How do thalamocortical interactions shape causal structure among resting state brain networks during aging?

The human brain undergoes both significant structural and functional changes across the lifespan. It is important to understand the underlying causal relationship of the emerging dynamical changes in functional connectivity with age. On average, functional connectivity within resting-state networks weakens in magnitude while connections between resting-state networks tend to increase with age. Further, few recent studies show that effective connectivity within and between large scale resting-state functional networks changes over the healthy lifespan. Motivated by these findings we move one step forward to investigate the effect of the thalamus in the context of healthy aging. Using directed connectivity and weighted net causal outflow measures on resting-state fMRI data, we examine the age-related changes in both cortical and thalamocortical causal interactions within and between resting-state networks. The three of core neurocognitive networks DMN, SN, CEN networks are identified independently by carrying out ICA as well as spatially matching of hub regions with the important RSNs previously reported in the literature. Thereafter, multivariate GCA was performed to test for causality index between ROIs with and without the inclusion of left and right thalamus. There are two major findings, firstly, we observe that within network causal connections become progressively weaker with age, however, between network causal connections are getting stronger with age among core neurocognitive networks, primarily a reflection of within and between network resting-state functional connectivity. Secondly, significant modifications were found in causal connections and net causal outflows in the presence of thalamus. Finally, we found that the thalamus plays a crucial role as an exogenous drive in the reorganization of within network causal outflow, while Salience network plays a critical role in mediating between network causal outflow with age among cortical networks. Our findings with the weighted causal outflow measures strengthen the hypothesis that balancing within and between network connectivity is perhaps critical for the preservation of cognitive functions with aging.

and if we calculate unconditional 157 granger causality between X and Y, spurious causalities may occur due to common dependency 158 on Z. Thus, to eliminate the possibility of spurious causalities between two time series 159 Multivariate Granger Causality analysis (GCA) was performed to assess the causal influence 160 between nodes of SN, CEN and DMN based on the methods described in ( (Barrett,Barnett,& 161 Seth, 2010)). In MVGC spurious causalities are eliminated by conditioning out the common 162 dependencies. According to MVGC approach, if we wish to test the causality from 163 164 Y to X, we have to consider the full and reduced regressions of the following form 165 This motivates the definition of conditional G-causality statistic as the appropriate log-likeli-181 hood ratio. The conditional G -causality from Y to X is defined by 182

198
We further calculated the weighted net granger causal flow to characterise the causal networks 199 in young and elderly groups. To calculate the weighted net granger causal flow, we selected 200 the significant causal outflow connections between the nodes. Weighted out-degree of a node 201 is defined by sum of the strength (granger causal indexes) of significant causal connections 202 from a node in a network to any other node. Likewise weighted in-degree of a node is sum of 203 the strength of causal inflow connections to a node in the network from any other node. 204 Weighted net causal flow was defined as weighted (Out-In) degree. For example, weighted net 205 granger causal outflow of node X, say ∆ can be expressed using the following formula: 206  To investigate the effect of thalamus in shaping within and between network causality, we 278 included thalamus as an additional node in our analysis. For each of the three resting state 279 networks, for within network analysis, we included the right and the left thalami, as the seventh 280 and the eighth nodes. Since the mechanism of the MVGC changes with the number of nodes, 281 the causal strengths were not directly comparable between two networks having different num-282 ber of nodes. In other words, it is not meaningful to quantify the difference between causal 283 strengths of the same RSN networks before and after the inclusion of the thalami. Instead of 284 that, we focused our analysis to find out the changes in the pattern of causal connections and 285 net causal outflows. 286 287

288
In elderly group all the connections except, rRMFG to lSPL remained significant (p < 0.05, 289 false discovery rate (FDR) corrected FDR corrected). The connections from the rrMFG to the 290 lSPL was mediated by the lThal. Some significant causal connections were emerging from both 291 the thalami, from the lthal to the lRMFG, the rRMFG, the lSPL, the rSPL, from the rthal to the 292 In between network analysis, salience network exhibited prominent causal influence on both 346 default mode network and central executive network in both groups ( figure 7A and 7B). 347 Causal strengths were significantly higher in old groups compared to young (p<0.01). Other 348 directed connectivity between these three nodes were also significant, with presence of higher 349 causal strength in old compared to young suggesting an increase in between network causality 350 with aging. 351 Next we estimated the weighted net granger causal outflow in between network nodes. Among 352 the three RSNs, the SN was causal outflow hub in the between network analysis. Causal out-353 flows were significantly different in young and elderly groups (p< 0.01) ( figure 8A). Next we investigated between network causality in presence of the thalamus for both groups. 363 Causal connections from the thalamus to all three network nodes, namely the DMN, the SN, 364 and the CEN in both age groups were found significant (p< 0.05, false discovery rate (FDR) 365 corrected). Thalamus was also causally driven by CEN and SN for both elderly and young amus acts as a causal outflow hub for elderly group. SN received highest causal outflow for 369 young group (figure 8B). Unlike the within network results, in between network analysis 370 causal outflows were greater in the elderly group. Among other three networks, the SN had 371 positive outflow and the DMN had negative outflow for both the groups. After inclusion of 372 thalamus, outflows in the CEN changed its direction in old cohorts. 373 Overall, thalamus had not changed the causal connectivity pattern between three resting state 374 networks. After inclusion of thalamus, the causality dynamics between three resting state net-375 works remained unaltered. However, thalamus received causal influences from SN and CEN, 376 also influenced resting state networks. Considering the net causal outflows, causal outflows for 377 the elderly group was affected with inclusion of the thalamus. The effect of thalamus was dis-378 similar in between network analysis compared to within network analysis. Effect was higher 379 in elderly group. 380 381

383
We identified a group of 50 young and 50 elderly participants from the publicly available Cam-384 bridge Aging Neuroscience dataset (https://camcan-archive.mrc-cbu.cam.ac.uk//dataaccess/) 385 in the age range of 18-80 years who did not differ in mean age, gender distribution, or IQ from 386 Berlin dataset. Using this new dataset for independent verification, we conducted identical 387 functional and effective connectivity analyses as in the original dataset and also analysed those 388 measures by varying parcellation and node numbers in each of the three-core neurocognitive 389 resting state networks of interest. We have performed our replication analysis, with different cohort and different brain parcella-513 tion atlas. Even though, different nodes were used for three RSN, some of the major findings 514 with the earlier dataset were replicated. In within network analysis, greater number of causal 515 connections with higher strengths in younger group compared to old individuals are consistent 516 with our findings with Berlin data set. Effect of thalamus is also revealed by higher number of 517 causal connections from thalamus to other nodes in younger group compared to old individuals. 518

519
We acknowledge several limitations of our study that should be addressed by the future studies. 520 The result should be replicated in a larger lifespan cohort comprise of middle young and middle 521 elderly groups to get additional insights and true estimate of lifespan trend in causal outflow in 522 resting state networks. Future research should also investigate the sensitivity of the analysis to 523 the choice of different brain parcellation schemes. 524 525 Finally, thalamus is a heterogenous structure composed of several nuclei; each of which sends 526 distinct afferent inputs to cortical regions as well as driven by cortical outputs (add reference). 527 Thus, probing the influence of different nuclei of thalamus on reorganization of within-and 528 between-network causality of different RSNs would help to better describe the complex neu-529 rophysiological processes taking place in the brain with aging. In general, the analysis could 530 be extended to other subcortical regions to understand how cortical-subcortical connectivity 531 impact the cognitive ability across age. 532 533 In conclusion, the results of the present study demonstrate that effective connectivity analysis 534 can provide crucial insights regarding within-and between-network information flow across 535 lifespan over and above insights provided by functional connectivity measures. This study also 536 establishes thalamus as a common driver of organization and reorganization of RSNs with ag-537 ing, a conclusion that should encourage future research to explore the influences exerted by 538 subcortical structures on cortical networks and their clinical implications. 539 540