Abstract

Introduction Understanding the emotional responsivity style and neurocognitive profiles of depression-related processes in at-risk youth may be helpful in revealing those most likely to develop affective disorders. However, the multiplicity of biopsychosocial risk factors makes it difficult to disentangle unique and combined effects at a neurobiological level.Methods In a population-derived sample of 56 older adolescents (aged 17-20), we adopted partial least squares regression and correlation models to explore the relationships between multivariate biopsychosocial risks for later depression, emotional response style, and fMRI activity, to rejecting and inclusive social feedback.Results Behaviorally, higher depressive risk was associated with both reduced negative affect following negative social feedback and reduced positive affect following positive social feedback. In response to both cues of rejection and inclusion, we observed a general neural pattern of increased cingulate, temporal, and striatal activity in the brain. Secondly, in response to rejection only, we observed a pattern of activity in ostensibly executive control- and emotion regulation-related brain regions encompassing fronto-parietal brain networks including the angular gyrus.Conclusion The results suggest that risk for depression is associated with a pervasive emotional insensitivity in the face of positive and negative social feedback.


Authors

Stretton, Jason;  Walsh, Nicholas D.;  Mobbs, Dean;  Schweizer, Susanne;  van Harmelen, Anne-Laura;  Lombardo, Michael;  Goodyer, Ian;  Dalgleish, Tim

Publons users who've claimed - I am an author

No Publons users have claimed this paper.

Contributors on Publons
  • 1 reviewer
  • pre-publication peer review (FINAL ROUND)
    Decision Letter
    2020/11/28

    28-Nov-2020

    Dear Dr. Stretton:

    It is a pleasure to accept your manuscript entitled "How biopsychosocial depressive risk shapes behavioral and neural responses to social evaluation in adolescence." in its current form for publication in Brain and Behavior. If there were further comments from the reviewer(s) who read your manuscript, they will be included at the foot of this letter.

    Please note although the manuscript is accepted the files will now be checked to ensure that everything is ready for publication, and you may be contacted if final versions of files for publication are required.

    The final version of your article cannot be published until the publisher has received the appropriate signed license agreement. Once your article has been received by Wiley for production the corresponding author will receive an email from Wiley’s Author Services system which will ask them to log in and will present them with the appropriate license for completion.

    Payment of your Open Access Article Publication Charge (APC):

    All articles published in Brain and Behavior are fully open access: immediately and freely available to read, download and share. Brain and Behavior charges an article publication charge (APC).

    Before we can publish your article, your payment must be completed. The corresponding author for this manuscript will have already received a quote email shortly after original submission with the estimated Article Publication Charge; please let us know if this has not been received. Once your accepted paper is in production, the corresponding author will receive an e-mail inviting them to register with or log in to Wiley Author Services (www.wileyauthors.com) where the publication fee can be paid by credit card, or an invoice or proforma can be requested. The option to pay via credit card and claim reimbursement from your institution may help to avoid delays with payment processing.

    If your paper contains Supporting Information:
    Materials submitted as Supporting Information are authorized for publication alongside the online version of the accepted paper. No further Supporting Information can be submitted after acceptance. It is the responsibility of the authors to supply any necessary permissions to the editorial office.

    Thank you for your fine contribution. On behalf of the Editors of Brain and Behavior, we look forward to your continued contributions to the Journal.

    Sincerely,
    Dr. Amanda Bischoff-Grethe
    Editor in Chief, Brain and Behavior
    agrethe@ucsd.edu

    Associate Editor Comments to Author:

    Reviewer(s)' Comments to Author:

    P.S.- Would you be interested in publishing your proven experimental method as a detailed step-by-step protocol? Current Protocols in Neuroscience welcomes proposals from prospective authors to disseminate their experimental methodology in the rapidly evolving field of neuroscience. Please submit your proposal here: https://currentprotocols.onlinelibrary.wiley.com/hub/submitaproposal

    P.S. – You can help your research get the attention it deserves! Wiley Editing Services offers professional video abstract and infographic creation to help you promote your research at www.wileyauthors.com/eeo/promotion. And, check out Wiley’s free Promotion Guide for best-practice recommendations for promoting your work at www.wileyauthors.com/eeo/guide.

    This journal accepts artwork submissions for Cover Images. This is an optional service you can use to help increase article exposure and showcase your research. For more information, including artwork guidelines, pricing, and submission details, please visit the Journal Cover Image page at www.wileyauthors.com/eeo/covers.

    Decision letter by
    Cite this decision letter
    Author Response
    2020/11/27

    Reviewer(s)' Comments to Author: Reviewer: 2 Comments to the Author The author has responded to all of my comments regarding the intro and methods/analysis. The details added to the methods helped me better assess and interpret this work. I have two minor comments regarding the discussion that I hope can be addressed prior to publication. 1. In their response letter, the authors note, "The lack of uniqueness associated to any single variable reinforces our view that depressive risk is not related to any one single variable, but is formed via the interplay between a constellation of biological, psychological and social factors." I think this is an important point to consider adding to the discussion. Thank you for your comment. We agree this is an important point and have included it in our main discussion (see page 15, paragraph 2); ‘In a previous study assessing cognitive reappraisal of emotion in the same sample of adolescents, we showed those with a history of CA (relative to those without) had an enhanced capacity to downregulate both positive and negative affect (Schweizer et al., 2016). We interpreted this as the CA environment serving as a practice ground to hone explicit emotion regulation skills. In the current study a higher risk of depression predicted lower positive and negative affect following social feedback, in the absence of any emotion regulation instructions. The lack of uniqueness associated to any single variable reinforces our view that depressive risk is not related to any one single variable, but is formed via the interplay between a constellation of biological, psychological and social factors. The current results therefore extend these previous findings and support a more general notion of emotion attenuation associated with biopsychosocial risk.’ 2. I am still having some trouble with the interpretation of neural findings for the Positive + Negative vs. Neutral conjunction analysis. Referencing the Morgan et al. (2016) paper and frontostriatal reactivity seems unintentionally misleading here, as no connectivity analyses were performed and no regions in the prefrontal cortex came out for the Positive + Negative vs. Neutral conjunction analysis. Thus, I'm not sure how findings would reflect an attenuated frontostriatal response. Referring to posterior cingulate and temporal regions as "emotion-regulation regions" (pg. 17) also seems a little misleading in the context of striatal findings - if there is evidence to suggest that these regions work with the caudate to support emotion regulation capabilities, please cite this. Further, it doesn't seem necessary to reference hyperactivity of the striatum to monetary reward anticipation as support for present findings, as social evaluation feedback is quite different from monetary reward anticipation. I'm wondering if there are other possible interpretations of heightened striatal/temporal/PCC activity to social evaluation, potentially related to self-referential processing or even to resilience (Dennison et al., 2016 showed that adolescents with a history of maltreatment and high striatal activation to rewards reported lower depressive symptoms)? Authors could consider alternative interpretations and/or should consider tempering current interpretations given lack of consistent evidence in this field and the novelty of this work. Thank you for your comment. We have removed the reference to Morgan et al. (2016) as we would not want to unintentionally mislead the reader into thinking we ran any connectivity analyses. The posterior cingulate and temporal regions are involved in regulation of emotions, specifically there is evidence to show activity in the posterior cingulate and striatum increases and decreases in response to up- and down-regulation of socially driven emotions (Grecucci et al., 2013). Thus an alternative explanation could be that increased activation of these regions in response social evaluation represents the downstream effects of the enhanced emotion regulation capabilities of the sample (Schweizer et al., 2016). This would be in line with the Dennison et al., study (2016) and in line with an interpretation favouring resilient mechanisms. We have amended the discussion accordingly on Pages 16-17, to read the following; “Research on brain reward-region responsivity in association with biopsychosocial risk for depression has been mixed. Reward processing in those who have experienced early adversity has been accompanied by reduced activation of the ventral striatum in a number of studies (Goff et al., 2013; Hanson et al., 2016; Mehta et al., 2010), and this has been interpreted as an adaptive avoidant response during approach-avoidance conflict situations (Teicher & Samson, 2016) conferring long-term risk. In contrast, and in line with the present findings of augmented activity in reward-related brain networks as a function of risk, Dennison and colleagues (Dennison et al., 2016) reported increased striatal response, specifically in the left nucleus accumbens and putamen, while passively viewing positive relative to neutral social stimuli in a group of maltreated older adolescents relative to a control group with no history of maltreatment. Further, in a longitudinal community-based study of adolescent girls, low parental warmth - a risk factor for subsequent MDD (Hipwell et al., 2008) - measured at age 11 was associated with increased striatal activity during reward anticipation measured at age 16 (Casement et al., 2014). Finally, the posterior cingulate and striatum show increases and decreases in response to up- and down-regulation of socially driven emotions during neuroeconomic strategy games (Grecucci et al., 2013). The increased activation of these regions in response social evaluation in our study could therefore represent the downstream effects of the enhanced emotion regulation capabilities of the sample (Schweizer et al., 2016), potentially reflecting a putative resilience mechanism to social evaluation.”



