Differential effects of Down's syndrome and Alzheimer's neuropathology on default mode connectivity

Abstract Down's syndrome is a chromosomal disorder that invariably results in both intellectual disability and Alzheimer's disease neuropathology. However, only a limited number of studies to date have investigated intrinsic brain network organisation in people with Down's syndrome, none of which addressed the links between functional connectivity and Alzheimer's disease. In this cross‐sectional study, we employed 11C‐Pittsburgh Compound‐B (PiB) positron emission tomography in order to group participants with Down's syndrome based on the presence of fibrillar beta‐amyloid neuropathology. We also acquired resting state functional magnetic resonance imaging data to interrogate the connectivity of the default mode network; a large‐scale system with demonstrated links to Alzheimer's disease. The results revealed widespread positive connectivity of the default mode network in people with Down's syndrome (n = 34, ages 30–55, median age = 43.5) and a stark lack of anti‐correlation. However, in contrast to typically developing controls (n = 20, ages 30–55, median age = 43.5), the Down's syndrome group also showed significantly weaker connections in localised frontal and posterior brain regions. Notably, while a comparison of the PiB‐negative Down's syndrome group (n = 19, ages 30–48, median age = 41.0) to controls suggested that alterations in default mode connectivity to frontal brain regions are related to atypical development, a comparison of the PiB‐positive (n = 15, ages 39–55, median age = 48.0) and PiB‐negative Down's syndrome groups indicated that aberrant connectivity in posterior cortices is associated with the presence of Alzheimer's disease neuropathology. Such distinct profiles of altered connectivity not only further our understanding of the brain physiology that underlies these two inherently linked conditions but may also potentially provide a biomarker for future studies of neurodegeneration in people with Down's syndrome.


| INTRODUCTION
Down's syndrome, or trisomy 21, is the most common identified cause of intellectual disability worldwide (Bittles, Bower, Hussain, & Glasson, 2007;Contestabile, Benfenati, & Gasparini, 2010;Dierssen, 2012) and is also strongly associated with the development of amyloidosis-a hallmark of Alzheimer's disease. The amyloid cascade hypothesis (Hardy & Higgins, 1992;Selkoe, 1991) suggests that this increased risk is due to the supernumerary copy of the amyloid precursor protein (APP) gene that is present in Down's syndrome, which in turn is posited to increase the levels of beta-amyloid (Aβ) present in the brain. Indeed, histological studies have found Aβ plaques in the brains of children and teenagers with Down's syndrome (Lemere et al., 1996;Leverenz & Raskind, 1998), while this neuropathology appears in nearly 100% of people with Down's syndrome over 40 years of age (Mann, 1988). Moreover, studies of people with Down's syndrome using positron emission tomography (PET) and radioligands such as 11 C Pittsburgh  and 18 F Florbetaben have demonstrated that the presence of fibrillar Aβ is detectable in a very high proportion of people by age 50 (Annus et al., 2015;Hartley et al., 2014;Hartley et al., 2017;Jennings et al., 2015;Lao et al., 2016;Lao et al., 2017).
In the typically developing population with Alzheimer's disease, the topography of amyloid deposition is known to overlap substantially with regions that form a large-scale functional brain network known as the default mode network (DMN; Buckner et al., 2005;Buckner et al., 2009).
On the other hand, investigations into the DMN's functional connectivity architecture in people with Down's syndrome have been limited, with only a small number of studies reporting on increased inter-network connectivity of the DMN in this population (Anderson et al., 2013;Vega, Hohman, Pryweller, Dykens, & Thornton-Wells, 2015). Despite the demonstrated links to both Alzheimer's disease pathology and the DMN, there has been no investigation into this network's neurodevelopmental alterations in people with Down's syndrome, nor of the potential additive influence of Alzheimer's disease neuropathology.
As such, the present study employed multimodal neuroimaging techniques (fMRI and 11 C-PiB PET) with the aim of (a) determining potential functional connectivity alterations of the DMN in people with Down's syndrome, (b) examining the relationship between DMN connectivity and age, IQ and performance on tasks of memory and executive function in people with Down's syndrome and (c) investigating differences in DMN connectivity in people with Down's syndrome with and without fibrillar Aβ accumulation, indicative of Alzheimer's disease neuropathology.

