Referee 1's review (Mark Eldaief):

Manuscript Title
Harrison BJ, Pujol J, Ortiz H, Fornito A, Pantelis C and Yucel M.
“Modulation of Brain Resting-State Networks by Sad Mood Induction”

Reviewer: Mark C. Eldaief, M.D.

Manuscript Summary

The authors examined whether a self-induced mood state would serve to modulate slow variations in blood oxygen level-dependent (BOLD) activity in brain resting state networks (RSNs). To this end they employed a mood induction paradigm (adapted from that used by Damasio). Twenty-four healthy subjects underwent continuous functional magnetic resonance imaging during two autobiographical recall conditions: a neutral recall (NR) condition-during which they were asked to recapitulate an “unemotional” day in their lives; and a sad recall (SR) condition-during which they pondered an episode of sadness from their past. Scanning commenced when subjects had indicated (via button press) that the appropriate mood state had been achieved. To confirm the existence of the desired mood states during scanning, subjects completed a verbal response rating scale immediately following scanning. This analysis yielded a statistically significant change in sadness between the NR and SR conditions. BOLD data was then analyzed using an independent component analysis (ICA) paradigm. This resulted in two ICA data sets (one for each condition), which were subsequently compared.

The authors probed seven different resting state networks that had been documented to demonstrate functional connectivity at rest in prior studies. In five of the seven networks there were statistically significant changes in functional connectivity as a result of the sad mood induction. Specifically, Harrison et al. identified relevant changes in: the so-called “default mode network” (in which there was decreased connectivity), the “right and left frontoparietal networks” (where increased connectivity was seen with sad mood induction) and an auditory cortex network (where functional connectivity was also increased). Perhaps most saliently, they noted increased functional connectivity within the “paralimbic network,” specifically in the dorsal anterior cingulate, supplementary motor area, left anterior insula and left operculum. In addition, resonating past studies, they also identified both a “visual cortex” and a “sensorimotor network,” both of which were observed only during the NR condition.

Relevance of the current study

Resting state activity in the brain has garnered increasing amounts of attention. However, despite the fact that many experiments have sought to delineate the structure of these networks, relatively few studies have endeavored to empirically modulate activity within them. The importance of doing so is that it sheds light on possible functional roles for these networks. Such functional roles of the RSNs cannot be undervalued, as many authors have estimated that a large percentage of the brain’s large metabolic expenditure is devoted to the maintenance of resting activity. A particularly innovative aspect of the current study is the examination of the role that affective states play in modulating RSN connectivity. This line of investigation has broad implications for affective neuroscience as well as having ramifications for a better understanding of mood and anxiety disorders in psychiatric populations. It also continues the methodology, put forth by Damoiseaux and others, of analyzing functional connectivity across many networks. Therefore, the current study is particularly relevant and original. It puts forth a plausible method for altering functional connectivity in the resting brain and shows convincing evidence for having accomplished its stated objective.
Comments and Questions

Introduction:

The background portion is written in a concise manner, allowing the reader to logically follow the development of the clearly stated hypotheses. The authors begin with a general description of resting state activity. They then go on to describe relevant functional connectivity studies that have been done with RSNs. Specifically, they cite previous studies that have made use of independent component analysis (ICA) to delineate several resting networks. Next, the authors put forth a compelling rationale for believing that functional connectivity within the networks can be experimentally modified.

Methods:

1. Both the behavioral and the imaging techniques are documented in sufficient detail to allow replication.

2. Perhaps the most salient methodological issue is the timing of the image acquisitions. The authors cite several studies based on “continuous fMRI recordings” of resting state activity (page 5). However, one might argue that acquiring data during the SR and NR conditions does not reflect “resting” activity, but rather the measurement of mentally “active” processing. Subjects may have been using cognitive and emotional resources to continually update and maintain the desired sadness state. Indeed, the authors state (on page 5) that the “process of inducing emotion using memory recall techniques requires explicit and intense cognitive effort.” This issue is mitigated by the fact that scanning commenced only after subjects felt they had achieved the desired moods and not as the subjects were actually attempting to induce the state. However, it is not clear what instructions subject received after the button was pressed, other than the fact that they were told to lay with their eyes closed (p. 18). Were they told to try to continue to ruminate on the sad episodes (which would have resulted in the marshalling of several memory sub-systems)? If not, it seems likely that subjects’ thoughts would not be “unconstrained,” but rather would continue to center on the episode they were asked to recall. Therefore, the authors may have been clearer as to how they defined “resting” activity as the described paradigm seems to blur the line between “resting” and “active” states. An external mood induction paradigm (e.g. emotional pictures or reading emotional script) might have made the task more cognitively “passive.” However, one might argue that the induced moods may not have been as intense. Also, the authors may have seen intrinsic value in specifically studying self-generated mood states.

3. As a follow-up question to the above discussion, did the subject’s mood state vary across the scanning period (i.e. were they less sad at the end of the scanning period then they were upon pressing the button)?

4. The use of the musical piece in addition to the autobiographical recall may be problematic. The authors point out that these musical pieces have been used before to successfully induce mood states, so their efficacy is not in question. However, their use may introduce a confounding factor in that one is studying a mood state that is both extrinsically and internally driven. Have there been mood induction studies that have used both stimuli?

5. Self-report of mood states has its pitfalls, and some have tried to use more “objective measures” (e.g. skin conductance responses or startle responses in the case of fear induction). The authors allude to future studies using psychophysiological monitoring (p. 15). However, given the constraints of self-reporting, the authors properly addressed the issue of whether the desired mood was induced with rating scales performed immediately after the conditions.

Data Analysis and Processing:

1. The statistical analysis of the data obtained from image acquisitions is clearly documented in sufficient detail to allow replication.
2. The use of independent component analysis (ICA) is also excellently detailed. The employment of this method over “seed-based” correlational analysis is also laudable, and renders the results obtained even more compelling.


Results and Discussion:

1. The experimental data support the claims made in the paper. The appending tables and figures are clear and appropriately descriptive to allow the reader to interpret the relevant data.

2. The authors also explore several alternate explanations for their findings. For example, they discuss the possibility (on page 14) of non-neuronal physiologic variations affecting the resting BOLD signal.

3. There is an intriguing discussion of the possible effects of mood induction on the auditory and the right and left frontoparietal networks. There is also a speculation that decreased connectivity in the “default network” may be related to the attentional demands of the recall task.

4. The significance of the paralimbic network is discussed with respect to its role in “autonomic arousal and introceptive awareness” (page 11). However, a small suggestion is to go into slightly greater depth about the paralimbic network, specifically with regard to its individual components and its structural connectivity (e.g. what is known from human neuroanatomical studies and from comparative anatomy studies across species). This could be included in either the background or discussion section, and would lead to an even stronger justification for why mood induction should result in RSN connectivity modulation there. Also, there might have been more speculation as to why a decrease in connectivity within the paralimbic network was found.

Grammatical Corrections:

1. On page 9, in the first full paragraph after table 1 the authors state: “To assess that the associated time-courses of all identified RSNs, we performed a detailed analysis of the power spectral distribution (PSD) of each (detrended) component pattern.” The word “that” should be deleted to render the sentence less awkward.

2. On page 20, in the first full paragraph, under the heading “Group ICA,” the second sentence should read: “Group ICA methods have gained recent attention in the analysis of fMRI studies because of their robust and flexible modeling nature, which has included (switched to the past tense) several recent studies…

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements.