folded or otherwise), can thus aid the separation of insoluble proteins into subtypes based either on their macromolecular structure or on whether the proteins remain folded (Figure 7). Figure S1). Of these, the low stringency protocol detects not only aggregated unfolded proteins, but also actin which is insoluble as a result of its forming high molecular weight, but still correctly folded, F-actin fibers. This protocol was originally developed as a more sensitive manner of detecting insoluble proteins, and has previously been used to identify a number of actin-related proteins as experiencing proteostatic changes in rats with cognitive impairment [25]. In comparison to the "actin-positive" low stringency protocol, the high stringency protocol is "actin-negative", detecting aggregating transgenic DISC1, but not actin. This is the protocol, or close variations of it, that has been used in the majority of studies to date regarding protein aggregation in mental illness. Curiously, while DISC1 was present strongly in the high stringency insoluble fraction, Tau was present at a more reduced level. This is consistent with the fact that formic acid extraction, an established method for purifying insoluble Tau [28], also leads to insoluble fractions which contain actin [29], and are therefore also "actin-positive". The combination of these two protocols, one optimized for high stringency detection of only the most insoluble aggregated proteins in the brain, and the other optimized for more sensitive detection of a wider array of insoluble protein complexes (folded or otherwise), can thus aid the separation of insoluble proteins into subtypes based either on their macromolecular structure or on whether the proteins remain folded (Figure 7).

cols (Supplementary
Having previously used the high stringency protocol to determine that TRIOBP-1 is present in the pooled aggregated protein of a group of schizophrenia brain samples [9] we here demonstrate the existence of aggregated TRIOBP-1 in a broader and independent subset of patients with major mental illness. Specifically, TRIOBP-1 could be detected in the high stringency protein fraction of approximately half of the brain samples from patients examined. It is interesting to note that this occurred in both schizophrenia and major depression, suggesting that it is a mental-illness related event which crosses clinical diagnostic boundaries. This is consistent, however, with our previous identification of DISC1, dysbindin-1 and CRMP1 aggregation across multiple psychiatric diagnoses [2,7,8]. Representation of different species of TRIOBP-1, DISC1, Tau and actin described in the brain, arranged by solubility. Total brain homogenate contains all of these species, while the insoluble fractions derived from the low and high stringency purification protocols described here include only increasingly smaller subsets of insoluble protein. Representation of different species of TRIOBP-1, DISC1, Tau and actin described in the brain, arranged by solubility. Total brain homogenate contains all of these species, while the insoluble fractions derived from the low and high stringency purification protocols described here include only increasingly smaller subsets of insoluble protein.
Having previously used the high stringency protocol to determine that TRIOBP-1 is present in the pooled aggregated protein of a group of schizophrenia brain samples [9] we here demonstrate the existence of aggregated TRIOBP-1 in a broader and independent subset of patients with major mental illness. Specifically, TRIOBP-1 could be detected in the high stringency protein fraction of approximately half of the brain samples from patients examined. It is interesting to note that this occurred in both schizophrenia and major depression, suggesting that it is a mental-illness related event which crosses clinical diagnostic boundaries. This is consistent, however, with our previous identification of DISC1, dysbindin-1 and CRMP1 aggregation across multiple psychiatric diagnoses [2,7,8].
Our results suggest that TRIOBP-1 aggregation, as defined by the high stringency protocol, identifies a substantial subset of mental illness patients as sharing a common disrupted biological process, potentially more so than can be identified by DISC1 aggregation [2]. While it does not necessarily follow that aggregation of TRIOBP-1 must be a causative factor in mental illness, these data strongly imply that it forms a distinct part of the cellular pathology of these conditions. Two of the 50 samples from control individuals also displayed aggregated TRIOBP-1. These may be false positives; however, the presence of aggregated TRIOBP-1 in two chronic mental illnesses also suggests the possibility that it in fact exists in a broader range of conditions, and it thus cannot be discounted that these two control individuals may have carried an unreported, presumably subtler, condition which is also associated with this aggregation event. Replication of our findings in even bigger patient samples, ideally with differing ethnicities, would therefore prove revealing.
In comparison, when the low stringency protein purification technique was employed on the brain samples, we observed an inverse correlation of the amount of pelleted material with that of the high stringency protocol. Since the low stringency protocol also purifies F-actin, it is likely that the TRIOBP species detected in this fraction are present due to their direct association with F-actin, as part of the known role of TRIOBP-1 in actin polymerization [11,14,15]. In contrast to findings with the high stringency protocol, the abundance of these TRIOBP species is actually decreased in mental illness. It is therefore possible that their actin-binding function is inhibited, either as a cause or as a consequence of biological changes involved in mental illness.
When TRIOBP-1 was first found to aggregate in mental illness, bioinformatic analyses were performed to predict whether certain regions of the protein may have the propensity to induce aggregation. This implicated the PH domain, although deletion of this region did not affect aggregation of mouse TRIOBP-1 [9] ( Figure 6A). Subsequent studies involved expressing only the C-terminal half of the protein, consisting of most of the coiled-coil domains, but lacking the first coiled-coil and/or the loop between the first two coiled-coils. These experiments confirmed that this loop, mapped down to 25 amino acids (AA 324-348), was critical for aggregation [15] ( Figure 6B). Crucially, a longer construct incorporating the unstructured "mid-region" of TRIOBP-1 and the first coiled-coil, but lacking these 25 amino acids, still did not form aggregates [15] ( Figure 6C). Here we began by also using C-terminal fragments of TRIOBP-1 including only a portion of the 25 amino acid-long loop between the first two coiled-coils, allowing us to narrow down the likely aggregation domain to just 8 amino acids (AA 333-340, Figure 6D). We, therefore, anticipated that when these amino acids were deleted from otherwise full-length TRIOBP-1, the aggregation propensity of the process would be abolished, as was seen with a longer construct previously ( Figure 6C). Surprisingly, however, knocking out of these 8 amino acids, or other portions of the loop, had no obvious effect on TRIOBP-1 aggregation ( Figure 6E), indicating that the optionally translated N-terminal region of TRIOBP-1 may also be able to induce aggregation. This was confirmed by the deletion of both the N-terminal untranslated region (AA 1-59) and the 8 amino acids from between coiled-coils 1 and (AA 333-340), which together abolished aggregation ( Figure 6F). It is therefore apparent that either of these regions alone is capable of inducing aggregation of TRIOBP-1. However, based on assays knocking out one or both sites, it is clear that aggregation occurs most consistently when both sites are present in the protein.
While the 652 AA long form of TRIOBP-1 therefore seemingly has two mechanisms through which it can form aggregates, one based on its extreme N-terminus, and one based on the loop between two coiled-coils in the center of the protein, it should be noted that this form of TRIOBP-1 likely represents a minority of TRIOBP-1 protein species in the cell [15]. The presence of a second Kozak sequence and start codon means that TRIOBP-1 can also be expressed as a 597 AA long protein, starting at AA 60, as counted in this paper. Indeed, many studies, both classic and recent, assume that this is the principle TRIOBP-1 species. 597 AA long TRIOBP-1, referred to as Tara in many publications, therefore lacks the Nterminal unstructured region, and possesses only one aggregation-critical region, implying only one mechanism of aggregation. Indeed, it is possible that the doublet of TRIOBP bands seen around 72 kDa in the brain samples represents the 652 and 597 AA long forms of TRIOBP-1, although other possibilities, including protein phosphorylation [19,22,23,30], could also account for it. Similarly, longer splice variants of TRIOBP, such as TRIOBP-5 and TRIOBP-6 should also form protein aggregates [15,17], presumably as a result of them having extensive sequences overlapping with TRIOBP-1. While these long TRIOBP isoforms contain the central aggregation region of TRIOBP-1, they lack the N-terminal 59 AA unstructured region [12][13][14], again suggesting that their aggregation can occur only through the central aggregation region and its associated mechanism. Finally, smaller TRIOBP species of 40-60 kDa were detected in insoluble fractions of brain samples here, and were also seen previously in cell systems [9], which are detectable using antibodies against C-terminal parts of TRIOBP-1. While it is unclear whether these species represent the products of additional TRIOBP splice variants or processed forms of TRIOBP-1, it is likely that they would also possess only the one, central, aggregation critical region.
Determining mechanistic details of the two paths by which TRIOBP can form protein aggregates, therefore, represents an important next step in investigating this emerging aspect of mental illness molecular pathology. Additionally, the existence of plasmid constructs encoding near-full-length TRIOBP-1 that does not aggregate provides a powerful tool for future experiments to investigate the consequences of aggregation. For example, while rodent models of DISC1 and dysbindin-1 aggregation show clear effects on behavior, brain anatomy and biochemistry [27,31,32], it remains to be confirmed to what extent these phenotypes arise from the presence of human aggregation protein in the brain, and how much from over-expression of the human protein in the brain at all. This could be addressed, to a significant extent, in a system involving three lines of animals, one control lacking a transgene, one expressing full-length aggregating protein, and one expressing a mutant version of the protein that lacks or has a significantly reduced tendency to aggregate, such as the TRIOBP-1 lacking amino acids 1-59 and 333-340.
