Glial immune-related pathways mediate effects of closed head traumatic brain injury on behavior and lethality in Drosophila

In traumatic brain injury (TBI), the initial injury phase is followed by a secondary phase that contributes to neurodegeneration, yet the mechanisms leading to neuropathology in vivo remain to be elucidated. To address this question, we developed a Drosophila head-specific model for TBI termed Drosophila Closed Head Injury (dCHI), where well-controlled, nonpenetrating strikes are delivered to the head of unanesthetized flies. This assay recapitulates many TBI phenotypes, including increased mortality, impaired motor control, fragmented sleep, and increased neuronal cell death. TBI results in significant changes in the transcriptome, including up-regulation of genes encoding antimicrobial peptides (AMPs). To test the in vivo functional role of these changes, we examined TBI-dependent behavior and lethality in mutants of the master immune regulator NF-κB, important for AMP induction, and found that while sleep and motor function effects were reduced, lethality effects were enhanced. Similarly, loss of most AMP classes also renders flies susceptible to lethal TBI effects. These studies validate a new Drosophila TBI model and identify immune pathways as in vivo mediators of TBI effects.

TBI is an important subject to study and work in Drosophila might potentially help to understand human pathology. However, I do have some concerns regarding the brain specific trauma induction, this approach in my view is likely to generate disruptions of (1) nerves running in the neck of the fly and disruptions in the foregut which is also located in the neck. In addition, the forces used in this study (8.34 N) appear quite high in particular when comparing to the forces obtained in the device of Katzenberger et al., (2013, 2 N). This should at least be discussed. And more importantly it must be analyzed whether the neck is not damaged.
We now acknowledge in the Discussion the potential that other head tissues including the neck may have been damaged in our assay as is the case also for mammalian TBI assays and other assays used to address TBI in flies. We believe that most if not all of the phenotypes observed can be attributed to brain damage (i.e. we observe cell death in the brain and glial gene expression changes) but acknowledge that we cannot exclude damage to other tissues/structures contributing.
We also make note in the Discussion of the difference in force used in our assay relative to the Katzenberger paper. We may need a higher force because brain damage is caused by the direct impact of the solenoid to the fly head, where the fly head moves with the solenoid (sup movie 1) rather than full body injury or compression injuries used in the other Drosophila TBI assays.
We also note in the Discussion that our assay allows us to deliver strikes directly to fly head, without having to anesthetize flies, thus avoiding anesthesia effects on behavior and TBI outcomes.
The present work provides a number of interesting finding -which however are not entirely new. The Bonini group just published a paper "Dynamic neural and glial responses of a head-specific model for traumatic brain injury in Drosophila" presenting a head-specific TBI model (Saikumar et al., PNAS 2020). In addition, Wassarman and colleagues published paper in 2015 where they showed that TBI results in blood-brain barrier permeability defects (Katzenberger 2015). Both of these two important papers are not mentioned at all -despite the fact that 130 references are presented.
We have added references to both studies throughout the revised manuscript. The Bonini paper was published just prior to our submission. We thank the reviewer for pointing out this issue.
The finding that genes of innate immunity response are upregulated has been made before. The authors should compare their sequence data with the previously published datasets. The specificity of the TRAPseq method should be documented (e.g. expression of alrm, wunen-2, repo, gliotactin and moody should be compared to neuronal gene expression e.g. nsyb or elav). It should be stated that gliotactin is not a marker for peripheral glial cells but is expressed by the subperineurial glia -as moody and wunen-2 is not a marker for astrocytes but shows a rather broad expression in the adult brain (see Stein Aerts data single cell seq dataset). The finding that CG40470 is the only gene that is persistently downregulated in all days tested is interesting but unfortunately not analyzed in further detail.
Thank you for these remarks. We have added a comparison to previous datasets with regards to upregulated immune response genes in the Discussion section.
With regards to TRAP-seq, we have updated Fig 4A in the following ways: we have removed wunen and moody as markers for glial expression and we have added elav and nsyb as neuronal markers (expression levels for both were not significantly altered after TBI). * We agree that CG40470 is an interesting gene to follow up on in future studies and make note of this finding in the text. Major comments: 1. One complicated feature of TBI is the heterogeneity of outcomes in injured animals, which is also demonstrated in this study. Here, it appears that a large proportion of flies die within ~1-2 days post-TBI, after which it seems that the remaining flies have only a slightly accelerated aging curve. It is appropriate, then, to separate these flies into different groups for analysis. For example, do flies that die rapidly exhibit stronger sleep impairments? If these rapidly-dying flies are separated from the analysis, does the survival curve of the remaining TBI-treated flies look more comparable to sham-treated controls?

