GTPBP1 resolves paused ribosomes to maintain neuronal homeostasis

Essential revisions:

Listed below are the points we find especially important to address in a revised version of the manuscript, that we hope to receive from you in short order.

1) The manuscript features ribosome profiling and RNA-seq that are presented sequentially in the subsection “GTPBP1 is a novel ribosome-rescue factor”. However, which dataset is being analyzed and discussed is not always clear to the reader. It should be clarified whether the profiling was normalized for RNA- translational efficiencies. This normalization point should be clarified in the Results text and the legend for Figure 2. Figure 2D refers to analysis of differential gene expression, a generic term that is used later for the RNA-seq analysis. More precision in the language through the translational efficiency and mRNA levels would be helpful. Furthermore, Figure 3 should clarify whether the "analysis of differential gene expression" refers to mRNA measurements.

We appreciate the input as this may also confuse readers. We have clarified our language to specifically identify the datasets we are referring to. In addition, we defined in the text, legends, and Materials and methods, that translation efficiency (TE) data reflect normalization of the abundance of ribosome-protected fragments to that of RNA-Seq reads and refer to these data more consistently as TE Gtpbp1 or TE Gtpbp2. Our RNA-Seq data were utilized to measure mRNA abundance and we now refer to changes in mRNA abundance as transcriptional differences in gene expression (DE Gtpbp1 or DE Gtpbp2) in the text and in the legend for Figure 3.

2) At several points in the manuscript it is stated that ribosome stalling induces GCN2 and mTORC1 signaling. The evidence for GCN2 induction by ribosome stalling is strong based on research from this lab and others. The idea that ribosome stalling directly induces neuron-specific changes in mTOR signaling warrants some caution (as opposed to indirect regulation through changes in the integrated stress response, amino acid pools, or other mechanisms).

We agree with the reviewer and did not mean to imply that changes in mTOR signaling are a direct result of ribosome stalling. We have included a statement in the Discussion to this effect.

3) GTPBP1 and GTPBP2 are very similar and act by a similar mechanism. It is thus very surprising that only 50% of AGA pausing genes with increased AGA occupancy intersected between B6J.Gtpbp1^-/- and B6J-Gtpbp2^-/- mice (Figure 2—figure supplement 1B, Supplementary file 1). It is even more surprising that the biological processes as defined by Gene Ontology (GO) analysis of pausing genes

revealed enrichment in different biological processes between Gtpbp1 and Gtpbp2 (Figure 2—figure supplement 1C).

Pausing on AGA codons is very robust in the Gtpbp1^-/- cerebellum as shown in Figure 2 A and B, and is consistent with our previous data from the Gtpbp2^-/- cerebellum. However, the genes with pause sites in each biological replicate for either genotype only overlap by about 50% (Figure 2C). This suggests that ribosome pausing is stochastic, i.e., it can occur at any AGA codon in any gene, but doesn’t occur at all codons at a given time. Capturing all stochastic events would require a substantially higher sequencing depth and therefore, we likely only scratched the surface in defining pausing events in these mice. Furthermore, as pointed out in the Results and the Materials and methods, we only considered strong pauses for our analysis (z-score above 10, that is pausing is at least 10 standard deviations above background). Our thresholds are picked to have high confidence in our pausing data. However, weaker pauses may be missed at the expense of high confidence. Thus, all things considered, we wouldn’t expect to observe a higher overlap between two different genotypes.

As we would expect from stochastic pausing, pausing genes are enriched for many GO terms. Most of TOP GO terms (17 out of 24) overlap between Gtpbp1 and Gtpbp2 and the GO terms that appear unique to pauses which occur in a given genotype, are again likely due to the depth of our ribosome profiling datasets. Indeed, to circumvent the limitations of single/individual experiments, our conclusion of Gtpbp1 and Gtpbp2 are likely acting in the same pathway is not just derived from our profiling data, but rather is a collective hypothesis based on both our genomic and genetic data.

All this consistent with the finding that GTPBP1 and GTPBP2 are not mechanistically redundant, but the mechanistic insight, although not known should be mentioned and discussed.

Our study, together with the recent biochemical study from Zinoviev et al., 2018, are the first studies that analyze Gtpbp1 and Gtpbp2 side by side. Thus, we only have limited understanding about the shared or unique functions of these two proteins and we have included a statement in the Discussion to point out that the function of these proteins is still unknown. The biochemical experiments by Zinoviev et al. revealed striking differences in mRNA decay for these two proteins which we have discussed in the Discussion.