| Participants
Ethical review and approval of this study were provided by the National Research Ethics Service, East of England Committee. The majority of participants in this study were able to provide informed consent. However, in the minority of cases where participants were not able to consent for themselves, the procedures outlined in the Mental Capacity Act (2005) were followed, and an appropriate consultee was identified who provided permission for the person with Down's syndrome to participate, while the participant provided their assent. A total of 36 participants with Down's syndrome underwent neuropsychological testing, resting state fMRI and 11 C-PiB PET. However, one participant was excluded from subsequent analysis due to the presence of ventriculomegaly, which prevented registration to standard templates, and another was excluded due to suspected co-morbid Parkinson's disease. Therefore, a total of 34 people with Down's syndrome were included (Table 1). A group of 20 age-matched typically developing control participants underwent resting-state fMRI only.
Meanwhile, from the CAMCOG-DS neuropsychological assessment, the memory for new learning subtest (scored out of 21) was used in subsequent analyses, given the propensity for Alzheimer's disease to affect this type of memory. Moreover, the subscale for language (scored out of 27) was also employed in our analyses. In addition, participants completed a modified version of the Tower of London executive function test (Krikorian, Bartok, & Gay, 1994) adapted for people with intellectual disability by Ball and colleagues (Ball, Holland, Treppner, Watson, & Huppert, 2008), scored out of 12. IQ was measured using the raw IQ scores from the Kaufmann Brief Intelligence Test, 2nd edition (KBIT-II; Kaufman & Kaufman, 2004), since using standard scores would result in many participants scoring at floor (Sinai, Hassiotis, Rantell, & Strydom, 2016).

| PET data acquisition
All neuroimaging procedures were carried out at the Wolfson Brain Imaging Centre (WBIC), University of Cambridge, UK. The presence (or absence) of fibrillar Aβ neuropathology in the brains of participants with Down's syndrome was determined using 11 C-PiB PET. All 11 C-PiB PET scanning and data analysis protocols have been previously described in detail (see Landt et al. (2011) and Annus et al. (2015)). A brief summary of these procedures can be found in the Supporting Information. Based on the outcome of the 11 C-PiB PET scan, participants were allocated to PiB-negative and PiB-positive groups on the basis of a bimodal distribution in striatal BP ND values, as described in Annus et al. (2015). 11 C-PiB PET data were not collected for typically developing controls. volumes, using the unified-segmentation method (Ashburner & Friston, 2005). Finally, images were smoothed with an 8 mm FWHM Gaussian kernel.

| MRI data acquisition
We anticipated that participants with Down's syndrome may display substantial movement during scanning. Therefore, following recent reports (Ciric et al., 2017), a motion scrubbing procedure was implemented using the artefact detection tools (ART) toolbox, to identify volumes that were motion outliers, while group differences in motion were tested using two measures; the root mean square of frame-to-frame percentage change in BOLD signal ( The medial prefrontal cortex is a hub region of the DMN that has been previously used in seed-based investigations of DMN connectivity (Keller et al., 2015). In the first level analysis, Pearson correlation coefficients between the residual BOLD time-series of the seed region and that of every other voxel in the brain were calculated using the general linear model (GLM). These correlation coefficients were subsequently Z-transformed, yielding a correlation map for each participant. For comparison, an analysis was also carried out using a seed located in the posterior parietal cortex (PPC; see Supporting Information PiB-positive Down's syndrome) were investigated using the chisquare test of association, while statistical differences in age across these three groups were investigated using the Kruskal-Wallis H test, followed by post hoc tests using Dunn's procedure. Finally, the relationship between performance on tests of IQ/cognitive function and default mode connectivity in the Down's syndrome (all) group was investigated using a whole-brain approach in SPM.

| RESULTS
There were no statistically significant differences between the typically developing control and Down's syndrome (all) groups in terms of age (U = −0.224, p = .823) or gender distribution (χ 2 = 0.318, p = .573). As expected, the two groups differed significantly in terms of motion during scanning, with the Down's syndrome (all) group showing a greater number of volumes affected by motion (see Table S1).

