Impaired Neurovascular Coupling and Increased Functional Connectivity in the Frontal Cortex Predict Age‐Related Cognitive Dysfunction

Abstract Impaired cerebrovascular function contributes to the genesis of age‐related cognitive decline. In this study, the hypothesis is tested that impairments in neurovascular coupling (NVC) responses and brain network function predict cognitive dysfunction in older adults. Cerebromicrovascular and working memory function of healthy young (n = 21, 33.2±7.0 years) and aged (n = 30, 75.9±6.9 years) participants are assessed. To determine NVC responses and functional connectivity (FC) during a working memory (n‐back) paradigm, oxy‐ and deoxyhemoglobin concentration changes from the frontal cortex using functional near‐infrared spectroscopy are recorded. NVC responses are significantly impaired during the 2‐back task in aged participants, while the frontal networks are characterized by higher local and global connection strength, and dynamic FC (p < 0.05). Both impaired NVC and increased FC correlate with age‐related decline in accuracy during the 2‐back task. These findings suggest that task‐related brain states in older adults require stronger functional connections to compensate for the attenuated NVC responses associated with working memory load.


Standardized cognitive assessments
S2] The battery of tests is optimized to detect age-related changes in cognitive domains mainly concerning fluid abilities, including reaction time, practice, attention, memory (visual episodic and verbal recognition memory, attention, short-term memory, visual recognition), and executive function (working memory and strategy).Each assessment began with a motor screening test (MOT) and continued with the following tests from the battery: Delayed Match to Sample (DMS, short-term visual memory), Paired Association Learning (PAL, visual memory, and learning), Reaction Time (RTI, psychomotor speed), Rapid Visual Processing and Attention (RVPA, sustained attention), Spatial Working Memory (SWM, working memory and strategy).Below a brief description of these tests is provided.After evaluating sensorimotor function and comprehension with MOT, the participant was administered the DMS test and instructed to match the target object with visually identical stimuli presented together with visually similar items, simultaneously or with delay (0, 4, 12 seconds).In the case of the PAL test, the participant was required to recall the location previously paired with an object.RTI and RVPA tests measured the response to a target stimulus, which is an object at a specific location presented simultaneously with non-target objects or a given sequence of successively presented numbers in a series of numbers with nonmatching sequences, respectively.Finally, the SWM test implemented in CANTAB specifically assessed the participant's ability to find all hidden tokens in each number of boxes in the lowest number of attempts opening attempt.
Table S1.Age group-specific descriptive statistics and comparison of all cognitive outcome measures obtained by CANTAB performance between young and elderly participants.Unpaired comparisons revealed significant differences in all assessed cognitive domains.

Young (n=21)
Elderly (n=30) Table S1 reports the key outcome measures for the assessed cognitive domains reflecting fluid abilities in the study population.All participants successfully completed the screening tests, which indicated no significant age-related difference in the sensorimotor function of young and elderly persons.We found a statistically significant age effect in the performance of all neuropsychological tests (DMS, PAL, RTI, RVP, SWM) assessing different cognitive domains.The percentage of correct responses during the Delayed Matching Sample test (DMSPC) was lower in the elderly group indicating poorer memory function (non-verbal).We also observed a significantly worse performance on a PAL test in the elderly group (PALFAMS higher value means better performance, PALTEA28higher value means worse performance), which also implies memory impairment (specifically, episodic memory).RTIFMDMT and RVPMDL are key measures of processing speed that were significantly lower in the elderly group.RVP test also demonstrated attention deficits and slower processing in the aged group, given their lower RVPA and higher RVPPFA scores.Finally, we observed more errors during the Spatial Working Memory paradigm in older adults, captured by SWMBE468 (BE468 is a composite score referring to a number of errors for all mandatorily administered task conditions) parameters along with lower strategy score (SWMS), indicating poor working memory and executive function.CANTAB Connect Tool provides additional standardized measures that allow for a comprehensive characterization of different cognitive domains, further revealing the impact of aging on cognition.The readers are encouraged to download the data files characterizing cognitive outcome measures obtained by CANTAB Connect Research Tool from Physionet [project title: "Functional near infrared spectroscopy data recorded during n-back memory task and cognitive outcome measures from healthy young and older adults"].