4) Figure 1C – It appears that there is much more GTPBP2 than GTPBP1. Is it correct? What is the ratio? Also, the two proteins do not co-localize. Localization is even mutually exclusive. What do these results mean?

Figure 1C shows our in situ hybridization experiments for mRNA as mentioned in the Results and figure legend. Thus, the fluorescent punta do not indicate protein localization, nor does the nature of this experiment (different probes for each gene that are labeled with different fluorophores) allow us to measure levels of Gtpbp1 and Gtpbp2 transcripts. However, these experiments do show that the mRNAs for Gtpbp1 and Gtpbp2 are found in the same neurons and are expressed in both neurons that degenerate and those that do not.

5) Figure 3C – why is the amount of S6 and its phosphorylated form reduced in the hippocampus, but not in the cerebellum. Any differences in mTOR or upstream effectors amounts between these two structures?

We do not know why the mTOR pathway responds to defects in translation in cell type-specific fashion. Both positive and negative regulators of the mTOR pathway (e.g. AKT, ERK, PTEN, TSC1/2, etc.) are ubiquitously expressed (as far as we know) and we detected expression these genes in our transcriptome datasets from the cerebellum. However, proteomics data might be more insightful as many mTOR regulators act at the protein level and/or thru post-translational modifications. Even so to point to differences in upstream effectors that alter mTOR would require systemic genetic and/or molecular studies across different neuronal populations.

A large body of literature has shown that the mTOR pathway can respond to numerous cellular insults and that the activity of this signaling pathway is influenced by changes in cellular homeostasis including protein synthesis, neuronal activity and energy balance. The particular alteration(s) in cellular homeostasis in the Gtpbp1^-/- and Gtpbp2^-/- brain that triggers a decrease in mTOR activity is unknown, making it even more difficult to speculate about the cell type-specific responses we observed. In the Discussion, we suggest that defects in translation in different cell types lead to differences in the amount and type of perturbations, perhaps due to the differences in basal translation between cells, and this in turn could result in differential homeostatic changes.

6) The experimental evidence provided here strongly supports the conclusion that the two GTPBP's function in the same pathway. This might be expected, given the similarity of the encoded proteins but surprisingly, the compound BL6/J GTPBP1∆/∆;GTPBP2∆/∆ is no more compromised than either BL6/J GTPBP1∆/∆ or BL6/J GTPBP2∆/∆ single mutants. This last observation provides genetic evidence that despite considerable similarity in their encoded proteins, the two genes act at different steps in the same pathway. This is a surprising finding that stands to influence thinking in this area and will generate a search for the mechanistic basis of differences in function between GTPBP1 and GTPBP2 and the specific steps they act on in coping with stalled ribosomes (see also point 3, above). Less interesting would be a study that concluded that these two proteins do more or less the same thing and that the gene duplication that gave rise to their presence in the mammalian genome represents some sort of fine tuning, the basis of which remains for now obscure. The critical experiment examining this point is shown in Figure 1F and G, which compare the number of granule cells in two lobes of the cerebellum at a single time point (6 weeks) in three mice each with the relevant genotypes. The equivalence of the loss of cells between the three relevant genotypes (GTPBP1∆/∆;GTPBP2∆/∆, GTPBP1∆/∆ and GTPBP2∆/∆) is impressive and fits well the conclusion that GTPBP1 and GTPBP2 are active on different steps in the same pathway. However, the lack of a temporal component to the measurements, limits the strength of that key conclusion. For example, it is possible that by six weeks of age the histological changes had plateaued, obscuring important differences in the rate at which that plateau had been reached by the different genotypes. If the compound mutants were actually to degenerate more quickly than either single mutant, the paper's key conclusion would be wrong in that it would suggest that the two GTPBP's are in fact redundant and carry out the same step biochemically. Addressing this issue seems key to the significance of this paper.

We included additional time points (3 and 4 weeks) in Figure 1 for the double mutants. Our data show that onset and progression of granule cell death are very similar between Gtpbp1^-/-, Gtpbp2^-/-, and Gtpbp1^-/-; Gtpbp2^-/- cerebella.