Individual behavioral trajectories shape whole-brain connectivity in mice
Curation statements for this article:-
Curated by eLife
eLife assessment
This is an important paper that is methodologically solid and highlights structural covariance as the neuroanatomical basis underlying individuality in genetically identical mice. The approach to individuality is very well designed, and the use of brain imaging and anatomical covariance as the underlying mechanism is well thought out. The statistical methods, while overall sound, require further justification and exploration. This paper will be of broad interest to neuroscientists, especially those working in brain plasticity or understanding unique and shared environmental influences on individuality.
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
- Neuroscience (eLife)
Abstract
It is widely assumed that our actions shape our brains and that the resulting connections determine who we are. To test this idea in a reductionist setting, in which genes and environment are controlled, we investigated differences in neuroanatomy and structural covariance by ex vivo structural magnetic resonance imaging in mice whose behavioral activity was continuously tracked for 3 months in a large, enriched environment. We confirmed that environmental enrichment increases mouse hippocampal volumes. Stratifying the enriched group according to individual longitudinal behavioral trajectories, however, revealed striking differences in mouse brain structural covariance in continuously highly active mice compared to those whose trajectories showed signs of habituating activity. Network-based statistics identified distinct subnetworks of murine structural covariance underlying these differences in behavioral activity. Together, these results reveal that differentiated behavioral trajectories of mice in an enriched environment are associated with differences in brain connectivity.
Article activity feed
-
-
Author Response
Reviewer #1 (Public Review):
This is an elegant and fascinating paper on individuality of structural covariance networks in the mouse. The core precepts are based on a series of landmark papers by the same authors that have found that individuality exists in inbred mice, and becomes entrenched when richer environments are available. Here they used structural MRI to provide whole brain analyses of differences in brain structure. They first replicated brain (mostly hippocampal) effects of enrichment. Next, they used their roaming entropy measurements to show that, after dividing their mice into two groups based on their roaming entropy, that there were no differences in brain structure between the two groups yet significant differences in brain networks as measured by structural covariance. Overall I enjoyed this paper, …
Author Response
Reviewer #1 (Public Review):
This is an elegant and fascinating paper on individuality of structural covariance networks in the mouse. The core precepts are based on a series of landmark papers by the same authors that have found that individuality exists in inbred mice, and becomes entrenched when richer environments are available. Here they used structural MRI to provide whole brain analyses of differences in brain structure. They first replicated brain (mostly hippocampal) effects of enrichment. Next, they used their roaming entropy measurements to show that, after dividing their mice into two groups based on their roaming entropy, that there were no differences in brain structure between the two groups yet significant differences in brain networks as measured by structural covariance. Overall I enjoyed this paper, though am confused (and possibly concerned) about how they arrived at their two groups and have some less important methods questions.
The division of mice into two groups (down and flat) is confusing. The methods appear to suggest that k-means clustering combined with the silhouette method was used (line 380). The actual analyses used involves 2 groups of 15 mice each. The body of the manuscript suggests that 10 intermediate mice were excluded (line 100), but the methods (line 390) suggest that 8 mice were excluded, 2 for having intermediate results and 6 for having high RE slope values.
This leads to a series of questions:
- How many mice were excluded and for what reasons, given the discrepancy between body and methods?
The discrepancy was an oversight that has been corrected. The statement with the exclusion of six upward sloping and two intermediates is correct. For the rationale see above and the inserted text in the discussion.
- Was the k-means clustering actually used? It appears that the main division of mice was based on visual assessments.
The superfluous paragraph in the method section was removed.
- If the clustering was used, did it result in 2 or 3 groups?
Slope distribution did not reveal clear groups, so it did not offer an advantage over the arbitrary decision based on slope values and described above. We have now added a graphic depiction of the slope values next to the ‘flat’ or ‘down’ matrices for greater clarity (Fig. 3B).
- The intermediate group bothers me (if it was indeed 10 intermediate mice as indicated by the body rather than 2 as indicated in the methods): if these are indeed intermediate shouldn't they be analyzed and shown to be intermediate on the graph or other measures?
These were only 2 mice, for which the matrix cannot be calculated.
- Please explain the reasoning for excluding mice for having too high of a slope (if there were indeed mice excluded for having too high of a slope).
We went to long discussions among the authors and finally decided in favor of two equally-sized groups with homogenous patterns. The effect that we observed is so large and obvious that it survives all sorts of regrouping. We have also followed the suggestion to present the continuous correlation across the whole range of animals (Fig. 2)
I'd also appreciate more discussion about the structural covariance differences between flat and down mice. It is not clear what the direction of effects are - it appears that flats show mostly increases in covariance?
Yes, covariance is greater in the top (flat) than bottom (down) group.
The structural covariance matrix for those mice with a ‘flat’ RE suggests a much higher degree of inter-regional correlation in comparison to ‘down’ or STD mice, findings confirmed and extended by the NBS analysis.
Reviewer #2 (Public Review):
Lopes et al. use genetically identical mice to address a topic of broad interest: how does variation in roaming behaviour across individuals (here, quantified via the roaming entropy) arise over time when exposed to an enriched environment, and how does this variation in behaviour relate to brain structure and networks. Specifically, by examining the roaming entropy of mice and the sizes of brain structures, the authors convincingly show 1) an increase in variability in roaming behaviour over a period of 12 weeks, 2) that mice that roam more contain an increased number of doublecortin positive cells in the dentate gyrus (indicating higher levels of neurogenesis), and 3) that roaming is associated with widespread differences in neuroanatomy. The authors additionally partition mice into two groups characterized by roaming trajectories (continuous "flat" roamers and habituating "down" roamers), construct structural covariance networks for these groups, and show that the structural covariance network for "down" roamers is similar to mice housed in standard conditions and contrasts that of "flat" roamers.
A major strength of this study is the wealth of roaming data generated by the RFID setup; the high temporal resolution, fair spatial resolution, and long period of observation (3 months) allow for measures such as roaming entropy to be precisely quantified and tracked over time. Coupled with high-resolution whole brain structural MRI and histological measurements of neurogenesis in the dentate gyrus, the dataset generated is an incredibly valuable one to probe brain-behaviour relationships. Importantly, this study showcases the power of animal studies--because the subject mice are inbred, they are virtually identical in their genetics, and therefore any variation in the data collected should arise from the non-shared environment.
An area of improvement for this study follows from its strength: the dataset collected here contains far more information on mouse behaviours than is analyzed. For instance, the sizes of a broad set of regions were found to be statistically associated with roaming behaviour, but determining how much of this anatomical variation is specifically related to differential exploration of the static environment as opposed to social contact with other animals (which could presumably be determined from the RFID data) would make this study much more impactful and interesting to the community.
An important limitation in the network analyses performed is the small number of mice. Due to sampling variation, a large number of individuals are required to estimate correlation coefficients with reasonable precision. While large-scale similarities and differences between the structural covariance (correlation) matrices are visually apparent and quite striking, confidence in these results would be increased with the inclusion of more subjects, and/or a replication cohort.
We fully agree to this judgement. It is not straightforward, however, to further increase N in these studies, both for cost and logistic reasons. Rather than investing into further improving this current study, we decided to learn from our findings and design follow-up studies that take the next steps.
Finally, while both roaming behaviour and brain structure are assessed, relationships between these measures are associative. Since brain structure was only examined at one timepoint (post-enrichment), the direction of causation cannot be assessed. It remains to be seen if behavioural variation drives anatomical variation through plasticity, or whether anatomical variation present before enrichment is predictive of future behaviours. To their credit, the authors are careful not to make causal inferences. In the context of this brain-behaviour studies, this is an important limitation to recognize, but this does not detract from the important relationships between roaming behaviour and brain structure found by the authors in this study.
In summary, while there is much more to do in studying relationships between the environment, brain structure, and behaviour, Lopes et al. take an important step ahead in describing relationships between individual roaming behavioural trajectories, brain structure, and structural covariance networks.
-
eLife assessment
This is an important paper that is methodologically solid and highlights structural covariance as the neuroanatomical basis underlying individuality in genetically identical mice. The approach to individuality is very well designed, and the use of brain imaging and anatomical covariance as the underlying mechanism is well thought out. The statistical methods, while overall sound, require further justification and exploration. This paper will be of broad interest to neuroscientists, especially those working in brain plasticity or understanding unique and shared environmental influences on individuality.
-
Reviewer #1 (Public Review):
This is an elegant and fascinating paper on individuality of structural covariance networks in the mouse. The core precepts are based on a series of landmark papers by the same authors that have found that individuality exists in inbred mice, and becomes entrenched when richer environments are available. Here they used structural MRI to provide whole brain analyses of differences in brain structure. They first replicated brain (mostly hippocampal) effects of enrichment. Next, they used their roaming entropy measurements to show that, after dividing their mice into two groups based on their roaming entropy, that there were no differences in brain structure between the two groups yet significant differences in brain networks as measured by structural covariance. Overall I enjoyed this paper, though am confused …
Reviewer #1 (Public Review):
This is an elegant and fascinating paper on individuality of structural covariance networks in the mouse. The core precepts are based on a series of landmark papers by the same authors that have found that individuality exists in inbred mice, and becomes entrenched when richer environments are available. Here they used structural MRI to provide whole brain analyses of differences in brain structure. They first replicated brain (mostly hippocampal) effects of enrichment. Next, they used their roaming entropy measurements to show that, after dividing their mice into two groups based on their roaming entropy, that there were no differences in brain structure between the two groups yet significant differences in brain networks as measured by structural covariance. Overall I enjoyed this paper, though am confused (and possibly concerned) about how they arrived at their two groups and have some less important methods questions.
The division of mice into two groups (down and flat) is confusing. The methods appear to suggest that k-means clustering combined with the silhouette method was used (line 380). The actual analyses used involves 2 groups of 15 mice each. The body of the manuscript suggests that 10 intermediate mice were excluded (line 100), but the methods (line 390) suggest that 8 mice were excluded, 2 for having intermediate results and 6 for having high RE slope values.
This leads to a series of questions:
- How many mice were excluded and for what reasons, given the discrepancy between body and methods?
- Was the k-means clustering actually used? It appears that the main division of mice was based on visual assessments.
- If the clustering was used, did it result in 2 or 3 groups?
- The intermediate group bothers me (if it was indeed 10 intermediate mice as indicated by the body rather than 2 as indicated in the methods): if these are indeed intermediate shouldn't they be analyzed and shown to be intermediate on the graph or other measures?
- Please explain the reasoning for excluding mice for having too high of a slope (if there were indeed mice excluded for having too high of a slope).I'd also appreciate more discussion about the structural covariance differences between flat and down mice. It is not clear what the direction of effects are - it appears that flats show mostly increases in covariance?
-
Reviewer #2 (Public Review):
Lopes et al. use genetically identical mice to address a topic of broad interest: how does variation in roaming behaviour across individuals (here, quantified via the roaming entropy) arise over time when exposed to an enriched environment, and how does this variation in behaviour relate to brain structure and networks. Specifically, by examining the roaming entropy of mice and the sizes of brain structures, the authors convincingly show 1) an increase in variability in roaming behaviour over a period of 12 weeks, 2) that mice that roam more contain an increased number of doublecortin positive cells in the dentate gyrus (indicating higher levels of neurogenesis), and 3) that roaming is associated with widespread differences in neuroanatomy. The authors additionally partition mice into two groups …
Reviewer #2 (Public Review):
Lopes et al. use genetically identical mice to address a topic of broad interest: how does variation in roaming behaviour across individuals (here, quantified via the roaming entropy) arise over time when exposed to an enriched environment, and how does this variation in behaviour relate to brain structure and networks. Specifically, by examining the roaming entropy of mice and the sizes of brain structures, the authors convincingly show 1) an increase in variability in roaming behaviour over a period of 12 weeks, 2) that mice that roam more contain an increased number of doublecortin positive cells in the dentate gyrus (indicating higher levels of neurogenesis), and 3) that roaming is associated with widespread differences in neuroanatomy. The authors additionally partition mice into two groups characterized by roaming trajectories (continuous "flat" roamers and habituating "down" roamers), construct structural covariance networks for these groups, and show that the structural covariance network for "down" roamers is similar to mice housed in standard conditions and contrasts that of "flat" roamers.