Relationship between cognitive outcome measures, age and fNIRS parameters
In contrast to the n-back paradigm administered during fNIRS measurements, CANTAB Connect Research Tool provides metrics for different domains of cognitive function that are standardized.The relationship of these metrics is of interest for evaluating the domain-specific impact of aging on cognition.Therefore, we correlated RT and d' with key cognitive outcome measures yielded by the CANTAB Connect Research Tool (Table S2).We also found a significant association between d' corresponding to 2-back condition and key performance measures (DMSPC, PALFAMS28, PALTEA28, RVPA, SWMBE468, SWMS) in all assessed cognitive domains; in case of error score, these were inversely related.Measures of reaction latency (DMSMDL, RVPMDL) and reaction time, including the motor component of the response (RTIFMDMT), were also significantly correlated with 2-back RT.Please note that the correlation within the whole population is typically stronger than within age groups.On the one hand, we assume that the lower age-specific correlations can be attributed to the different difficulties of the tasks: either too easy for young participants or too hard for the elderly.On the other hand, the higher contribution of age-independent (assessment-specific) factors to the variability of these outcome measures may account for less correlated cognitive outcome measures, as in this case, age is not driving the relationship.The inference is that both tasks capture the age-related impairment of the cognitive performance, and while the n-back paradigm is assumed to measure working memory function, it correlates with other measures of fluid cognitive abilities typically declining with aging.

Table S2. Relationship between cognitive outcome measures yielded by CANTAB Connect
Research Tool and verbal n-back paradigm for the young group, elderly group and for the whole study population.Correlation analyses revealed significant associations in all assessed cognitive domains.To understand age-independent and age-specific associations, we utilized cognitive outcome measures from the n-back paradigm only from the 2-back task condition sensitive to group effect and performed correlation analysis with fNIRS parameters (Table S3).In the whole study population, only one parameter showed a significant association with βHbO characterizing NVC response during the 2-back task: first attempt memory score, a key measure of the PAL test.Conversely, increased global connection strength associate with lower PALFAMS28 and higher error scores (PALTEA28).In case of  ̅ , we found a significant negative correlation with PALFAMS28 and a positive correlation with PALTEA28, performance and error scores of the PAL test, respectively.Age-group specific analyses confirm the positive correlation between PALFAMS28 and HbO responses during the 2-back task condition, which was the only significant relationship when data of young and aged participants were processed separately.Taken together, measures of impaired NVC responses and increased FC significantly associate with key variables of cognitive function yielded by the PAL test, while other performance metrics do not correlate with such fNIRS-based parameters.If they always begin a search from the same starting point, we infer that the subject employs a planned strategy for finding the tokens.Therefore, a low score indicates high strategy use (1 = they always begin the search from the same box), and a high score indicates that they are beginning their searches from many different boxes.Calculated across assessed trials with 6 tokens or more.

Analysis of confounding variables
Sex-related differences in neurovascular coupling responses and functional connectivity Neurovascular coupling (NVC)-related hemodynamic responses evoked by cognitive n-back task were recorded by functional near-infrared spectroscopy (fNIRS).Oxy-and deoxyhemoglobin (HbO and HbR, respectively) were compared between male and female participants.With the aid of an fNIRS analytical pipeline similar to what was used to assess the impact of aging, we obtained HbO and HbR maps of t-statistics for the male-female contrast.
Analyses were carried out separately for the young and elderly groups.The impact of sex was significant in three circumscribed, separate frontal cortical areas suggesting increased NVC responses in young females compared to young males and decreased NVC responses in elderly females compared to elderly males (Figure S1).Detailed descriptive statistics are provided as Supplementary Data File.

Figure S1. Impact of sex on neurovascular coupling (NVC) responses elicited by cognitive nback task in the prefrontal and motor brain cortices.
Color-coded t-values are mapped with a cutoff at q<0.05 (obtained after false discovery rate correction) in the montage space.Significant sex-related differences revealed by the statistical contrast of HbO changes are localized both in the young (A) and in the elderly group (C).HbR maps do not show any differences between males and females, neither in the young group (B) nor in the elderly group (D).
Changes in global measures of functional connectivity (FC) during our cognitive n-back paradigm were compared between male and female participants.Figure S2 shows the normalized global node degree and normalized connection strength characterizing the male and female group for each session.Analyses were carried out separately for the young and elderly groups.Although GLM revealed a significant group effect (F(1,49)=4.592, p=0.0372), sex did not have a significant influence on any of these parameters, neither in the young nor in the aged group. between elderly males (n=12) and elderly females (n=9), respectively.None of the sexrelated differences were significant (p>0.05 for all comparisons); for further details, see supplementary text.

