Lower synaptic density is associated with depression severity and network alterations

Synaptic loss and deficits in functional connectivity are hypothesized to contribute to symptoms associated with major depressive disorder (MDD) and post-traumatic stress disorder (PTSD). The synaptic vesicle glycoprotein 2A (SV2A) can be used to index the number of nerve terminals, an indirect estimate of synaptic density. Here, we used positron emission tomography (PET) with the SV2A radioligand [11C]UCB-J to examine synaptic density in n = 26 unmedicated individuals with MDD, PTSD, or comorbid MDD/PTSD. The severity of depressive symptoms was inversely correlated with SV2A density, and individuals with high levels of depression showing lower SV2A density compared to healthy controls (n = 21). SV2A density was also associated with aberrant network function, as measured by magnetic resonance imaging (MRI) functional connectivity. This is the first in vivo evidence linking lower synaptic density to network alterations and symptoms of depression. Our findings provide further incentive to evaluate interventions that restore synaptic connections to treat depression.

The main finding is that SV2A density was negatively correlated with severity of depressive symptoms clinical subjects in dlPFC (r=-0.633, p=0.001), ACC (r=-0.634, p=0.001) and hippocampus (r=-0.487, p=0.012). As noted below, it was not immediately clear to me this these findings are the results of a prior hypothesis testing or if the study is meant to be exploratory/hypothesis generating. This is an important point in the context of rigor. Either approach is reasonable, but it is incumbent upon the authors to present the analyses in light of one of the two approaches, I believe. The supportive analyses related to group effects of fMRI are interesting and appropriately presented as exploratory. Overall this is a very interesting paper with some potential limitation related to sample size and lack of clarity about pre-specified features of the statistical plan.
I have a few specific comments below.
1. Intro -maybe a minor quibble, but the paper starts with, "the human brain displays remarkable neuroplasticity in response to stress...Persisting high levels of stress or repeated exposure to extreme stress results in loss of synapses in circuits underlying affective and cognitive processes1." But the data the authors are referring to is in rodent models, and I do not know if we know what happens re synaptic plasticity in humans under stress.
2. Intro, last paragraph, no specific hypotheses are stated. Authors refer to an exploratory approach, which is OK but should be consistent throughout. For example in the Results there is finding presented as primary of correlations between SV2A and a few select brain regions (where they defined a prior prior to data collection?) Then there is what is presented in the results as 'exploratory' but it is unclear if the authors are using these terms in the technical/statistical sense or just general language to indicate what they feel is most important. Since these terms do have technical meanings in this context, I suggest using the terms in that way. appropriate to combine MDD, PTSD and MDD+PTSD in this first study; [2] limited samples sizes that include the number of "pure" PTSD (n=5), sizes of the LS and HS groups, and only a subset of HC with fMRI; [3] significantly different unlabeled mass doses administered across groups; [4] no correction for non-displaceable distribution volume of radiotracer; and [5] the use of "[C-11]UCB-J SV2A VT" and "synaptic density" as equivalent terms (e.g., axis labels). In reference 53 (Fig 1D), it appears that the correlation between SV2A VT and SV2A OD is r2=0.31. How might this translate to the present work -is this lower correlation a concern?
Despite the authors' efforts, some of the work seems a bit exploratory (as noted by the authors) and larger sample sizes and further investigation are likely needed to establish the major findings. A few additional comments/questions remain. Did the authors observe any input function differences across the groups, e.g., did in vivo radiotracer metabolism differ across groups? Could gray matter atrophy impact data interpretation (beyond PVC for the PET and in terms of the MRI ICD metric)? Were regional gray matter volumes compared across groups or considered in the statistical analysis? The authors should describe UCB-J SV2A VT and VT/fp test-retest variability. Which outcome is considered more accurate or reliable, VT or VT/fp? Has outcome reliability been assessed yet in the relevant patient groups?
Overall, it is very exciting novel research.

Methods
• VT (line 423) also includes blood volume? • Is the PVC method related to other methods more commonly applied in the past (e.g. GTM)? Please offer more clarification.
• fMRI: how many functional volumes were acquired? Were the PET data acquisition parameters listed (e.g dynamic acquisition)? Please provide further evidence that the parametric data were well fit. • how does 5 mm fMRI resolution relate to PET resolution?

Results
• Scatter graphs would be more informative, if different symbols or colors indicated HC and/or patient sub-group membership • It would be more informative to show scatter graphs rather than bar graphs when sample sizes are small • Please list sample sizes in figure legends, when possible • Why not show all results on the same y-axis scale? Please include "SV2A" on y-axis scale, although it really is "[C-11]UCB-J VT" • Fig, 2b: Image planes appear to capture different anatomy for the 3 subjects (not spatially normalized?) • Fig, 5: Graphs E and F are less convincing in terms of a linear relationship Discussion • line 226, ". . . exploratory analysis of additional brain regions points to a global effect." Were these data shown? • The discussion is long and sometimes speculative (e.g., the paragraph on sleep, line 229)consider shortening and further focusing the Discussion • It is not always clear how the individuals in the current study (in a major depressive episode at screening) compare to subjects studied in some of the literature findings cited • Line 305, the word "Second" may be missing here • Line 322 The third aspect of the limitation discussion could be more clearly presented/described Supplemental Data • Table S3: Thalamus values appear to be missing Reviewer #3 (Remarks to the Author): The authors investigated SV2A with the radioligand [(11)C]UCB-J in healthy controls and an unmedicated clinical sample comprising subjects with either MDD, PTSD, or with comorbid MDD and PTSD. Both groups were matched for smoking and the ratio of male to female participants. In addition to PET imaging, functional brain connectivity was measured in the clinical sample and a fraction of the healthy controls with fMRI. Participants received a clinical assessment and computerized cognitive tasks. The full clinical sample and the controls did not differ in SV2A VT. In the clinical sample increasing depressive symptom load corresponded to decreasing SV2A VT. When split into subjects with high and low depressive symptom load based on their HAMD score the high load group displayed lower SV2A VT than controls in the DLPFC, ACC, and hippocampus. The full clinical sample displayed reduced functional connectivity of the DLPFC. Lower DLPFC SV2A VT corresponded to a higher DLPFC-PCC connectivity. Higher DLPFC-PCC connectivity corresponded to higher depressive symptom load and lower performance in a working memory task. This carefully conducted investigation uses a novel radiotracer and presents the first in vivo evidence for altered SV2A VT in MDD. It bears a high potential impact for the field and would be of significant interest to the readership of this journal. Consider the points below to improve readability and address some important points in discussing the exciting findings of this study. Major points Stratification for symptom severity resulted in an uneven distribution of diagnoses across groups, with no subjects with PTSD without comorbid MDD in the high depressive symptom load group. This could be emphasized more clearly in the text. The findings on the relation between depressive symptoms and SV2A VT are impressive. An additional aspect of the results would be that no relation was found between SV2A VT and symptoms of PTSD, or anxiety symptoms in general. Consistently, the very informative explorative analysis involving SV2A VT and the factorial structure of HAMD, revealed that decreased SV2A VT corresponds to higher scores on the depression, insomnia, and somatic, but not on the anxiety factor of HAMD. Consider expanding on this in the discussion. Minor points Information on the sensitivity and specificity of [(11)C]UCB-J, such as on page 13, lines 326-328, is central for understanding and appreciating the results. Therefore, consider moving it from the discussion to the introduction and extending it. In the introduction the authors state that "binding to SV2A provides an estimate of the density of nerve terminals" (p. 4, line 94). However, in the results and discussion they use the terms "SV2A density" and "SV2A VT" synonymously, whereas in the figures they only use "SV2A VT" / "VT". Consider using "SV2A VT" throughout, or explicitly stating that alterations in SV2A VT indicate impaired SV2A density. The discussion is interesting and informative but rather long and not very focussed. Consider shortening the text on page 10 (inflammation, HPA axis, mTORC1) to topics directly related to the specific methods and results of the present investigation. The authors report a comparison of SV2A VT ACC-HIP correlation between subjects with high and low depressive symptom load on page 7, lines 160-165 (figure 4), and discuss it on page 9, lines 217-228. This analysis is refreshing, informative, and deserves more highlighting. Analyses on the cross-regional signal pattern in PET studies are highly underrepresented in spite of their potential to open new perspectives -they deserve more attention and appreciation.

Reviewer comments, author responses and manuscript changes (NCOMMS-18-32857-T)
We are grateful to each reviewer for the attentive reading of our work and their constructive comments. The comments are encouraging and all reviewers agree that the presented research is exciting and novel. Below is a point by point response to the reviewers' comments. We have carried out additional data analyses and updated the manuscript according to the reviewers' comments, and believe this makes for an improved manuscript.