    Cite this author response
  • pre-publication peer review (ROUND 2)
    Decision Letter
    2020/11/11

    11-Nov-2020

    We recognise that the impact of the COVID-19 pandemic may affect your ability to return your revised manuscript to us within the requested timeframe. If this is the case, please let us know.

    Dear Dr. Stretton:

    Manuscript ID BRB3-2020-03-0259.R1 entitled "How biopsychosocial depressive risk shapes behavioral and neural responses to social evaluation in adolescence." which you submitted to Brain and Behavior, has been reviewed very favorably and minor revisions have been requested. I invite you to respond to the comments appended below and revise your manuscript.

    Before submitting your revisions:

    1. Prepare a response to the reviewer comments appended below in point-by-point fashion. In order to expedite the processing of the revised manuscript, please be as specific as possible in your response and indicate the page numbers in the manuscript where you have addressed each comment.

    2. Prepare a revised manuscript (word document), highlighting the changes you’ve made. Save this new document on your computer as you will be asked to upload it during the revision submission process. NOTE: Please be sure to keep in mind reviewer comments and incorporate your responses within the manuscript. There may well be areas where you disagree; for example, you may want to write, "A reviewer suggests that... However, I disagree because...". In any case, please try to address all of the concerns that are raised within the manuscript.

    3. In addition to your revised manuscript with changes highlighted, please also save a “clean” copy where the changes are not marked.

    To submit your revised manuscript:

    1. Log in by clicking on the link below

    PLEASE NOTE: This is a two-step process. After clicking on the link, you will be directed to a webpage to confirm.

    https://mc.manuscriptcentral.com/brainandbehavior?URL_MASK=e76d740e47154056b2a8e7fba7d4bbbe

    OR

    Log into https://mc.manuscriptcentral.com/brainandbehavior and click on Author Center. Under author resources, use the button “Click here to submit a revision”. PLEASE DO NOT SUBMIT YOUR REVISIONS AS A NEW MANUSCRIPT.

    1. Follow the on-screen instructions. First you will be asked to provide your “Response to Decision Letter”—this is the response to reviewer comments that you prepared earlier.

    2. Click through the next few screens to verify that all previously provided information is correct.

    3. File Upload: Delete any files that you will be replacing (this includes your old manuscript). Upload your new revised manuscript file with changes highlighted, a “clean” copy of your revised manuscript file, any replacement figures/tables, or any new files. Once this is complete, the list of files in the “My Files” section should ONLY contain the final versions of everything. REMEMBER: figures/tables should be in jpeg, tiff, or eps format. We hope that you will designate one of the figures in your paper to be considered for our online covers and potential publication on our blog.

    4. Review and submit: please be sure to double-check everything carefully so that your manuscript can be processed as quickly as possible.

    Deadlines:
    Because we are trying to facilitate timely publication of manuscripts submitted to Brain and Behavior,your revised manuscript should be uploaded as soon as possible. If it is not possible for you to submit your revision in 2 months, we may have to consider your paper as a new submission. If you feel that you will be unable to submit your revision within the time allowed please contact me to discuss the possibility of extending the revision time.

    Wiley Editing Services Available to All Authors
    Should you be interested, Wiley Editing Services offers expert help with manuscript, language, and format editing, along with other article preparation services. You can learn more about this service option at www.wileyauthors.com/eeo/preparation. You can also check out Wiley’s collection of free article preparation resources for general guidance about writing and preparing your manuscript at www.wileyauthors.com/eeo/prepresources.

    Once again, thank you for submitting your manuscript to Brain and Behavior and I look forward to receiving your revision.

    Sincerely,
    Dr. Amanda Bischoff-Grethe
    Editor in Chief, Brain and Behavior
    agrethe@ucsd.edu

    Associate Editor Comments to Author:

    Reviewer(s)' Comments to Author:

    Reviewer: 2

    Comments to the Author
    The author has responded to all of my comments regarding the intro and methods/analysis. The details added to the methods helped me better assess and interpret this work. I have two minor comments regarding the discussion that I hope can be addressed prior to publication.

    1. In their response letter, the authors note, "The lack of uniqueness associated to any single variable reinforces our view that depressive risk is not related to any one single variable, but is formed via the interplay between a constellation of biological, psychological and social factors." I think this is an important point to consider adding to the discussion.

    2. I am still having some trouble with the interpretation of neural findings for the Positive + Negative vs. Neutral conjunction analysis. Referencing the Morgan et al. (2016) paper and frontostriatal reactivity seems unintentionally misleading here, as no connectivity analyses were performed and no regions in the prefrontal cortex came out for the Positive + Negative vs. Neutral conjunction analysis. Thus, I'm not sure how findings would reflect an attenuated frontostriatal response. Referring to posterior cingulate and temporal regions as "emotion-regulation regions" (pg. 17) also seems a little misleading in the context of striatal findings - if there is evidence to suggest that these regions work with the caudate to support emotion regulation capabilities, please cite this. Further, it doesn't seem necessary to reference hyperactivity of the striatum to monetary reward anticipation as support for present findings, as social evaluation feedback is quite different from monetary reward anticipation. I'm wondering if there are other possible interpretations of heightened striatal/temporal/PCC activity to social evaluation, potentially related to self-referential processing or even to resilience (Dennison et al., 2016 showed that adolescents with a history of maltreatment and high striatal activation to rewards reported lower depressive symptoms)? Authors could consider alternative interpretations and/or should consider tempering current interpretations given lack of consistent evidence in this field and the novelty of this work.

