Uncovering narrative aging: an underlying neural mechanism compensated through spatial constructional ability

Abstract

The narrative” is a complex cognitive process that has sparked a debate on whether its features age through maintenance or decline. To address this question, we attempted to uncover the narrative aging and its underlying neural characteristics with a cross-validation based cognitive neuro-decoding statistical framework. This framework used a total of 740 healthy older participants with completed narrative and extensive neuropsychological tests and MRI scans. The results indicated that narrative comprises macro and micro structures, with the macrostructure involving complex cognitive processes more relevant to aging. For the brain functional basis, brain hub nodes contributing to macrostructure were predominantly found in the angular gyrus and medial frontal lobe, while microstructure hub nodes were located in the supramarginal gyrus and middle cingulate cortex. Moreover, networks enriched by macrostructure included the default mode network and fronto-parietal network, indicating a higher functional gradient compared to the microstructure-enriched dorsal attention network. Additionally, an interesting finding showed that macrostructure increases in spatial contribution with age, suggesting a compensatory interaction where brain regions related to spatial-constructional ability have a greater impact on macrostructure. These results, supported by neural-level validation and multimodal structural MRI, provide detailed insights into the compensatory effect in the narrative aging process.

Full text

communications biology Article
https://doi.org/10.1038/s42003-025-07501-5
Uncovering narrative aging: an underlying
neural mechanism compensated through
spatial constructional ability
Check for updates
Yumeng Li1,2, Junying Zhang3,XinLi 1,2 & Zhanjun Zhang1,2
“The narrative”is a complex cognitive process that has sparked a debate on whether its features age
through maintenance or decline. To address this question, we attempted to uncover the narrative
aging and its underlying neural characteristics with a cross-validation based cognitive neuro-
decoding statistical framework. This framework used a total of 740 healthy older participants with
completed narrative and extensive neuropsychological tests and MRI scans. The results indicated that
narrative comprises macro and micro structures, with the macrostructure involving complex cognitive
processes more relevant to aging. For the brain functional basis, brain hub nodes contributing to
macrostructure were predominantly found in the angular gyrus and medial frontal lobe, while
microstructure hub nodes were located in the supramarginal gyrus and middle cingulate cortex.
Moreover, networks enriched by macrostructure included the default mode network and fronto-
parietal network, indicating a higher functional gradient compared to the microstructure-enriched
dorsal attention network. Additionally, an interesting finding showed that macrostructure increases in
spatial contribution with age, suggesting a compensatory interaction where brain regions related to
spatial-constructional ability have a greater impact on macrostructure. These results, supported by
neural-level validation and multimodal structural MRI, provide detailed insights into the compensatory
effect in the narrative aging process.
For the vast majority of people, experience in the real world can be repre-
sented as a first-person-dominated narrative1.“The narrative”can be
regarded as a complex process of integrating meaningful units of infor-
mation from multiple structural scales2,3. Narrative applies various cognitive
operations beneath the surface of language production to generate a fra-
mework of meaning and logic4,5. Thus, the maintenance of narratives will
enhance individuals’perceptions of this world. Because life expectancy has
risen substantially, the importance of exploring the effect of aging on “the
narrative”has increased.
Although detailed neuropsychological profilesintheprogressionof
aging demonstrate deficits in language6,7,thefeaturesreflecting linguistic
decline are still unclear. Previous studies have indicated that a potentially
sensitive measure of aging progression is “the narrative”, also called con-
nected speech, which is also regarded as a marker of Alzheimer’sdiseaseat
the earliest stage8,9. For older adults, the cohesion between their discourse
units becomes weaker, and the correlation between these units and narrative
themesalso decreases significantly10,11. Moreover, measures of semantic and
lexical content, as well as syntactic complexity, were detectable from the
prodromal stage in patients with Alzheimer’sdisease
9. However, among the
different measures of narratives, the trajectories of age-related changes even
varied in a positive direction. There were no significant differences in the
type-to-token ratio (TTR) or the number of different words (NDW)
between young patients and elderly patients. The oral vocabulary and lexical
diversity of young individuals was also greater than that of elderly
individuals12–14.
The results presented above demonstrate that substantial heterogeneity
among different narrative measures and tasks is still existing. One possible
explanation is that “the narrative”itself involves multiple cognitive opera-
tions which exhibit distinct patterns of aging. Thus, the current research
draws upon interactive-construction model theory in order to isolate the
specific operation that could capture subtle changes in aging progression.
According to the theory, “the narrative”has two structures. On the one
hand, the formation of basic language structures requires interconnections
between linguistic elements, which can be called microstructures. On the
1State Key Laboratory of Cognitive Neuroscience and Learning, Beijing Normal University, Beijing, 100875, China. 2Beijing Aging Brain Rejuvenation Initiative
(BABRI) Centre, Beijing Normal University, Beijing, 100875, China. 3Institute of Basic Research in Clinical Medicine, China Academy of Chinese Medical Sciences,
Beijing, 100700, China. e-mail: lixin99@bnu.edu.cn;zhang_rzs@bnu.edu.cn
Communications Biology | (2025) 8:104 1
1234567890():,;
1234567890():,;
Content courtesy of Springer Nature, terms of use apply. Rights reserved

other hand, the macrostructure emerges as an organizational representation
of conceptual units, involving a top-down framework determining the logic
of the entire narrative and reflecting the mind of the narrator1,15,16.These two
structures correspond to two perspectives on aging in linguistic features.
One perspective proposes that the aging of narrative ability exhibits
language-specific characteristics. The automatic processing nature of lin-
guistic features17 and their independence from general cognitive
resources18,19 lead the traditional view to suggest that, in the early stages of
aging, there is no significant decline in the core narrative processes20.
However, evidence also indicates that the diminished connections at the
semantic, phonological, or orthographic levels are significant factors that
contribute to the adverse impacts of aging on narrative production, parti-
cularlyinspokenmandarin
21,22. Other perspective suggests that the narrative
aging is primarily influenced by non-linguistic-specificcognitive
operations23–25. The decline in processing speed among older individuals has
been shown to result in prolonged latency in discourse generation26.Also,
they exhibit reduced capacity to inhibit irrelevant information, leading to
more redundant language in their narrative, lacking emphasis and
coherence27. However, this theory is restricted in its focus on cognitive
functions such as processing speed, executive function, and memory,
neglecting other cognitive dimensions. The current study presents a method
to evaluate macrostructure which likened to a top-down conceptual fra-
mework serving as a cognitive map for language generation. As a result,
spatial construction and spatial memory abilities are included along with the
classic three cognitive functions. Another alternative perspective posits that
language specific cognitive processed is note limited by cognitive resources,
resulting in narrative aging determined by language processing system
itself 18,20. However, aging signifies a comprehensive decline in cognitive
function, guiding this study to favor the previous perspective where the
associated macrostructure better portrays the development of narrative
skills in the elderly.
At present, it is unclear whether these aspects of “the narrative”show
different trajectories in aging. A scarcity of comprehensive research is also
delving into the neural underpinnings of macrostructure and micro-
structure. Furthermore, previous studies have shown that cortices serve as a
top-down organizational resource for narrative production via the contrast
of different stimulus conditions28–31. These regions are susceptible to the
effects of aging. Moreover, the dorsolateral prefrontal cortex has been found
to play a more specific role in the narrative process among older adults than
among young adults32.
In addition, recent research has indicated that since “the narrative”is
relatively complex and coherent and contains rich meaning, the higher-
order cortical networks, such as the DMN and the posterior medial network
(PMN), may be more widely involved in the narrative than the brain regions
in which discrete random stimuli are activated33–35.RegionsofthePMN
have been shown to exhibit decreased narrative-evoked responses as a
consequence of aging, and this neural process significantly predicts indivi-
dual differences in episodic memory36. Additionally, one previous study
investigated cerebral plasticity in the narrative task and described both
intrahemispheric and interhemispheric reorganization in the elderly
group37. Unfortunately, few studies have investigated the underlying neural
mechanisms of narrative aging, and limited neural correlates and the role of
aging in this type of neural processing have been demonstrated38.
The purpose of the current study was to integrate a wide range of
clinical indicators relevant to narratives and evaluate their diverse trajec-
tories and cognitive contributions to aging via theoretical categorization.
Another aim of this study was to establish the basis of the evolution of neural
correlates around narratives in individuals with normal aging and propose a
standardized framework (Fig. 1) to investigate the individual differences
within individuals with cognitive-neural aging. In line with these objectives,
we propose a cognitive neuro-decoding approach that associates indivi-
duals’latent age-related trajectory of narratives with brain measures. This
strategy mainly relies on large scale dataset of narrative and neuropsycho-
logical tests, whole-brain resting-state functional connectivity and multi-
mode structural MRI39,40.
In summary, our hypothesis revolves around the different structures of
narratives in aging and is divided into the following three points:
1. Macrostructure involves more complex cognitive processes than
microstructure and is therefore more closely related to non-linguistic
cognitive abilities. This is evidenced by the fact that the resting-state
functional connectivity network supporting the cognitive process of
macrostructures is located at a higher functional gradient level.
2. Since macrostructure involves more complex cognitive processes, they
are more susceptible to the effects of aging. The patterns of cognitive
contributions to the aging of macrostructures differ from those of
microstructures due to the influence of other cognitive abilities.
Similarly, these patterns are supported by the corresponding resting-
state functional connectivity.
3. Finally, the substrates supporting narrative aging and its underlying
neural characteristics have a corresponding structural basis involving
both gray matter and white matter.
Results
Differences in cognitive deconstruction between macro- and
microstructures
We used the “comic discourse”paradigm to record participants’nar-
rative texts and sco red the macro- and microstructure usin g quantitative
encoding (see “Methods”). The macrostructure was classified into seven
items based on the logical development of the story, while the micro-
structure was classified into ten items according to fundamental lan-
guage elements.
Multiple cognitive abilities, including memory, executive function,
attention, and language, were used as independent variables in a multiple
linear regression model with narrative scores as the dependent variable. The
analysis revealed that these cognitive abilities were significantly associated
with macrostructure. However, among these cognitive abilities, only
language-related ability (verbal fluency) was significantly associated with
microstructure (Supplementary Table 4). Furthermore, Seemingly unre-
lated regression (SUR) analysis indicated that episodic memory (χ2= 27.29,
p< 0.001) and attention ability (χ2= 10.39, p< 0.001) contributed sig-
nificantly more to macrostructure than to microstructure in narratives
(Supplementary Fig. 2) (Supplementary Table 5).
Age-related trajectories of the narrative and cognitive contribu-
tions vary with age
With increasing age, there was a noticeable reduction in macrolevel indi-
cators of narrative ability (r = −0.393, p< 0.001), while microlevel indicators
did not significantly change with age (r = 0.044, p>0.05).Nonlinearfitting
of the aging trends in narrative ability showed that a turning point for
macrolevel indicators appeared at approximately 72 years; after this point,
the rate of age-related changes increased faster. However, changes in
microlevel indicators remained relatively stable (Supplementary Fig. 3).
The subjects were divided into higher and lower age groups using
this turning point as a cutoff, and general linear regression models were
constructed for both groups (Supplementary Table 6). The results sug-
gest that spatial constructional ability significantly contributes to mac-
rolevel narrative ability (χ2=4.37,p< 0.05), while the contribution of
episodic memory decreases (χ2= 4.03, p< 0.05). Only language skills
contributed significantly to microlevel narrative ability with increasing
age (χ2=5.48,p< 0.05) (Supplementary Fig. 4). The results indicate that
macrolevel indicators are more vulnerable to aging than microlevel
indicators are. The aging of macrolevel narrative ability may become
quicker at a later life stage due to the decrease in episodic memory,
although spatial constructional ability can compensate to some extent.
In comparison, microlevel narrative ability consistently relies on lan-
guage ability. The above results indicate that macrostructure is more
correlated with non-linguistic cognitive abilities than microstructure
and is more sensitive to aging. Thus, our behavioral data validates
hypothesis one. Next, we will further emphasize the reasonableness of
hypothesis one from the perspective of brain function.
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 2
Content courtesy of Springer Nature, terms of use apply. Rights reserved