Effect of education on neurovascular coupling responses and functional connectivity
Since the level of education was considerably different between young and elderly groups, it may affect the primary outcome parameters obtained by fNIRS.To assess this effect we defined a group of participants with higher (having at least an MSc/MA degree) and lower level of education (everyone else).This categorization led to bins with sufficient number of observations: young participants with higher (Hi, n=13) or lower (Lo, n=8) level of education as well as n=7 and n=23 elderly participants in the Hi and Lo subgroup, respectively.First, we checked whether higher education level was associated with better neurovascular coupling responses.NVC responses were analyzed with a pipeline similar to what we used to assess the effect of age or sex (Figure S3).We obtained HbO and HbR maps of t-statistics for the Hi-Lo contrast.The impact of education level in the young group was significant in two focal areas in the prefrontal cortex suggesting altered NVC responses in participants with more education compared to young participants with less education.In the case of the elderly group, our analysis revealed frontal areas whose NVC responses were significantly higher in participants having higher level of education compared to those with lower level of education.

Figure S3
. Impact of education on neurovascular coupling (NVC) responses elicited by cognitive n-back task in the prefrontal and motor brain cortices.Hi ed.: higher education level, refers to participants with who have at least an MSc/MA degree; Lo ed.: lower education level all, refers to everyone else.Color-coded t-values are mapped with a cutoff at q<0.05 (obtained after false discovery rate correction) in the montage space.Significant education level-related differences revealed by the statistical contrast of HbO changes are localized both in the young (A) and in the elderly group (C).HbR maps do not show any differences between males and females, neither in the young group (B), nor in the elderly group (D).
Changes in global measures of functional connectivity (FC) during our cognitive n-back paradigm were compared between participants with higher education level and participants with lower education level.Figure S4 shows the normalized global node degree and normalized connection strength characterizing the Hi and Lo group for each session.Analyses were carried out separately for the young and elderly groups.Level of education did not have a significant impact on any global network metrics corresponding to different task conditions. between elderly participants with higher education levels (n=7) and with lower education levels (n=23), respectively.None of the education level-related differences were significant (p>0.05 for all comparisons), for further details, see supplementary text.

Analysis of dynamic functional connectivity (DFC)
Age-and task-related changes in DFC in the frontal cortex Dynamic analyses of FC revealed the impact of aging on the temporal evolution of functional brain networks reconstructed from correlated hemodynamics of different durations.S4.Notably, the young and elderly group differs for all n-back sessions in terms of mean and variance  ̅  () captured in 10 and 30 seconds, while task-related effects are not significant for any age group and any correlation window lengths.
Statistical tests were conducted on the mean and variance of    ̅̅̅̅̅̅̅ (), corresponding results are also reported in Table S4.Significant effect of age group was verified by the Mann-Whitney test on the mean of    ̅̅̅̅̅̅̅ () for all task conditions analyzed at 10 (Figure S4B), 30 (Figure S4D) and 60 (Figure S4F) seconds, and for the first 0-back and 2-back conditions analyzed at 90 seconds (Figure S4H).Variance of  associated with to the 0BS reference brain state corresponding to 0-back session, but less associated with the 2BS reference brain state corresponding to 2-back session.
Table S5.Transition probability analysis of brain states.Trajectories of local brain network metrics define clusters with characteristically different functional connectivity and was calculated as described in the Supplementary text.-0±0 0±0 -0BSa brain state corresponding to the average functional connectivity during 0-back sessions 1BSa brain state corresponding to the average functional connectivity during 1-back session 2BSa brain state corresponding to the average functional connectivity during 2-back session Bold: significantly higher compared to the corresponding age-comparison group, p<0.05

Figure S2 .
Figure S2.Sex-related differences in global network metrics characterizing static functional connectivity (FC) in the frontal cortex during sessions of n-back task.Individual global (i.e., referring to the whole frontal cortex) network metrics values are shown in dark purple squares (female group) or dark cyan dots (male group) separately for 0-back_1 (0b1), 1-back (1b), 0-back_2 (0b2) and 2-back (2b) sessions.Corresponding median values and interquartile ranges are displayed as magenta (female) or cyan (male) boxplots.Panel A and Panel B compare the normalized global node degree ( ̅ ) and normalized global connection strength (  ̅  ) between young males (n=12) and young females (n=9), respectively.Panel C and Panel D compare  ̅ and  ̅ between elderly males (n=12) and elderly females (n=9), respectively.None of the sexrelated differences were significant (p>0.05 for all comparisons); for further details, see supplementary text.