| Default mode network in Down's syndrome
Our first aim was to determine the differences between typically developing controls and people with Down's syndrome irrespective of Aβ pathology. As such, the initial analyses investigated DMN connectivity within and between the typically developing and Down's syndrome (all) groups.
The within-group profiles of DMN connectivity (using the medial prefrontal cortex seed) in these two groups can be seen in Figure 1.
While this analysis yielded the expected pattern of DMN connectivity for the control group (Andrews-Hanna, 2012), the Down's syndrome (all) group did not display a typical profile of DMN connectivity.
Rather, functional connectivity of the DMN seed to the rest of the brain was far more extensive in the Down's syndrome (all) group. Furthermore, although the control group displayed a similar pattern of anti-correlation with the DMN to that seen in previous studies (Fox et al., 2005), there was a strikingly different pattern of anti-correlation with the DMN in the Down's syndrome (all) group, in that almost no anti-correlation with other cortical regions was observed.
To quantify the differences seen in the within groups analysis, paired comparisons were conducted between the typically developing control

| Association of default mode alterations and cognitive performance
As the Down's syndrome (all) group also represented various degrees of risk for cognitive decline and Alzheimer's disease, the relationship of DMN connectivity to IQ and cognition were investigated in this group (Figure 6) using a whole-brain approach in SPM. There was a strong positive correlation between DMN connectivity to the right thalamus and raw IQ score (r = .674, p < .0001), memory for recently learned information (r = .721, p < .0001) and language (r = .681, p < .0001), indicating that better scores on tests measuring these faculties were correlated with stronger connectivity between the DMN and the right thalamus. Language score also cor-  Down's syndrome group was reminiscent of that which has been shown to occur in sporadic and familial Alzheimer's disease (Chhatwal et al., 2013;Gili et al., 2011;Greicius et al., 2004;He et al., 2007;Yi et al., 2015;Zhou et al., 2010). That is, the alterations in DMN connectivity seen in this study encompassed regions including the posterior cingulate cortex and precuneus, a part of the brain that also shows altered metabolic function in people with Alzheimer's disease and is a key site of Aβ deposition (Buckner et al., 2005;Buckner et al., 2009). As such, this study serves to highlight the similarities in the effects of Alzheimer's disease on the brains of people with Down's syndrome and typically developing controls, indicating the utility of involving people with Down' syndrome in research, since the development of Alzheimer's disease can more readily be predicted in this patient population, but the results may be, in large part, generalisable.
Alterations in the DMN connectivity of the posterior cingulate cortex and precuneus have also been seen in an asymptomatic population at high risk for Alzheimer's disease compared to asymptomatic individuals with no increased risk (Chhatwal et al., 2013). This is particularly interesting given the largely preclinical nature of the cohort involved in the present study, and taken together these findings may corroborate the notion that disrupted functional connectivity of the DMN is an early biomarker of Alzheimer's disease neuropathology.
However, this finding was not replicated in a subsequent study that included slightly younger asymptomatic populations at risk of sporadic Alzheimer's disease due to the presence of the risk allele APOE ε4 (Thomas et al., 2014). Yet it must be noted that in the case of the study of a familial autosomal dominant Alzheimer's disease cohort by Chhatwal et al. (2013), and in the case of people with Down's syndrome as in the present study, the onset of Alzheimer's disease neuropathology is a near certainty provided that the required mutation or additional copy of APP is present, and therefore may represent a qualitatively different population to those with other risk factors such as APOE ε4 alleles, which are not determinant but greatly increase risk.
Thus, the similarities between the present findings and those of other studies involving different clinical populations further highlight the candidacy of people with Down's syndrome for future Alzheimer's disease research. Moreover, given that the development of neuropathology is more predictable in this cohort as compared to sporadic Alzheimer's disease, the design of clinical trials for preventative treatments may be less costly and more efficient. However, before any such future studies can be carried out, it is necessary to consider the problem inherent in all Alzheimer's disease research, but which is particularly relevant when conducting studies involving people with Down's syndrome; that is, the problem of age. Given the robust link between age and amyloid deposition in people with Down's syndrome, effects attributed to the presence of amyloid cannot readily be disentangled from effects due to age. In the present study, age was not entered as a covariate into the present analysis for this very specific reason, since by removing the effect of age from any analysis, one is certain to also remove a great deal of variance that may be due to the presence of Aβ neuropathology. This pattern of DMN hyper-connectivity is similar to findings reported from previous fMRI investigations of brain connectivity involving people with Down's syndrome (Anderson et al., 2013;Vega et al., 2015). These studies found increased inter-network connectivity in people with Down's syndrome compared to controls, including greater positive connectivity between the default mode network and numerous other large-scale functional brain networks. Together, these studies suggest that the organisation of the brain into segregated networks (Fox et al., 2005) is highly disrupted in people with Down's syndrome.