Uncovering the aggregation critical regions of TRIOBP, therefore, gives a powerful experimental tool to help understand the mechanisms behind this potentially pathological event, which is now implicated in the pathology of both schizophrenia and major depressive disorder, but also to understand the consequences of that aggregation for mental illness.

Brain Samples
Post mortem human brain specimens from BA23, the anterior cingulate cortex, were obtained from 50 control individuals, 25 patients with schizophrenia and 25 patients with major depressive disorder from the Human Brain and Tissue Repository of the Lieber Institute for Brain Development. BA23 was dissected under visual guidance using a hand-held dental drill, just superior to the rostrum of the corpus callosum. Details of the informed consent, diagnostic procedures, and curation are elaborated in detail in a previous publication from our group [33]. Demographic details of these can be found in Supplementary Table S1. Brains from a transgenic rat expressing human DISC1 [27] were extracted according to methods approved by the LANUV (State Agency for Nature, Environment and Consumer Protection), North Rhine-Westphalia, Germany. Brains from mice over-expressing a 383 AA form of human Tau with a P301S mutation [26] were a gift from Dr. Michel Goedert (University of Cambridge, Cambridge, UK).

High Stringency Insoluble Protein Fraction Purification Protocol
Brain samples were homogenized to 10% (w/v) in 50 mM HEPES pH 7.5/250 mM sucrose/5 mM magnesium chloride/100 mM potassium acetate/2 mM PMSF containing protease inhibitor cocktail, and were lysed by the addition of Triton X-100 to a final concentration of 0.5% (v/v). Lysate was then centrifuged at 20,000× g for 20 min. Pellet was resuspended in the same buffer (including Triton X-100) and centrifuged a second time. Pellet was resuspended in 50 mM HEPES pH 7.5/1.6 M sucrose/100 mM potassium acetate/1 mM PMSF/0.5% Triton X-100 and ultracentrifuged at 130,000× g for 45 min. This was repeated a second time. Pellet was resuspended in 50 mM HEPES pH 7.5/1 M sodium chloride/20 mM magnesium chloride/30 mM calcium chloride containing 40 units/mL DNaseI and protease inhibitor cocktail. Lysate was then incubated at 4 • C for 16 h on a rotary wheel. Lysate was ultracentrifuged at 130,000× g for 45 min. Pellet was resuspended in this same buffer (minus the DNaseI) and ultracentrifuged at 130,000× g for 45 min. Pellet was resuspended in 50 mM HEPES pH7.5/0.5% sarkosyl and ultracentrifuged at 130,000× g for 45 min. This was repeated, and the final pellet was resuspended in dissolution buffer (10 mM Tris/0.5 M triethylammonium bicarbonate/8 M urea/1% SDS). This entire procedure, including centrifugation steps, was done at 0-4 • C, is adapted and optimized from one of our previously published protocols [2] and is summarized in Supplementary Figure S1.

Plasmids
Plasmids encoding full-length TRIOBP-1 and two N-terminally truncated versions of it, have been described previously [15]. Additional constructs encoding the coiled-coil regions of TRIOBP-1 were made by subcloning reading frames, and then transferring into either a Gateway entry vector, either pDONR/Zeo (using BP clonase recombination, enzyme and plasmids: Thermo Fisher Scientific, Waltham, NJ, USA) or pENTR1A no ccDB (Dr. Eric Campeau, supplied by AddGene, Watertown, MA, USA, clone 17398 [34], by ligation). Vectors encoding full-length TRIOBP-1 with internal deletions were produced by subcloning of 5 and 3 fragments of the gene, which were then sequentially ligated into pENTRA1A no ccDB. Reading frames were transferred from entry vectors into destination vectors using LR clonase II recombination (Thermo Fisher Scientific). Destination vectors used were pdcDNA-FlagMyc (B. Janssens, supplied by the BCCM/LMBP Plasmid Collection, Zwijnaarde, Belgium, clone LMBP 4705) and pdECFP (Dr. S. Wiemann, BCCM/LMBP Plasmid Collection, clone LMBP 4548 [35]). All vectors were confirmed by sequencing. Details of all plasmids used are in Supplementary Table S2 and the primers used to clone  them are in Supplementary Table S3.