Reviewer #1
This is a very interesting study using an indirect measure of synaptic density in patients with MDD and PTSD. The translational relevance of the study is high and the experiment was carried out by an established group of investigators. The investigators take an interesting transdiagnostic approach, examining the SV2A marker in MDD, PTSD and comorbid, as well as HC. The sample sizes are small with "the clinical group" comprised of 11 individuals with MDD, 5 with PTSD and 10 with comorbid MDD and PTSD, which is a limitation.
The main finding is that SV2A density was negatively correlated with severity of depressive symptoms clinical subjects in dlPFC (r=-0.633, p=0.001), ACC (r=-0.634, p=0.001) and hippocampus (r=-0.487, p=0.012). As noted below, it was not immediately clear to me this these findings are the results of a prior hypothesis testing or if the study is meant to be exploratory/hypothesis generating. This is an important point in the context of rigor. Either approach is reasonable, but it is incumbent upon the authors to present the analyses in light of one of the two approaches, I believe. The supportive analyses related to group effects of fMRI are interesting and appropriately presented as exploratory. Overall this is a very interesting paper with some potential limitation related to sample size and lack of clarity about pre-specified features of the statistical plan.
I have a few specific comments below.

Comment 1:
Intro -maybe a minor quibble, but the paper starts with, "the human brain displays remarkable neuroplasticity in response to stress...Persisting high levels of stress or repeated exposure to extreme stress results in loss of synapses in circuits underlying affective and cognitive processes1." But the data the authors are referring to is in rodent models, and I do not know if we know what happens re synaptic plasticity in humans under stress.

Response:
The reference cited (Duman et al. 2016 1 ) refers to both clinical and preclinical studies. It is well established that the human brain demonstrates neuroplastic changes in response to stress. For example, imaging studies show stress-induced functional changes in stress paradigms, and both functional and structural changes have been consistently shown in chronic stress disorders such as MDD and PTSD. We acknowledge that our statement that persisting levels of stress results in loss of synapses is, whilst directly supported by preclinical work, only indirectly supported by human studies (e.g. post-mortem studies in MDD). We have therefore changed the second sentence to "persisting high levels of stress are thought to result in loss of synapses in circuits… ". Comment 2: Intro, last paragraph, no specific hypotheses are stated. Authors refer to an exploratory approach, which is OK but should be consistent throughout. For example in the Results there is finding presented as primary of correlations between SV2A and a few select brain regions (where they defined a prior to data collection?) Then there is what is presented in the results as 'exploratory' but it is unclear if the authors are using these terms in the technical/statistical sense or just general language to indicate what they feel is most important. Since these terms do have technical meanings in this context, I suggest using the terms in that way.

Response:
We have restructured the introduction to clarify that investigating the relationship between depression severity and synaptic density in a transdiagnostic sample was the primary focus of the study. We have also added our primary hypothesis to the end of the introduction: "Our primary hypothesis was that synaptic density, as measured by [ 11 C]UCB-J PET, would be negatively associated with severity of depressive symptoms transdiagnostically".
The regions were selected a priori based on existing research -the introduction outlines this research and justifies our a priori choice of regions. We also state this in the last paragraph of the introduction.
We have now clarified the primary and secondary analyses (see the data analysis section of the methods -pg. 19). Primary analyses included the transdiagnostic association between depression severity and synaptic density; and the group differences in synaptic density. Secondary analyses included the MRI results and the associations between synaptic density, cognitive function and the symptom clusters. Findings from the primary analyses were corrected for multiple comparisons. The secondary analyses relating to cognitive function and symptom clusters were not corrected for multiple comparisons, which is stated in the methods. It is important to note that although exploratory, the fMRI data were processed according to current best practices for multiple comparisons correction due to the high risk of type 1 error in fMRI research 2 .

Comment 3:
Unclear if the p values presented in results section (e.g., for correlations between SVA and severity in 3 different regions, Fig 1) are corrected for multiple comparisons.
Response: These correlations survived correction for multiple comparisons, which is stated in the results (pg. 5).

Comment 4:
There is no explicit statement of hypothesis testing, sample size determination and the like. It is not explicitly stated if the hypotheses and the analytic plan for this experiment as presented was formulated before the data was collected. This is important in evaluating the results and in particular the risk of type II error. It is also important for general transparency. It would be helpful if the accompanying research protocol were available.