A major strength of this study is the wealth of roaming data generated by the RFID setup; the high temporal resolution, fair spatial resolution, and long period of observation (3 months) allow for measures such as roaming entropy to be precisely quantified and tracked over time. Coupled with high-resolution whole brain structural MRI and histological measurements of neurogenesis in the dentate gyrus, the dataset generated is an incredibly valuable one to probe brain-behaviour relationships. Importantly, this study showcases the power of animal studies--because the subject mice are inbred, they are virtually identical in their genetics, and therefore any variation in the data collected should arise from the non-shared environment.
An area of improvement for this study follows from its strength: the dataset collected here contains far more information on mouse behaviours than is analyzed. For instance, the sizes of a broad set of regions were found to be statistically associated with roaming behaviour, but determining how much of this anatomical variation is specifically related to differential exploration of the static environment as opposed to social contact with other animals (which could presumably be determined from the RFID data) would make this study much more impactful and interesting to the community.
An important limitation in the network analyses performed is the small number of mice. Due to sampling variation, a large number of individuals are required to estimate correlation coefficients with reasonable precision. While large-scale similarities and differences between the structural covariance (correlation) matrices are visually apparent and quite striking, confidence in these results would be increased with the inclusion of more subjects, and/or a replication cohort.
Finally, while both roaming behaviour and brain structure are assessed, relationships between these measures are associative. Since brain structure was only examined at one timepoint (post-enrichment), the direction of causation cannot be assessed. It remains to be seen if behavioural variation drives anatomical variation through plasticity, or whether anatomical variation present before enrichment is predictive of future behaviours. To their credit, the authors are careful not to make causal inferences. In the context of this brain-behaviour studies, this is an important limitation to recognize, but this does not detract from the important relationships between roaming behaviour and brain structure found by the authors in this study.
In summary, while there is much more to do in studying relationships between the environment, brain structure, and behaviour, Lopes et al. take an important step ahead in describing relationships between individual roaming behavioural trajectories, brain structure, and structural covariance networks.
-
Reviewer #3 (Public Review):
The present study is a comparison of brain magnetic resonance imaging (MRI) of mice who developed in an enriched environment laboratory environment, in which some mice become habituated while other mice maintain active exploration of the environment over time. Between these groups, differences are shown in the pattern of correlations between brain regions of interindividual variability, which may indicate differences in brain connectivity or other shared maturational processes between regions. Because the mice are genetically inbred and have the same shared environment, these differences are attributed to individual-level differences in environment and behavior, which are extremely difficult to isolate in non-laboratory settings.
My comments are focused on aspects of the paper that overlap with my area of …
Reviewer #3 (Public Review):
The present study is a comparison of brain magnetic resonance imaging (MRI) of mice who developed in an enriched environment laboratory environment, in which some mice become habituated while other mice maintain active exploration of the environment over time. Between these groups, differences are shown in the pattern of correlations between brain regions of interindividual variability, which may indicate differences in brain connectivity or other shared maturational processes between regions. Because the mice are genetically inbred and have the same shared environment, these differences are attributed to individual-level differences in environment and behavior, which are extremely difficult to isolate in non-laboratory settings.
My comments are focused on aspects of the paper that overlap with my area of expertise which is human brain MRI methods. The strengths of the paper include the unique environmental paradigm that provides support for important hypotheses about individual-level variation. The imaging methods are rigorous and sound, and there is a nice convergence with human work. The application of structural covariance is interesting. The weaknesseses of the paper are that the writing could be clearer. Alternative explanations for structural covariance and alterations in "down roamers" should be more fully considered. The statistical approaches could be more rigorous in places. The areas of novelty relative to past work should be more explicitly articulated.
-