Whether and how this hyper-connectivity of the DMN is related to developmental cognitive dysfunction and intellectual disability seen in people with Down's syndrome is an important question and one that should be taken up by future research. The present study may act as a catalyst for such studies, having found associations between performance on tests of IQ, memory for new learning, language and DMN connectivity to various regions, most prominently the right thalamus.
As noted above, the hyper-connectivity of the DMN in people with DS revealed in this study was coupled with a lack of anticorrelation with the DMN. In the control group, the presence of a network that is anti-correlated with the DMN was evident from the within-group analysis and encompassed many of the regions of the "task-positive network," described by Fox et al. (2005). Notably, the exact physiological nature of the anti-correlation between brain networks is still under debate (Chai, Castanon, Ongur, & Whitfield-Gabrieli, 2012;Murphy et al., 2009), with recent publications highlighting their potential biological basis (Fox, Zhang, Snyder, & Raichle, 2009;Keller et al., 2015;Spreng, Stevens, Viviano, & Schacter, 2016), and indicating that they may provide a level of segregation between two major brain systems for healthy cognitive processing (Gao & Lin, 2012;Uddin, Kelly, Biswal, Castellanos, & Milham, 2009;Vatansever et al., 2017).
The notion that anti-correlation between functionally distinct networks is necessary for normal brain function is somewhat intuitive, as it neatly lends itself to a mechanism by which distinct functional networks can be organised and segregated within the brain (Fox et al., 2005). Furthermore, it is an idea that is given some weight by consistent findings of reductions in anti-correlation with the DMN in other brain disorders, including Alzheimer's disease , attention-deficit/hyperactivity disorder (Castellanos et al., 2008), autism (Anderson et al., 2011), schizophrenia (Whitfield-Gabrieli et al., 2009 and behavioural frontotemporal dementia (Hafkemeijer et al., 2015).
Thus, the hyper-connectivity of the DMN in people with Down's syndrome (although largely weaker than in controls in areas such as the anterior cingulate) and the (almost) complete absence of anticorrelation may be a large contributing factor to the disorganisation of the DMN in this group of people.
Finally, it is important to consider the differences in brain morphology between people with Down's syndrome and typically developing controls, and its possible impact on functional connectivity estimates.
A number of volumetric MRI studies primarily employing ROI-based approaches (please see Annus et al., 2017 for a summary) have identified reduced overall grey matter volumes in people with Down's syndrome relative to controls (Beacher et al., 2010;Pearlson et al., 1998;White, Alkire, & Haier, 2003), while specific reductions in volume have also been reported in the frontal lobes, the hippocampus and the cerebellum (Aylward et al., 1999;Beacher et al., 2010;Koran et al., 2014;Pearlson et al., 1998;White et al., 2003). Meanwhile, increases in grey matter volume have been noted in the parahippocampal gyrus and in the parietal and occipital cortices (Beacher et al., 2010;White et al., 2003).
Notably, however, where cortical thickness in Down's syndrome is  Figure S1) show a slightly different pattern of default mode connectivity differences in the Down's syndrome group relative to controls, with more posterior dominance, the overall direction of results remains the same, with substantial regions of reduced anti-correlation to the DMN being seen in participants with Down's syndrome. Moreover, when comparing the PiB-negative and PiBpositive Down's syndrome groups using the PPC seed, a near identical pattern of results to that obtained using the mPFC seed is observed, in that the PiB-positive group showed a specific reduction in positive connectivity of the PPC seed to the posterior cingulate cortex. As such, the results of this additional analysis indicate that the differences we have observed in this study can be generalised to the whole DMN and are not likely to be the consequence of differences in brain morphology or the selection of a specific seed region. Furthermore, the findings of Annus et al. (2017) regarding cortical thinning in PiB-positive participants with Down's syndrome in the regions encompassing the PPC seed indicate that the use of an mPFC seed for comparing DMN connectivity between these two groups is the more appropriate choice.
In summary, the present study has demonstrated widespread disruption of the DMN in people with Down's syndrome, which is further altered in the presence of Aβ neuropathology in a pattern that is reminiscent of that seen in other populations with Alzheimer's disease. The significance of these findings may be further highlighted given the largely preclinical nature of our cohort, indicating that functional connectivity of the DMN may be useful in future studies as an early biomarker of altered neuronal function due to Alzheimer's disease.

ACKNOWLEDGMENTS
The authors would like to thank all the participants with Down's syndrome, their families and their carers for their support of, and dedication to, this research. We also thank the PET imaging technologist, radio-