Rev. 2:
This is an excellent point. We reanalyzed our mortality and sleep data to test whether flies that die early behave differently in these assays.
For the mortality data, we removed early deaths from both controls and TBI flies cumulatively, for up to two weeks after TBI, and performed log rank test on the remaining flies. In all cases, survival rate is still significantly decreased in the TBI group, suggesting that the increased mortality is not entirely due to flies that die early. This data is now in Sup Fig 2. For sleep data, we removed all flies that died during the seven days from the analysis and analyzed sleep for two conditions: Condition 1 compares sleep in flies that survived TBI to controls. Here we see that the reduced sleep phenotype is still present validating our original finding.
Condition 2 compares sleep in all flies that died during the seven days post TBI to controls. Here we observe an interesting phenotype. While sleep is unaffected at day 1 post-TBI, sleep is significantly increased during post-TBI days 2 and 3. This is only due to increased sleep during the light phase, where bout lengths are significantly increased. Night time sleep is unaffected. Also, wake activity during the light phase does not differ between controls and TBI-treated flies, indicating that the increased sleep phenotype is not due to impaired locomotion. Thus the flies that die exhibit a distinct sleep phenotype from those that do not die.
These data are now in Sup Figs 3 and 4. Figure 2C shows a significant elevation in TUNEL staining to label apoptotic cells in the brain after TBI. Because the authors focus their sequencing studies on glia, it would be informative to test whether the TUNEL-positive cells are neuronal or glial. Do glia activate AMP expression in response to neuronal death? Or to their own apoptosis?

2.
This is an interesting point and we note the ambiguity of the identity of the cells in the text. Experimentally addressing this issue would be beyond the scope of this study. Figure 6 that the mortality and behavioral phenotypes can be partially dissociated. However, this is not consistently addressed. The AMP mutants used in Figure 7 are only tested for mortality; it remains unclear if these AMP mutants, either individually or in combination, could also account for the behavioral phenotypes. Do flies in the "ABC" class from Figure 7 show changes in climbing and sleep behavior after TBI?

The authors have demonstrated in
We tested whether TBI affects climbing and sleep in the ΔAMP flies. Baseline climbing did not differ between ΔAMP flies and controls without TBI and both groups showed a similar reduction in climbing behavior after TBI. However, TBI affected sleep in the opposite direction in ΔAMP flies. Where the wildtype controls show a reduction in sleep 24 hours after TBI, ΔAMP flies show increased sleep post-TBI. Thus, AMPs appear to be important for mediating TBI effects on sleep. This data is now in Sup Fig 8. 4. The authors use TUNEL staining for validation of the TBI model but do not address whether the immune genes that affect survival after TBI also affect neuronal death. To provide a more complete characterization of the role of immune genes, the authors could compare neuronal death in relish mutants and wild-type flies after TBI. This is also an interesting point and one that is worthwhile for a follow up study (noted in the Discussion) but beyond the scope of this study.

5.
The authors claim that the immune response mediates survival after TBI. However, it is possible that immune responses are not playing an active role after TBI but rather that the mutants are more sensitized/susceptible to TBI. The authors should either discuss this caveat in the text or address this experimentally using inducible genetic systems to knock down relish starting ~24h before TBI.
Yes, we cannot rule out a role for immune mutants rendering the flies susceptible to TBI. However, the finding that immune gene expression changes after TBI suggests a role in mediating the response. We more fully discuss this in the text.

Sleep architecture data (Figs. 3D-F) should show daytime and night time results separately.
Sleep architecture data are now split by daytime and nighttime. Results are described in response to reviewer 1.
Minor comments: 1. Are flies that die over the course of the 7-day sleep experiment excluded from the entire dataset, or only from the days after they died? These flies might provide insight into whether the severity of behavioral changes might correlate with mortality.
Flies that die over the course of the 7-day sleep experiment were included in the dataset. As noted above, (Major Comment #1) we found that excluding these flies from the dataset did not change our conclusions. However, analyzing sleep only in the flies that died over the course of the experiment showed that sleep is increased in this group on days 2 and 3 post-TBI. This is a really interesting point, we have added some discussion of the possibility that TBI-induced immune responses may be secondary to their effects on sleep. We have added a paragraph to the discussion on sleep responses after injury.

Figure 7 lacks n values
We have added n values to this figure.

Major comments: 1) Given the novelty of this assay the authors might consider additional descriptions of the behaviors themselves. For example, quantifying differences in behavior immediately following the TBI events through the first few hours of recovery.
This is an excellent point. To address it, we induced TBI at three intensities (x1, x5, x10) and used video tracking in individual flies, housed in small petri dishes, to record locomotor behavior in the four hours immediately after TBI induction. Locomotion metrics were compared to sham treated controls.
We found that after TBI, ~25% flies in the TBIx1 condition are immobile, versus ~55% in the TBIx5 and x10 conditions. Flies in the x1 condition started moving within seconds, while flies in the x5 and x10 conditions started moving after minutes (3.3 and 10 min respectively). Walking speed was reduced in all three groups during the first hour post-TBI, but the TBIx1 and x5 groups had recovered by the second hour. Walking speed remained impaired for all four hours in the TBIx10 group. Overall activity (% of time active) was significantly reduced in the TBIx5 and x10 groups for the first hour after TBI, but unaffected in the TBIx1 group. We also observed some locomotor defects (circling, slow walking, sideways walking, backwards walking, jumping) shortly after TBI onset, in a dose dependent manner (25%, 45% and 50% in the TBIx1, x5 and x10 groups respectively). These movement disorders only occurred in flies that were immobile immediately after TBI and were not observed in flies that immediately started walking.
This data in now shown in Sup Fig 1. 2) It is interesting that the behavioral deficits return to normal after a few days. I would be very interested to know if they have memory deficits that persist beyond this point. This may be beyond the scope of the paper, but would be worth discussing.
In their recently published Drosophila TBI model, Saikumar et al (2020) demonstrated that memory deficits, measured by courtship conditioning, are present ten days after severe TBI. Surprisingly, baseline courtship behavior was unaffected. We added a brief section to the discussion on this issue.

3) I understand the advantages of targeting the head, even with this precision is it possible to differentiate between neural injury and general stress? Is it possible that targeting a different body region would also lead to climbing/sleep deficits?
This is an interesting point which we acknowledge explicitly in the text.

4) Genetic background is certainly an important factor for sleep and longevity/aging and is therefore likely very important TBI response. Please describe efforts to account for genetic background.
We agree that genetic background is very important in sleep and, likely, the response to TBI. We have taken the following measures: All TBI experiments were carried out in young adult (3-7 days old) male iso31 flies, an isogenic w 1118 control strain commonly used for sleep research. Our NFkB null mutant, Rel[E20] is in a w 1118 background, and we used w 1118 iso31 as control. For the AMP null mutants, we compared survival rates to the iso31 control line provided by the LeMaitre lab.
We have clarified our methods section.

5) Localizing genes to subpopulations of glia would increase the impact of the findings. would be very helpful to sort TRAP-seq data based on the glial subtype that they express in. I understand this is not entirely straight forward and these data sets don't exist for all glia, but they do for Repo and Alrm, and this alone might be useful. An alternative would be to knock genes down in subsets of glia.
Testing the effect of knocking down hits from our glial TRAP-seq, both in all glia and in subpopulations is a follow up experiment we are considering but beyond the scope of the current study. We added some discussion of this in the text.
Minor Comments 1) Line 40: Is TBI really one of the leading causes of death? This seems unlikely. Also, (though perhaps too detailed to address here) I imagine most TBI deaths are in elderly patients, which leads me to wonder if the effects of TBI would differ in aged flies.
According to data compiled by the CDC, TBI is a leading cause of death among children and young adults rather than among the elderly. https://www.cdc.gov/traumaticbraininjury/pubs/tbi_report_to_congress.html To avoid confusion, we have rephrased 'leading' to 'major' on line 40.