    Reviewer: 1

    Comments to the Author
    I think the manuscript is improved and suitable for publication. My main remaining concern is the length of the manuscript itself, but I let the editor see to that.

    Decision letter by
    Cite this decision letter
    Reviewer report
    2020/11/08

    I think the manuscript is improved and suitable for publication. My main remaining concern is the length of the manuscript itself, but I let the editor see to that.

    Reviewed by
    Cite this review
    Reviewer report
    2020/11/02

    The author has responded to all of my comments regarding the intro and methods/analysis. The details added to the methods helped me better assess and interpret this work. I have two minor comments regarding the discussion that I hope can be addressed prior to publication.

    1. In their response letter, the authors note, "The lack of uniqueness associated to any single variable reinforces our view that depressive risk is not related to any one single variable, but is formed via the interplay between a constellation of biological, psychological and social factors." I think this is an important point to consider adding to the discussion.

    2. I am still having some trouble with the interpretation of neural findings for the Positive + Negative vs. Neutral conjunction analysis. Referencing the Morgan et al. (2016) paper and frontostriatal reactivity seems unintentionally misleading here, as no connectivity analyses were performed and no regions in the prefrontal cortex came out for the Positive + Negative vs. Neutral conjunction analysis. Thus, I'm not sure how findings would reflect an attenuated frontostriatal response. Referring to posterior cingulate and temporal regions as "emotion-regulation regions" (pg. 17) also seems a little misleading in the context of striatal findings - if there is evidence to suggest that these regions work with the caudate to support emotion regulation capabilities, please cite this. Further, it doesn't seem necessary to reference hyperactivity of the striatum to monetary reward anticipation as support for present findings, as social evaluation feedback is quite different from monetary reward anticipation. I'm wondering if there are other possible interpretations of heightened striatal/temporal/PCC activity to social evaluation, potentially related to self-referential processing or even to resilience (Dennison et al., 2016 showed that adolescents with a history of maltreatment and high striatal activation to rewards reported lower depressive symptoms)? Authors could consider alternative interpretations and/or should consider tempering current interpretations given lack of consistent evidence in this field and the novelty of this work.