Figure S4 .
Figure S4.Education level-related differences in global network metrics characterizing static functional connectivity (FC) in the frontal cortex during sessions of n-back task.Hi ed.: higher education level, refers to participants with who have at least an MSc/MA degree; Lo ed.: lower education level all, refers to everyone else.Individual global (i.e., referring to the whole frontal cortex) network metrics values are shown in dark purple squares (Lo group) or dark cyan dots (Hi group) separately for 0-back_1 (0b1), 1-back (1b), 0-back_2 (0b2) and 2-back (2b) sessions.Corresponding median values and interquartile ranges are displayed as magenta (Lo) or cyan (Hi) boxplots.Panel A and Panel B compare the normalized global node degree ( ̅ ) and normalized global connection strength (  ̅  ) between young participants with more (n=13) and with less education (n=8), respectively.Panel C and Panel D compare  ̅ and  ̅ between elderly participants with higher education levels (n=7) and with lower education levels (n=23), respectively.None of the education level-related differences were significant (p>0.05 for all comparisons), for further details, see supplementary text.

Figure
S5 clearly demonstrates the effect of age group on    ̅̅̅̅̅̅̅ (), which is, however, similar across task conditions (p>0.05) and for different durations of the time window.Please observe that the obtained network metrics fluctuate in a higher range in the aged group during all nback sessions (FigureS5A, S5C, S5E, S5G).Moreover, the connection strength is higher in the elderly group independently from the time window of interest along the    ̅̅̅̅̅̅̅ ().

Figure S5 .
Figure S5.Age-related changes in dynamic functional connectivity (DFC) in the frontal cortex during sessions of n-back task.Dynamic functional connectivity (DFC) analysis was carried out in a sliding window mannerwhere each window has an ID referring to the actual functional brain networkyielding a series of  ̅  ().The dynamics of  ̅  () is compared between the young (blue) and elderly groups (red); the thick line and shaded area refer to mean and standard error, respectively.Each panel shows the temporal pattern of network metrics for different time scales corresponding to the correlation window size of 10 (A), 30 (B), 60 (C) and 90 seconds (D).  () reveals a clear distinction between the trajectories of young and elderly participants during all sessions of the n-back paradigm captured in all window scales used for DFC analysis.The dynamics of normalized global network connection strengths ( w Dnorm) values are shown in blue squares (young group, n=21) or red dots (elderly group, n=30) separately for 0-back_1 (0b1), 1-back (1b), 0-back_2 (0b2) and 2-back (2b) sessions.Panels on the left depict the average of  ̅  () corresponding to correlation window size of 10 (A), 30 (C), 60 (E) and

Cognitive outcome measures Group Rank correlation analysis CANTAB n-back paradigm Strength (rho) Probability level (p)
Visual Processing and Attention, MDL: median response latency (RVPMDL), SWM: Spatial Working Memory, BE: between errors calculated as the number of an unnecessary revisit of the previously selected box across all assessed 4, 6 and 8 token trials (SWMBE468), SWMS: SWM strategy, the number of times a subject begins a new search pattern from the same box they started with previously.If they always begin a search from the same starting point we infer that the subject is employing a planned strategy for finding the tokens.Therefore, a low score indicates high strategy use (1 = they always begin the search from the same box), and a high score indicates that they are beginning their searches from many different boxes.Calculated across assessed trials with 6 tokens or more.

Table S3 .
Relationship between cognitive outcome measures yielded by CANTAB ConnectResearch Tool and fNIRS parameters obtained during 2-back task condition.Correlation analyses revealed significant associations in the elderly group and in the whole study population for key measures of the Paired Associates Learning test.

CANTAB parameter Correlation (r or rho) Probability level (p) Correlation (r or rho) Probability level (p)
Median Five-Choice Movement Time (RTIFMDMT); RVPA: Rapid Visual Processing and Attention, MDL: median response latency (RVPMDL), SWM: Spatial Working Memory, BE: between errors calculated as the number of an unnecessary revisit of the previously selected box across all assessed 4, 6 and 8 token trials (SWMBE468), SWMS: SWM strategy, the number of times a subject begins a new search pattern from the same box they started with previously.