2)
While not critical to the scientific content, Figure 1 could be improved to depict the assay. For example, a cartoon diagraming the components of the assay would be more useful to the image in A, which could be placed in the supplemental figures.
We have added a new schematic figure as suggested.
3) Line 258 describes the immediate response of flies to the TBI, and their recovery. Supplemental videos are provided but it would be very useful to quantify this given the novelty of the assay. In addition, the term 'dazed' may inadvertently imply changes in cognitive perception.
This overlaps with Major Comment 1.
We have updated the text to: "flies are often able to stand but only barely respond to tactile stimuli" 4) Figure 3. While not essential, it would be a useful control to show climbing and sleep data in animals given a single strike. Five strikes results in some death, and therefore phenotypes may derive from generalized deficiencies (although the finding that sleep returns to normal after 7 days suggests the effects are specific).
Climbing data for TBIx1 is shown in Fig 2B. There was no effect on climbing 24 hours after TBIx1. We also did not observe sleep effects after TBI x1 (data not shown).
However, our video analysis of locomotion effects in the first four hours immediately after TBI, we see that ~25% of flies are immobile immediately after TBI and flies show decreased walking velocity for the first hour post TBI. However, all locomotor effects in TBIx1 flies are gone by the 2 nd hour after TBI. Figure 2. When do flies die within the 24hrs following TBI? Is it immediate, or hours after? There are also some caveats about using negative geotaxis to infer sensory-motor function (though these are likely shared in rotarod studies). It does not rule out things like general arousal, endurance, or motivation.

5)
After reanalyzing our TBI sleep data for mortality, we found that 16% of flies die immediately after, and 22% of flies die within 24 hours after TBI induction.
We acknowledge the caveats with the negative geotaxis assay.

6) Recent work from the Donlea group showed that antennal axotomy results in increased sleep. It is worth commenting on the difference in sleep phenotypes that result from each type of neural injury.
This is an excellent point. Both groups demonstrated that antennal axotomy increases sleep, a process that facilitates clearance of debris, while our work demonstrates a reduction in sleep with TBI, suggesting that Wallerian degeneration is not a predominant component of TBI. We did note a transient increase in sleep in flies that die suggesting it may play a role there.

7) How was the strength of the TBI-inducing stimulus chosen?
When designing our TBI paradigm, we tested several commercially available solenoids for their ability to induce TBI and used the one that gave the best results. We may need a higher force because brain damage is caused by the direct impact of the solenoid to the fly head, where the fly head moves with the solenoid (sup movie 1) rather than full body injury or compression injuries used in the other Drosophila TBI assays.

8)
In some cases the language could be more precise. E.g. line 422 'Also, quite a few members of the turanadot… ' We have fixed these issues.

Rev. 4:
In this manuscript ("Glial immune-related pathways as mediators of closed head TBI effects on behavior and lethality in Drosophila"), van  Thank you for this comment. We have added the complete data sets of genes with altered expression levels for days 1,3 and 7 post TBI as Supplementary Files 1-3. We now also show in Fig 4A that expression levels of neuro-specific gene elav is not significantly changed after TBI., and nsyb is downregulated. Other neuronal genes (bruchpilot, neurexin, neuroligin) are detected but not up-or downregulated.
2. The authors state that differential gene expression analysis was identified using a p value of 0.1 (and Log2 value of 0.6) as a threshold. This value is high. The authors should clarify why this is an appropriate cutoff value (or choose to make the cutoff more stringent, for example p value of 0.05 or below).
Thank you for your remark. We use FDR corrected p-values (adjusted p-values) to control for the false discovery rate. The significance threshold is an arbitrary value that signals how many false positives one agrees to accept. These thresholds should take into consideration the experimental data and requirements for the downstream analysis, like GO. Depending on the application and further use of the data, values between 0.01-0.1 are generally accepted in the field [2][3][4][5][6][7]. Our choice of more relaxed FDR threshold takes into consideration correlation and range of values for replicates for analyzed conditions, as well as the fact that gene expression differences in the brain might be subtle on transcriptional level. We decided to go with the default threshold implemented in DESeq2 (FDR ≤0.1 [1]), we added additional filter in the form of the fold change ratio factor, requiring at least 50% of change (up or down) in the expression level in relation to the value in control samples, to support the biological relevance of results. These thresholds combined serve as a good indicator of observed effect, and are accepted in the field [i.e. 3,6,7]