    Reviewed by
    Cite this review
    Author Response
    2020/10/07

    23-Jul-2020 We recognise that the impact of the COVID-19 pandemic may affect your ability to return your revised manuscript to us within the requested timeframe. If this is the case, please let us know. Dear Dr. Stretton: Manuscript ID BRB3-2020-03-0259 entitled "How biopsychosocial depressive risk shapes behavioral and neural responses to social evaluation in adolescence." which you submitted to Brain and Behavior, has been reviewed. Some revisions to your manuscript have been recommended. Therefore, I invite you to respond to the comments appended below and revise your manuscript. Before submitting your revisions: 1. Prepare a response to the reviewer comments appended below in point-by-point fashion. In order to expedite the processing of the revised manuscript, please be as specific as possible in your response and indicate the page numbers in the manuscript where you have addressed each comment. 2. Prepare a revised manuscript (word document), highlighting the changes you’ve made. Save this new document on your computer as you will be asked to upload it during the revision submission process. NOTE: Please be sure to keep in mind reviewer comments and incorporate your responses within the manuscript. There may well be areas where you disagree; for example, you may want to write, "A reviewer suggests that... However, I disagree because...". In any case, please try to address all of the concerns that are raised within the manuscript. 3. In addition to your revised manuscript with changes highlighted, please also save a “clean” copy where the changes are not marked. To submit your revised manuscript: 1. Log in by clicking on the link below PLEASE NOTE: This is a two-step process. After clicking on the link, you will be directed to a webpage to confirm. https://mc.manuscriptcentral.com/brainandbehavior?URL\_MASK=326884292fdd4f159708f0b6dae4edfa OR Log into https://mc.manuscriptcentral.com/brainandbehavior and click on Author Center. Under author resources, use the button “Click here to submit a revision”. PLEASE DO NOT SUBMIT YOUR REVISIONS AS A NEW MANUSCRIPT. 2. Follow the on-screen instructions. First you will be asked to provide your “Response to Decision Letter”—this is the response to reviewer comments that you prepared earlier. 3. Click through the next few screens to verify that all previously provided information is correct. 4. File Upload: Delete any files that you will be replacing (this includes your old manuscript). Upload your new revised manuscript file with changes highlighted, a “clean” copy of your revised manuscript file, any replacement figures/tables, or any new files. Once this is complete, the list of files in the “My Files” section should ONLY contain the final versions of everything. REMEMBER: figures/tables should be in jpeg, tiff, or eps format. We hope that you will designate one of the figures in your paper to be considered for our online covers and potential publication on our blog. 5. Review and submit: please be sure to double-check everything carefully so that your manuscript can be processed as quickly as possible. Deadlines: Because we are trying to facilitate timely publication of manuscripts submitted to Brain and Behavior,your revised manuscript should be uploaded as soon as possible. If it is not possible for you to submit your revision in 2 months, we may have to consider your paper as a new submission. If you feel that you will be unable to submit your revision within the time allowed please contact me to discuss the possibility of extending the revision time. If you would like help with English language editing, or other article preparation support, Wiley Editing Services offers expert help with English Language Editing, as well as translation, manuscript formatting, and figure formatting at www.wileyauthors.com/eeo/preparation. You can also check out our resources for Preparing Your Article for general guidance about writing and preparing your manuscript at www.wileyauthors.com/eeo/prepresources. Once again, thank you for submitting your manuscript to Brain and Behavior and I look forward to receiving your revision. Sincerely, Dr. Amanda Bischoff-Grethe Editor in Chief, Brain and Behavior agrethe@ucsd.edu Associate Editor Comments to Author: Reviewer(s)' Comments to Author: Reviewer: 1 Comments to the Author Review BRB3-2020-03-0259 This study focuses on underlying risk markers for affective disorders, with the aim to identify through partial least square latent components that best describe the relationship between various “risk factors” and behavioral or fMRI data. They find on 56 late adolescents an association between some risk factors, mainly childhood adversity, and decreased negative affect after social rejection, decreased positive affect after social inclusion, and increased activity during social feedback in a network comprising ventral striatum, PCC, mid cingulate cortex, and temporal gyrus. Authors interpret these findings in the context of the Emotion Context Insensitivity model. I think the topic is very timely and the approach interesting. I support a dimensional approach using multivariate data-driven statistical technique. However I have some reservations on how these data were used and on the manuscript. It is very ambitious to cover multiple and diverse risk factors and how they predict clinical AND neuroimaging data, so I am afraid the sample does not allow for sufficient statistical power. Thank you for your comments. We agree that our sample size could be bigger, however we are confident that the PLS techniques applied in the manuscript are appropriate for this reason. PLS regressions and correlations are well suited to data with multiple variables and smaller samples, and we directly reference this in the methods section on pages 21-22 (Wold, Ruhe, Wold, & Dunn, 1984). Furthermore, the robustness of our neuroimaging analysis with 10,000 permutation tests and 10,000 bootstraps lend confidence to the validity of the brain-behaviour relationships we observed. Only patterns of voxels that survived such stringent measures have been included in the manuscript, with a pseudo Z statistic of (+/-) 3 equating to Family-Wise Error correction for multiple comparisons. In response to your comment we have now included a sentence in our limitations highlighting that an increased sample size would have been welcome, and that further studies in larger samples would be appropriate to validate our results (page 18, paragraph 2); ‘In addition, it should also be noted that we have a relatively small sample size (N=56) to assess depressive risk across multiple parameters. This is however, why we opted to use the PLS method which has been validated for use in data such as this (Wold, Ruhe, Wold, & Dunn, 1984). Furthermore, the robustness of our neuroimaging analysis with 10,000 permutation tests and 10,000 bootstraps lend confidence to the validity of the brain-behaviour relationships we observed’ I am also skeptical about putting BDI at T3 as a “risk factor”, since it is more a clinical description of psychopathology at T3, could be an outcome. Thank you for your comment. We believe that the BDI score used in the two weeks prior to scanning does qualify as a risk factor to clinical diagnosis of Major Depressive Disorder. The stated purpose of the BDI-II is not to diagnose major depressive episode and as such is a measure of symptoms. Furthermore and crucial to our stated aims, the PLS method looks for dimensions with high covariance between the behavioural variables. The focus of this approach is to identify a latent risk factor based on all of the parameters taken together, rather than trying to isolate the contribution of an independent risk factor in relation to the neuroimaging data. This was consolidated by the fact that there were no univariate activations associated with BDI (or any other unitary risk factor) in any of the analyses. The global interpretation is misleading : while some “risk factor” are associated with results coherent with a certain model of affective disorders, they don’t actually predict such disorders, and it is not a longitudinal study in the sense that variables from T1, T2, and T3 are mixed together. Thank you for your comment. We agree that ‘predict’ is too strong and rather, that risk factors increase the likelihood of developing depression. We have adapted the Abstract (Page 2) accordingly and changed ‘predicting’ to ‘revealing’. With regard to the longitudinal nature of the study, you are correct that the variables considered in the analysis are not treated longitudinally in the analysis. However, we do still feel our study qualifies as a longitudinal approach as data were collected at separate time points over several years. With this in mind, we have removed the longitudinal description from the Introduction (Page 4) as we do not want to mischaracterize our study, but kept reference to the longitudinal nature of the data collection in the methods section. The PLS results for neuroimaging data are not very strong (30% variance explained) and not very straightforward (different variables with different patterns). Maybe trying a model with less variables but better selected could be helpful. Thank you for your comment. Within the limited neuroimaging PLS regression literature, ~30% can be considered quite high (see Romero-Garcia, R., Warrier, V., Bullmore, E.T. et al. Synaptic and transcriptionally downregulated genes are associated with cortical thickness differences in autism. Mol Psychiatry 24, 1053–1064 (2019) for examples of data with just 10% variance explained). Further, our permutation tests (10,000) showed that the brain-behaviour relationship was significantly above chance in both analyses (p’s = 0.0003 & 0.037). A few additional comments: Abstract is not very clear, could be more precise on which population exactly and which inputs to PLS, Emotion Context Insensitivity model best kept for discussion I suggest. Thank you for your comment. We have removed reference to the Emotion Context Insensitivity model in the Abstract and have clarified that we used a population-derived sample of older adolescents. Introduction : In general, the introduction references too many concepts, some not necessary in my opinion. For example animal studies and stress reactivity: neither very relevant nor exhaustive; same apply for remitted MDD studies : what is the link with the study presented here? Thank you for your comment. We have removed references to the animal studies and stress reactivity (Page 8) as we agree that the link to the present study is not clear. The logical link between Emotion Context Insensitivity and reward processing is also missing. I think the introduction should be refocused on the current hypothesis. Thank you for your comment. We have included a reference to the link between emotional disengagement and reward processing to account for this oversight. See Page 8, the last line of paragraph 1 reads; ‘This is in line with event-related potential (ERP) data which consistently shows diminished activity to the processing of motivationally salient stimuli and to the receipt of reward, suggesting depression is associated with emotional disengagement and deficits in reward processing (Proudfit et al., 2015).’ Hypothesis : not clear, especially in the direction of expected changes in activity. Thank you for your comment. Our neural hypothesis was to expect a change in activity, but owing to the existing differential literature and the predictions of the ECI and positive attenuation models, we intentionally left directionality out of the hypothesis. Methods : Why exclusion of DSM IV axis I disorder ? Even is exploratory analyses later checking the if the results are different for the group with disorders. Thank you for your comment. The inclusion of participants with a current DSM IV axis I disorder would have confounded our results as we were specifically trying to elucidate psychiatric risk. While any subsequent analysis between participants with and without a current diagnosis is of interest, it is beyond the scope of the data. We did however run a sensitivity analysis for previous psychiatric history (see Methods section on Page 25) as it was important to test that any resulting brain-behavior pairs were not specific to prior mental health difficulties. How was CA measured ? Thank you for your comment. Childhood Adversity was measured using the Cambridge Early Experiences interview (CAMEEI). Full details are provided in the Supplementary Material on Page 2. Task : Did you use a jitter ? Thank you for your comment. We did not use jitter for the fMRI task, as the epochs used to create the contrasts of interest were counterbalanced and not event-related. Did you debrief to check if participants “believed” in the artificial social feedback? Thank you for your comment. Following the scan, as a manipulation check, participants were asked a series of questions aimed at assessing believability of the task and of the hyperscanning environment. As detailed in the Supplementary Materials (Page 2), three participants were removed from analysis as they showed doubts about the veracity of the cover story. We have included this additional comment on Page 2, paragraph 1 of the Supplementary Material to read; ‘One participant was removed due to imaging acquisition difficulties and we had doubts as to whether 3 participants fully believed in the veracity of the cover story for this particular task following the debriefing procedure.’ Reviewer: 2 Comments to the Author This manuscript by Stretton and colleagues provides important insight into how neural processing of social evaluation supports depression risk in late adolescence. Strengths of this paper include a relatively large (n=56) sample with rich longitudinal data and the inclusion of several risk factors for depression in one model. The introduction and discussion are thorough and well-researched. While the adoption of PLS correlation models for use with the fMRI data is interesting, I have some questions about this approach that made reading and interpreting the results difficult at times. Overall, providing more detail regarding the methods and results would be helpful. More specific comments are summarized below by topic. Analytic Approach As noted, I have some questions about the PLS approach and am hoping more details can be added to the manuscript for clarification. More information about PLS regression would be helpful to include in the “Behavioral Analysis” section of the methods on pg. 21 (e.g., what information do you get in the output and how do you evaluate that information?). Thank you for your comment. We have added additional information in the ‘Behavioral Analysis’ section of the methods on Page 21 to include more detail on the PLS Regression including the following; ‘The algorithm reduces the number of parameters using a technique similar to principal components analysis to extract a set of components that describes maximum correlation between the predictors and response variable(s). Components were evaluated for significance based on the percentage of variance explained in both the predictor variables and the response variable and were retained if they explained more than 10% (equivalent to small effect size (Cohen 1992)) of the variance in both variable sets. If a component was retained, the rotated factor loadings were then used to determine the importance of each variable to the component, measured as correlation coefficients +/- 0.4.’ From my understanding based on the manuscript, the PLS correlation approach used for the neuroimaging data allowed the authors to enter multiple variables and find what voxels are most robustly correlated with the set of depression risk variables. What I am less clear on, however, is whether this approach uses the entire set of 8 risk variables (as a latent factor) and finds patterns of neural activity that correlate with this latent factor, or whether this approach finds patterns of neural activity that correlate with one variable or a subset of the 8 variables. Relatedly, what does “one significant latent brain-behavior pair” refer to? Does it refer to one pattern of brain activation associated with heightened depression risk (from all 8 variables), or one combination of brain activity and behavioral data that expresses the largest covariance (and does this mean that only some of the behavioral data are included)? My assumption is that all 8 variables are used in the model, but that some variables may be driving the association with the brain more than others (hence the correlational graph), but please confirm. Thank you for your comment. The Behavioral PLS correlation approach essentially performs two multivariate regressions prior to correlating the resulting latent factors from each. In our case, one on the behavioral data (8 risk variables) and one on the neural data (activation maps of the Social Evaluation contrasts of interest). The PLS correlation then attempts to find the highest degree of shared information between these two regressions. The output is therefore two-fold, a behavioral component with factor loadings and a neural component with regional activation maps, and taken together these are interpreted as a ‘latent brain-behavior’ pair. The interpretation of the activation map is derived from the correlational graph in terms of stable correlations (correlations where the confidence interval does not include zero, see Page 13). Further, from a theoretical standpoint, I’m curious as to whether the authors think about these 8 risk variables as one latent factor of depression risk that would be associated with one pattern of brain activity, or whether they hypothesize that each variable might relate to distinct patterns of brain activity. Thank you for your comment. We do think about these eight risk variables as a latent factor influencing the activation maps related to positive and negative social evaluation. This is specific to the neuroimaging paradigm that was used in this study however, we would be very interested to see if the same latent brain-behavior pairs would be present in other activation maps derived from other paradigms sensitive to affective disorder. With regard to the uniqueness of the association between each risk factor and the brain activity, we ran the univariate ANCOVAs for this very purpose. Although the PLS correlation pointed towards certain variables (i.e. CA) being more important to the ‘brain-behaviour pair’ than others (See page 13-14) we wanted to test the univariate contribution of each psychosocial risk variable to each feedback contrast in order to increase confidence in the contribution of any one single variable to patterns of neural activity. One benefit of using PLS correlation is that the IVs (depression risk variables) are highly correlated. It would be helpful to see how correlated the measures are; please provide this information in a table or the supplement. This could strengthen the argument that these factors conceptually represent a latent factor of depression risk. If some variables are uncorrelated, please justify keeping them in the model. Thank you for your comment. We have included a correlation matrix in the Supplementary Materials, Table S3, and included reference to it on Page 21 under the ‘Biopsychosocial risk variables’ section. The correlation matrix is helpful in assessing the individual relationships between the variables, but importantly the PLS regression method also factors the dependent variable (in our case the social evaluation ratings from the task) and projects a latent factor as a function of the relationship between the predictor variables and the dependent variable. Thus, we believe removing variables from the model based on a correlation matrix would be overly conservative, particularly with strong a priori evidence that each of these variables was associated with Major Depressive Disorder. I’m a little confused about the univariate follow-up tests that were performed. Were all covariates included in one model, or were separate models conducted with each covariate? It sounds like all covariates were entered at once – if so, I wonder if you may have been underpowered to detect effects of one covariate controlling for 7 others, and/or how the correlation between these variables could have affected results. Also, what does it mean for your interpretations that these follow-up tests were not significant? Thank you for your comment. We performed follow up univariate t-tests/ANOVAs on variables with stable correlations in the latent brain-behavior pair. As noted above, this was to ensure the unique contribution of said variable. This was achieved by including all other covariates in the model and assessing the contribution of the variable of interest as detailed on Page 25-26. We performed a whole-brain analysis and images were assessed for cluster-wise significance using a cluster-defining threshold of p < 0.001 uncorrected; the 0.05 FWE-corrected critical cluster size was 350 voxels to assess the robustness of any unique contribution. We also performed these tests without controlling for the other variables and again found no robust unique contribution for any single variable. This leads us to interpret the findings within the context of ‘latent risk’ as a combination of the variables we included in our analysis. The lack of uniqueness associated to any single variable reinforces our view that depressive risk is not related to any one single variable, but is formed via the interplay between a constellation of biological, psychological and social factors. Results & Discussion 1. Critically, I believe that Figures 1 and 2 are switched. Please review this carefully. My remaining comments hinge on the assumption that they are switched. Thank you for your comment. You are correct in your assumption and we can only apologise for this oversight in the file uploads. We have rectified this error in the revision. 2. On page 11, the authors discuss “one optimal risk component that predicted affective response ratings to Negative (minus Neutral) feedback trials.” In the list of full component loadings (Table S3), however, it looks like some of the factor loadings are really poor (e.g., PHx). Are these included in the risk component or removed (I’m assuming included)? Please speak to how low factor loadings affect findings or interpretations, if at all. Thank you for your comment. The risk component includes all the variables in the model. Variables with low factor loadings were retained to the model. The factor loading is an indicator of the unique contribution of any one variable. Low factor loadings indicate relatively low importance to the projection of the latent variable but still contribute to the overall pattern of the latent factor. We have included this information within the Behavioural Analysis section of the methods on Page 22. ‘Low factor loadings indicate relatively low importance to the projection of the latent variable but still contribute to the overall pattern of the latent factor.’ 4. From a conceptual standpoint, I am also wondering how the brain and behavior findings fit together. Specifically, while high biopsychosocial risk is associated with blunted affective responses to negative and positive feedback, high biopsychosocial risk is associated with heightened responses to feedback. The authors mention that activity in some of these emotion-regulation brain regions may indicate dampening, but what about for other regions, particularly the striatum (a region highlighted in the introduction)? Associations between depression risk and heightened activity in the striatum to rejection and inclusion cues does not seem to support the overall conclusion that “risk for depression is associated with a pervasive emotional insensitivity in the face of positive and negative social feedback.” Thank you for your comment. This is an important point that we have not addressed adequately. There is clear evidence for frontostriatal dysfunction in depression during monetary reward tasks. Indeed, in a sample very similar to the one presented here, adolescent boys with a history of depression reported increased frontostriatal reactivity when receiving positive rewards relative to negative rewards (Morgan et al., 2016). The authors suggested this could reflect over-signaling of the medial PFC to dampen typical striatal response or enhance weak striatal response. This is one possible explanation for our results, in that the pattern of increased activation including emotion-regulation and striatal regions together could reflect an attenuated frontostriatal response to social evaluation. We have included this at the end of the first paragraph on page 17; ‘Finally, in a similar sample, adolescent boys with a history of depression reported increased frontostriatal reactivity when receiving positive rewards relative to negative rewards (Morgan et al., 2016). The authors suggested this could reflect over-signaling of the medial PFC to dampen typical striatal response or enhance weak striatal response (Morgan et al., 2016). In line with our findings, the pattern of increased activation including emotion-regulation and striatal regions together could reflect an attenuated frontostriatal response to both positive and negative social evaluation.’ 5. I’m hoping the authors can speak more to why we might be seeing negative correlations between BDI scores and brain activity to Negative > Neutral and positive correlations for CA, Parental PHx, and cortisol? This makes me questioning whether these variables together are really tapping depression risk, or another process. Could maybe be addressed in the discussion. Thank you for your comment. We consider these findings in the context of a resilience mechanism and thus fits well with the results of our sensitivity analysis. We have included this in our paragraph on page 18, which now reads the following; ‘It is important to consider the relationship between our findings and notions of stress inoculation and resilience (Rutter, 2012). Evolutionary theorists (Allen & Badcock, 2003; Beck & Bredemeier, 2016; Gilbert & Allan, 1998; Nesse, 2000) argue that depressed mood, including the pervasive emotional insensitivity that we find here, is in fact an adaptive or resilient response to risks of social exclusion, illness, or threats to valued resources. Depressed mood serves to withdraw the beleaguered individual from potentially disadvantageous social disputes and shifts the focus towards repair and resource conservation. It is only when this systemic response becomes entrenched or chronic that clinical depression occurs. This suggests that those biopsychosocial factors that confer a greater risk for clinical depression will also likely confer a greater risk for pervasive depressed mood, including emotion context insensitivity, as a putative resilient response. In this context, then, risk and resilience are two sides of the same coin because depressed mood – an adaptive or resilient response – also places the individual at risk of clinical depression – a maladaptive response – if that mood state becomes entrenched. This interpretation is in line with the negative correlation we observed between depressive symptom severity (BDI) and the neural response pattern to negative social evaluation. Moreover, this complexity is supported by the results of our sensitivity analysis that excluded participants with a prior mental health difficulty but revealed largely unchanged latent brain-behaviour pairs relative to the whole sample. This sensitivity analysis subsample has navigated the period of mid-adolescence associated with the greatest risk of onset of depression (and other disorders; (Spinhoven et al., 2010) without experiencing a psychiatric episode. For them, it therefore makes sense to characterise the relationship between elevated risk on our suite of biopsychosocial variables and emotion insensitivity as a putative marker of resilience.’ Minor: 1. Consider removing the mention of “numerous implicated genetic loci” from the introduction (pg. 7) since this was not included in your risk variables. Thank you for your comment. We have removed this from the manuscript. 2. The manuscript switches between “PLS regression” and “PLS correlation”. It is my understanding that these are different – please ensure that they are being used appropriately throughout. Thank you for your comment. We agree that the switch between regression and correlation methods between the behaviour and neural data has the potential to confuse. We have gone through the paper meticulously and used the appropriate nomenclature throughout. 3. The authors may think about incorporating the first paragraph of fMRI results (pg. 12) into the introduction. Thank you for your comment. We have opted to keep the results of the prior study within the Results section as after moving this to the Introduction, felt hat it would be confusing to the reader with regard to the neural hypotheses at the end of the introduction. 4. Please define PLS “behavioral saliences” for those unfamiliar with this approach. I believe they are similar to factor loadings? Thank you for your comment. You are correct. We have included this detail in the methods (see Page 13, paragraph 1) which reads the following; ‘Saliences are similar to the loadings in principal component analysis (PCA).’ 5. In supplemental Table S1, please make a note of what “CA+” and “CA-“ refer to. In Table S3, it does not look like anything is bolded, despite a nod to this in the note. Thank you for your comment. We have amended the tables accordingly. 6. Please consider re-wording the first sentence of the abstract (i.e., change "To understand" to "Understanding"). Thank you for your comment. We have changed the sentence accordingly.