Response:
We have added our primary hypothesis to the introduction and have outlined the primary and secondary analyses in the methods. The plan to adopt a transdiagnostic approach was formulated prior to data collection, and investigating the transdiagnostic association between synaptic density and depression severity was the primary aim of the study. Given that this is the first study of its kind there was no existing data on synaptic density and depression severity, and a precise power calculation was not possible. However, with anticipated large effect sizes based on pilot data and preclinical data on synaptic loss in models of depression, we determined that a clinical sample size of 24 would be sufficiently powered to detect a statistically significant (i.e., p<0.05) association between synaptic density and depressive symptoms. The protocol is not available for distribution as it incorporates several other studies and other confidential information.

Reviewer #2
This is an impressive study by a world-class imaging group. This is very novel work but there are concerns relevant to data interpretation, although it is important to note that the authors did well to address some key concerns through Supplemental data and the Discussion section. Example concerns: Comment 1: Need for further consideration of why it may be and/or why it might not be appropriate to combine MDD, PTSD and MDD+PTSD in this first study; Response: Consistent with emerging theoretical frameworks in psychiatry (e.g., NIMH Research Domain Criteria Project) we intentionally employed a transdiagnostic, dimensional approach instead of a categorical approach. Diagnostic categories are inherently heterogenous; a dimensional approach aims to understand the neurobiology underlying symptoms, which may lead to the discovery of targets and novel treatments that are effective for specific symptoms, in this case, depression. Indeed, depressive symptoms are common in PTSD, which is often comorbid with MDD, yet current antidepressants are often ineffective. We believe our transdiagnostic approach is a strength of this study, and that demonstrating an association between SV2A V T , which is indicative of synaptic density, and depressive symptoms in a transdiagnostic sample of individuals with MDD and/or PTSD in a first study highlights the potential importance of synaptic plasticity in relation to depressive symptoms specifically. However, we acknowledge that larger PET studies investigating synaptic density in distinct categorical disorders are vital in further unravelling the role of synaptic density in relation to psychiatric symptoms, and have specifically addressed this in the Discussion.

Comment 2:
Limited samples sizes that include the number of "pure" PTSD (n=5), sizes of the LS and HS groups, and only a subset of HC with fMRI; Response: We acknowledge that the number of individuals with 'pure' PTSD is relatively small. However, given the transdiagnostic approach, we were able to detect a significant association between synaptic density and symptoms of depression. We would also like to highlight that 26 individuals with psychiatric illness (and 21 healthy controls; 47 subjects in total) is relatively large for a PET study. Further, all 26 individuals were medication-free, which is rare to see in a PET study of such size. Given that the study is sufficiently powered to detect a significant association between synaptic density and depressive symptoms transdiagnostically and that this is the first study of its kind, we believe that despite the relatively small samples of each diagnostic category alone, these findings add important and novel insight into the neurobiological mechanisms underlying symptoms of depression -symptoms that span multiple psychiatric and neurological disorders that are often inadequately treated. Saying this, as stated previously, we do acknowledge that larger PET studies investigating synaptic density in discrete categorical disorders are necessary to determine whether there are distinct synaptic density patterns across disorders, and we have highlighted this in the Discussion.
In relation to not having fMRI data in the full HC sample, we have highlighted this in the Discussion. Furthermore, we took several steps to improve power. For example, a ROI-based approach based on the PET findings was used to examine global connectivity in the dlPFC -indeed this finding replicates previous studies showing abnormal PFC connectivity in MDD and PTSD. Further, we combined PET and fMRI modalities to explore a relationship between lower synaptic density and a 'blurring' of 2 typically opposed networks in the clinical sample, which provides novel insight into the possible molecular underpinnings of network dysfunction.

Comment 3: Significantly different unlabelled mass doses administered across groups;
Response: Based on our previously reported validation study 3 (https://journals.sagepub.com/doi/abs/10.1177/0271678X17724947), the mass doses in this study were determined to occupy <1% of receptors. Further, there was no observed correlation between injected mass and V T . We are therefore confident that the different mass doses have not substantially impacted our results. It must also be noted that any effect of a lower mass dose in high severity clinical group would have resulted in an underestimation of our observed group difference. This is because lower mass is associated with higher specific activity . Therefore, any effect of lower mass would have resulted in higher V T and thus would have minimized the difference between high severity patients and controls. We have addressed this issue in the Discussion (pg.12).

Comment 4:
No correction for non-displaceable distribution volume of radiotracer Response: Given the observed difference in centrum semiovale V T, and that MDD and PTSD have been associated with white matter pathology, we determined that V T was the more appropriate outcome measure over BP ND . V T has been previously validated as a reliable and accurate outcome measure for assessing synaptic density using [ 11 C]UCB-J 3,4 . Further, the non-displaceable uptake is very small compared to the grey matter uptake, especially after PVC and so nearly all of the grey matter V T value represents specific binding. We address this in the Discussion (pg. 11).
Response: Thank you for pointing this out -we have changed all axis labels to [ 11 C]UCB-J V T and use [ 11 C]UCB-J V T consistently throughout the results. We explicitly state that [ 11 C]UCB-J V T is indicative of SV2A (synaptic) density, as validated previously 3,4 .

Comment 6:
In reference 53 (Fig 1D), it appears that the correlation between SV2A VT and SV2A OD is r2=0.31. How might this translate to the present work -is this lower correlation a concern?
Response: The relatively low correlation between SV2A OD and SV2A V T for the GM was due to the relative low difference in SV2A signal across the GM regions. Further, we'd like to highlight that Western blot quantification is semi-quantitative and the relatively small sample size for V T (9 GM regions in one baboon). Considering the low sample size, it is important to also consider the high correlation between SV2A OD and SYN OD (direct comparison with the same method) and the good correlation between V T and the SV2A density (B max ) from the homogenate binding assay (a more accurate assay for SV2A). Further, blockade of [ 11 C]UCB-J binding by levetiracetam confirms SV2A density in vivo. We are thus comfortable that V T relates to SV2A density and that SV2A and SYN expression was highly correlated across the brain regions. Comment 7: Did the authors observe any input function differences across the groups, e.g., did in vivo radiotracer metabolism differ across groups?

Response:
We observed no between-group differences in plasma parent fraction at either 30 or 60 minutes post-injection. Further, there was no difference in the metabolite-corrected input function (40-60 minutes post-injection) across groups. We have added this to the manuscript and now report these data in SI (Table S4; shown below). Note that if there was a group difference in the tracer availability, the use of kinetic modelling would have provided a correction for this effect. Response: It is unlikely that gray matter atrophy could account for the ICD findings. As described in the response below, we did not detect any group differences in gray matter volume for the a priori ROIs. In our two previous papers using ICD in MDD and PTSD, we showed that regions of group differences in ICD were not the same regions as group differences in gray matter volume 5,6 . Further, we have shown that the correlation between gray matter volume and ICD is not significantly correlated across the gray matter in MDD 5 . Finally, ICD involves several steps to account for potential confounds related to brain size, include high resolution non-linear registrations, and within subject normalization of connectivity values.
Comment 9: Were regional gray matter volumes compared across groups or considered in the statistical analysis? Response: Regional gray matter volumes were calculated using tensor-based morphometry (TBM) as in our previous studies 5, 6 . The determinant of the Jacobian of the deformation field generated from a high resolution non-linear transformation into MNI common space was used to quantify local volume differences between the registered images and the template 7 . This metric provided an estimate of voxel-wise volume changes for all transformed images with respect to the group averaged template and was used for further analysis. As shown in Response Table 1, below, there were no differences in gray matter volume for these regions, suggesting that there were no differences in gray matter volume for the three regions of interest. As such, we did not consider gray matter volume in any further statistical analysis. Comment 10: The authors should describe UCB-J SV2A VT and VT/fp test-retest variability. Which outcome is considered more accurate or reliable, VT or VT/fp? Response: Our previous paper 3 describes test-retest variability of [ 11 C]UCB-J V T and V T /f p . The testretest reproducibility of [ 11 C]UCB-J V T was exceptionally good, TRV characteristics were virtually identical for ROI and voxel-based analyses. For voxel-based analysis, the regional mean TRV and aTRV for V T were -0.7±5.2% and 4.2±2. 6. Correction for f p worsened the TRV and ICC. Indeed the ICC was above 0.6 (the commonly used as a threshold) for V T /f p , but <0.6 for V T. Therefore, we chose V T as our primary outcome measure, but report V T /f p in SI for completeness. We have added a statement on test-retest variability of these outcomes to the methods section (under 'PET image analysis').
Comment 11: Has outcome reliability been assessed yet in the relevant patient groups?
Response: Although test-retest studies are not typically performed in psychiatric populations, we conducted repeat scans in two adults with MDD and observed test-retest reliability similar to