    Cite this author response
  • pre-publication peer review (ROUND 1)
    Decision Letter
    2020/07/23

    23-Jul-2020

    We recognise that the impact of the COVID-19 pandemic may affect your ability to return your revised manuscript to us within the requested timeframe. If this is the case, please let us know.

    Dear Dr. Stretton:

    Manuscript ID BRB3-2020-03-0259 entitled "How biopsychosocial depressive risk shapes behavioral and neural responses to social evaluation in adolescence." which you submitted to Brain and Behavior, has been reviewed. Some revisions to your manuscript have been recommended. Therefore, I invite you to respond to the comments appended below and revise your manuscript.

    Before submitting your revisions:

    1. Prepare a response to the reviewer comments appended below in point-by-point fashion. In order to expedite the processing of the revised manuscript, please be as specific as possible in your response and indicate the page numbers in the manuscript where you have addressed each comment.

    2. Prepare a revised manuscript (word document), highlighting the changes you’ve made. Save this new document on your computer as you will be asked to upload it during the revision submission process. NOTE: Please be sure to keep in mind reviewer comments and incorporate your responses within the manuscript. There may well be areas where you disagree; for example, you may want to write, "A reviewer suggests that... However, I disagree because...". In any case, please try to address all of the concerns that are raised within the manuscript.

    3. In addition to your revised manuscript with changes highlighted, please also save a “clean” copy where the changes are not marked.

    To submit your revised manuscript:

    1. Log in by clicking on the link below

    PLEASE NOTE: This is a two-step process. After clicking on the link, you will be directed to a webpage to confirm.

    https://mc.manuscriptcentral.com/brainandbehavior?URL_MASK=326884292fdd4f159708f0b6dae4edfa

    OR

    Log into https://mc.manuscriptcentral.com/brainandbehavior and click on Author Center. Under author resources, use the button “Click here to submit a revision”. PLEASE DO NOT SUBMIT YOUR REVISIONS AS A NEW MANUSCRIPT.

    1. Follow the on-screen instructions. First you will be asked to provide your “Response to Decision Letter”—this is the response to reviewer comments that you prepared earlier.

    2. Click through the next few screens to verify that all previously provided information is correct.

    3. File Upload: Delete any files that you will be replacing (this includes your old manuscript). Upload your new revised manuscript file with changes highlighted, a “clean” copy of your revised manuscript file, any replacement figures/tables, or any new files. Once this is complete, the list of files in the “My Files” section should ONLY contain the final versions of everything. REMEMBER: figures/tables should be in jpeg, tiff, or eps format. We hope that you will designate one of the figures in your paper to be considered for our online covers and potential publication on our blog.

    4. Review and submit: please be sure to double-check everything carefully so that your manuscript can be processed as quickly as possible.

    Deadlines:
    Because we are trying to facilitate timely publication of manuscripts submitted to Brain and Behavior,your revised manuscript should be uploaded as soon as possible. If it is not possible for you to submit your revision in 2 months, we may have to consider your paper as a new submission. If you feel that you will be unable to submit your revision within the time allowed please contact me to discuss the possibility of extending the revision time.

    If you would like help with English language editing, or other article preparation support, Wiley Editing Services offers expert help with English Language Editing, as well as translation, manuscript formatting, and figure formatting at www.wileyauthors.com/eeo/preparation. You can also check out our resources for Preparing Your Article for general guidance about writing and preparing your manuscript at www.wileyauthors.com/eeo/prepresources.

    Once again, thank you for submitting your manuscript to Brain and Behavior and I look forward to receiving your revision.

    Sincerely,
    Dr. Amanda Bischoff-Grethe
    Editor in Chief, Brain and Behavior
    agrethe@ucsd.edu

    Associate Editor Comments to Author:

    Reviewer(s)' Comments to Author:

    Reviewer: 1

    Comments to the Author
    Review BRB3-2020-03-0259

    This study focuses on underlying risk markers for affective disorders, with the aim to identify through partial least square latent components that best describe the relationship between various “risk factors” and behavioral or fMRI data. They find on 56 late adolescents an association between some risk factors, mainly childhood adversity, and decreased negative affect after social rejection, decreased positive affect after social inclusion, and increased activity during social feedback in a network comprising ventral striatum, PCC, mid cingulate cortex, and temporal gyrus. Authors interpret these findings in the context of the Emotion Context Insensitivity model.

    I think the topic is very timely and the approach interesting. I support a dimensional approach using multivariate data-driven statistical technique. However I have some reservations on how these data were used and on the manuscript. It is very ambitious to cover multiple and diverse risk factors and how they predict clinical AND neuroimaging data, so I am afraid the sample does not allow for sufficient statistical power. I am also skeptical about putting BDI at T3 as a “risk factor”, since it is more a clinical description of psychopathology at T3, could be an outcome. The global interpretation is misleading : while some “risk factor” are associated with results coherent with a certain model of affective disorders, they don’t actually predict such disorders, and it is not a longitudinal study in the sense that variables from T1, T2, and T3 are mixed together. The PLS results for neuroimaging data are not very strong (30% variance explained) and not very straightforward (different variables with different patterns). Maybe trying a model with less variables but better selected could be helpful.

    A few additional comments:

    Abstract is not very clear, could be more precise on which population exactly and which inputs to PLS, Emotion Context Insensitivity model best kept for discussion I suggest.
    Introduction : In general, the introduction references too many concepts, some not necessary in my opinion. For example animal studies and stress reactivity: neither very relevant nor exhaustive; same apply for remitted MDD studies : what is the link with the study presented here? The logical link between Emotion Context Insensitivity and reward processing is also missing. I think the introduction should be refocused on the current hypothesis.
    Hypothesis : not clear, especially in the direction of expected changes in activity.
    Methods : Why exclusion of DSM IV axis I disorder ? Even is exploratory analyses later checking the if the results are different for the group with disorders.
    How was CA measured ?
    Task : Did you use a jitter ? Did you debrief to check if participants “believed” in the artificial social feedback ?

    Reviewer: 2

    Comments to the Author
    This manuscript by Stretton and colleagues provides important insight into how neural processing of social evaluation supports depression risk in late adolescence. Strengths of this paper include a relatively large (n=56) sample with rich longitudinal data and the inclusion of several risk factors for depression in one model. The introduction and discussion are thorough and well-researched. While the adoption of PLS correlation models for use with the fMRI data is interesting, I have some questions about this approach that made reading and interpreting the results difficult at times. Overall, providing more detail regarding the methods and results would be helpful. More specific comments are summarized below by topic.

    Analytic Approach

    1. As noted, I have some questions about the PLS approach and am hoping more details can be added to the manuscript for clarification. More information about PLS regression would be helpful to include in the “Behavioral Analysis” section of the methods on pg. 21 (e.g., what information do you get in the output and how do you evaluate that information?). From my understanding based on the manuscript, the PLS correlation approach used for the neuroimaging data allowed the authors to enter multiple variables and find what voxels are most robustly correlated with the set of depression risk variables. What I am less clear on, however, is whether this approach uses the entire set of 8 risk variables (as a latent factor) and finds patterns of neural activity that correlate with this latent factor, or whether this approach finds patterns of neural activity that correlate with one variable or a subset of the 8 variables. Relatedly, what does “one significant latent brain-behavior pair” refer to? Does it refer to one pattern of brain activation associated with heightened depression risk (from all 8 variables), or one combination of brain activity and behavioral data that expresses the largest covariance (and does this mean that only some of the behavioral data are included)? My assumption is that all 8 variables are used in the model, but that some variables may be driving the association with the brain more than others (hence the correlational graph), but please confirm. Further, from a theoretical standpoint, I’m curious as to whether the authors think about these 8 risk variables as one latent factor of depression risk that would be associated with one pattern of brain activity, or whether they hypothesize that each variable might relate to distinct patterns of brain activity.

    2. One benefit of using PLS correlation is that the IVs (depression risk variables) are highly correlated. It would be helpful to see how correlated the measures are; please provide this information in a table or the supplement. This could strengthen the argument that these factors conceptually represent a latent factor of depression risk. If some variables are uncorrelated, please justify keeping them in the model.

    3. I’m a little confused about the univariate follow-up tests that were performed. Were all covariates included in one model, or were separate models conducted with each covariate? It sounds like all covariates were entered at once – if so, I wonder if you may have been underpowered to detect effects of one covariate controlling for 7 others, and/or how the correlation between these variables could have affected results. Also, what does it mean for your interpretations that these follow-up tests were not significant?

    Results & Discussion

    1. Critically, I believe that Figures 1 and 2 are switched. Please review this carefully. My remaining comments hinge on the assumption that they are switched.

    2. On page 11, the authors discuss “one optimal risk component that predicted affective response ratings to Negative (minus Neutral) feedback trials.” In the list of full component loadings (Table S3), however, it looks like some of the factor loadings are really poor (e.g., PHx). Are these included in the risk component or removed (I’m assuming included)? Please speak to how low factor loadings affect findings or interpretations, if at all.

    3. From a conceptual standpoint, I am also wondering how the brain and behavior findings fit together. Specifically, while high biopsychosocial risk is associated with blunted affective responses to negative and positive feedback, high biopsychosocial risk is associated with heightened responses to feedback. The authors mention that activity in some of these emotion-regulation brain regions may indicate dampening, but what about for other regions, particularly the striatum (a region highlighted in the introduction)? Associations between depression risk and heightened activity in the striatum to rejection and inclusion cues does not seem to support the overall conclusion that “risk for depression is associated with a pervasive emotional insensitivity in the face of positive and negative social feedback.”

    4. I’m hoping the authors can speak more to why we might be seeing negative correlations between BDI scores and brain activity to Negative > Neutral and positive correlations for CA, Parental PHx, and cortisol? This makes me questioning whether these variables together are really tapping depression risk, or another process. Could maybe be addressed in the discussion.

    Minor:

    1. Consider removing the mention of “numerous implicated genetic loci” from the introduction (pg. 7) since this was not included in your risk variables.

    2. The manuscript switches between “PLS regression” and “PLS correlation”. It is my understanding that these are different – please ensure that they are being used appropriately throughout.

    3. The authors may think about incorporating the first paragraph of fMRI results (pg. 12) into the introduction.

    4. Please define PLS “behavioral saliences” for those unfamiliar with this approach. I believe they are similar to factor loadings?

    5. In supplemental Table S1, please make a note of what “CA+” and “CA-“ refer to. In Table S3, it does not look like anything is bolded, despite a nod to this in the note.

    6. Please consider re-wording the first sentence of the abstract (i.e., change "To understand" to "Understanding").

    Decision letter by
    Cite this decision letter
    Reviewer report
    2020/07/23

    This manuscript by Stretton and colleagues provides important insight into how neural processing of social evaluation supports depression risk in late adolescence. Strengths of this paper include a relatively large (n=56) sample with rich longitudinal data and the inclusion of several risk factors for depression in one model. The introduction and discussion are thorough and well-researched. While the adoption of PLS correlation models for use with the fMRI data is interesting, I have some questions about this approach that made reading and interpreting the results difficult at times. Overall, providing more detail regarding the methods and results would be helpful. More specific comments are summarized below by topic.

    Analytic Approach

    1. As noted, I have some questions about the PLS approach and am hoping more details can be added to the manuscript for clarification. More information about PLS regression would be helpful to include in the “Behavioral Analysis” section of the methods on pg. 21 (e.g., what information do you get in the output and how do you evaluate that information?). From my understanding based on the manuscript, the PLS correlation approach used for the neuroimaging data allowed the authors to enter multiple variables and find what voxels are most robustly correlated with the set of depression risk variables. What I am less clear on, however, is whether this approach uses the entire set of 8 risk variables (as a latent factor) and finds patterns of neural activity that correlate with this latent factor, or whether this approach finds patterns of neural activity that correlate with one variable or a subset of the 8 variables. Relatedly, what does “one significant latent brain-behavior pair” refer to? Does it refer to one pattern of brain activation associated with heightened depression risk (from all 8 variables), or one combination of brain activity and behavioral data that expresses the largest covariance (and does this mean that only some of the behavioral data are included)? My assumption is that all 8 variables are used in the model, but that some variables may be driving the association with the brain more than others (hence the correlational graph), but please confirm. Further, from a theoretical standpoint, I’m curious as to whether the authors think about these 8 risk variables as one latent factor of depression risk that would be associated with one pattern of brain activity, or whether they hypothesize that each variable might relate to distinct patterns of brain activity.

    2. One benefit of using PLS correlation is that the IVs (depression risk variables) are highly correlated. It would be helpful to see how correlated the measures are; please provide this information in a table or the supplement. This could strengthen the argument that these factors conceptually represent a latent factor of depression risk. If some variables are uncorrelated, please justify keeping them in the model.

    3. I’m a little confused about the univariate follow-up tests that were performed. Were all covariates included in one model, or were separate models conducted with each covariate? It sounds like all covariates were entered at once – if so, I wonder if you may have been underpowered to detect effects of one covariate controlling for 7 others, and/or how the correlation between these variables could have affected results. Also, what does it mean for your interpretations that these follow-up tests were not significant?

    Results & Discussion

    1. Critically, I believe that Figures 1 and 2 are switched. Please review this carefully. My remaining comments hinge on the assumption that they are switched.

    2. On page 11, the authors discuss “one optimal risk component that predicted affective response ratings to Negative (minus Neutral) feedback trials.” In the list of full component loadings (Table S3), however, it looks like some of the factor loadings are really poor (e.g., PHx). Are these included in the risk component or removed (I’m assuming included)? Please speak to how low factor loadings affect findings or interpretations, if at all.

    3. From a conceptual standpoint, I am also wondering how the brain and behavior findings fit together. Specifically, while high biopsychosocial risk is associated with blunted affective responses to negative and positive feedback, high biopsychosocial risk is associated with heightened responses to feedback. The authors mention that activity in some of these emotion-regulation brain regions may indicate dampening, but what about for other regions, particularly the striatum (a region highlighted in the introduction)? Associations between depression risk and heightened activity in the striatum to rejection and inclusion cues does not seem to support the overall conclusion that “risk for depression is associated with a pervasive emotional insensitivity in the face of positive and negative social feedback.”

    4. I’m hoping the authors can speak more to why we might be seeing negative correlations between BDI scores and brain activity to Negative > Neutral and positive correlations for CA, Parental PHx, and cortisol? This makes me questioning whether these variables together are really tapping depression risk, or another process. Could maybe be addressed in the discussion.

    Minor:

    1. Consider removing the mention of “numerous implicated genetic loci” from the introduction (pg. 7) since this was not included in your risk variables.

    2. The manuscript switches between “PLS regression” and “PLS correlation”. It is my understanding that these are different – please ensure that they are being used appropriately throughout.

    3. The authors may think about incorporating the first paragraph of fMRI results (pg. 12) into the introduction.

    4. Please define PLS “behavioral saliences” for those unfamiliar with this approach. I believe they are similar to factor loadings?

    5. In supplemental Table S1, please make a note of what “CA+” and “CA-“ refer to. In Table S3, it does not look like anything is bolded, despite a nod to this in the note.

    6. Please consider re-wording the first sentence of the abstract (i.e., change "To understand" to "Understanding").

    Reviewed by
    Cite this review
    Reviewer report
    2020/06/24

    Review BRB3-2020-03-0259

    This study focuses on underlying risk markers for affective disorders, with the aim to identify through partial least square latent components that best describe the relationship between various “risk factors” and behavioral or fMRI data. They find on 56 late adolescents an association between some risk factors, mainly childhood adversity, and decreased negative affect after social rejection, decreased positive affect after social inclusion, and increased activity during social feedback in a network comprising ventral striatum, PCC, mid cingulate cortex, and temporal gyrus. Authors interpret these findings in the context of the Emotion Context Insensitivity model.

    I think the topic is very timely and the approach interesting. I support a dimensional approach using multivariate data-driven statistical technique. However I have some reservations on how these data were used and on the manuscript. It is very ambitious to cover multiple and diverse risk factors and how they predict clinical AND neuroimaging data, so I am afraid the sample does not allow for sufficient statistical power. I am also skeptical about putting BDI at T3 as a “risk factor”, since it is more a clinical description of psychopathology at T3, could be an outcome. The global interpretation is misleading : while some “risk factor” are associated with results coherent with a certain model of affective disorders, they don’t actually predict such disorders, and it is not a longitudinal study in the sense that variables from T1, T2, and T3 are mixed together. The PLS results for neuroimaging data are not very strong (30% variance explained) and not very straightforward (different variables with different patterns). Maybe trying a model with less variables but better selected could be helpful.

    A few additional comments:

    Abstract is not very clear, could be more precise on which population exactly and which inputs to PLS, Emotion Context Insensitivity model best kept for discussion I suggest.
    Introduction : In general, the introduction references too many concepts, some not necessary in my opinion. For example animal studies and stress reactivity: neither very relevant nor exhaustive; same apply for remitted MDD studies : what is the link with the study presented here? The logical link between Emotion Context Insensitivity and reward processing is also missing. I think the introduction should be refocused on the current hypothesis.
    Hypothesis : not clear, especially in the direction of expected changes in activity.
    Methods : Why exclusion of DSM IV axis I disorder ? Even is exploratory analyses later checking the if the results are different for the group with disorders.
    How was CA measured ?
    Task : Did you use a jitter ? Did you debrief to check if participants “believed” in the artificial social feedback ?

    Reviewed by
    Cite this review
All peer review content displayed here is covered by a Creative Commons CC BY 4.0 license.