Using the cross-validation model to identify the brain regions
involved with macro- and microstructures in narratives
We used functional connectivity (edges) to predict individuals’narrative
scores by applying a connectome-based predictive model (CPM)41;the
specific statistical methods and validation procedures are shown in Sup-
plementary Fig. 5. Briefly, after constraining the narrative-related edges with
a permutation test, the edges that satisfied the fitting criteria to predict the
participants’narrative performance were selected. These validated edges
then served to establish a mode l that predicted narrative performance for the
testing set.
The macrostructure of the narrative could be predicted by 554 vali-
dated edges, while the microstructure required 546 edges. The predictive
accuracies of the macrostructure (r = 0.79, p=1.39×e
−06)and
microstructure (r = 0.62, p= 5.25e−05) in the testing set were both significant.
The specific brain regions associated with macro- and microstructures were
subsequently identified via complex network analysis. We ranked nodes
according to hub value and extracted the top nodes for the macro- and
microstructures. For the macrostructure, the top nodes were the middle
temporal gyrus, angular gyrus, medial orbital frontal cortex and frontal pole
(Fig. 2a). For the microstructure, the top nodes were the supramarginal
gyrus, precuneus, superior temporal gyrus and middle cingulate cor-
tex (Fig. 2b).
Dice similarity analysis was used to examine the neural
dissociation between macro- and microstructures (Dice coefficient = 0.16,
P
permutation
= 0.203). The results indicated that the neural substrates of dif-
ferent narrative structures can be separated functionally.
Fig. 1 | The statistical framework of this study. a The dominant cognitive con-
tribution of different narrative structures was determined. bThe CPM method was
used to constrain the specific narrative edges and hub nodes. The edges linked to the
hub nodes representing the higher cognitive hierarchy were validated. cSliding age
windows were used to investigate how the narrative-related RSFC pattern changes
with age. Three methods were used to decode the cognitive representation among
brain regions specific to narrative aging. dStructural data were used to validate the
functional basis of narrative aging.
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 3
Content courtesy of Springer Nature, terms of use apply. Rights reserved

We also performed conjunction analysis to reveal the overlapping
brain regions between macro- and microstructures. The overlapping
regions were the supramarginal gyrus, precuneus, middle temporal gyrus
and part of the insula (Fig. 2c). Although these brain regions all contribute to
narrative ability, the degree of importance to different narrative structures
varies. For instance, the MTG serves as a hub area in the brain network of
macrostructures, while the supramarginal gyrus functions as the hub area
forthemicrostructure.Inadditiontothesharedbrainregions,brainregions
specific to macrostructures included the medial prefrontal cortex and
DMN-related areas compared to microstructure-correlated brain
regions (Fig. 2d).
Investigating the higher functional hierarchy of macrostructures
The behavioral data suggest that macrostructures hold greater significance
for higher-level cognitive functions than microstructures. Therefore, we
examined the distribution of contributing edges that support narrative
processing, aiming to identify the networks in which these edges were
mainly organized.
The enrichment of hub nodes for macro- and microstructures.We
further investigated the distribution of the contributing edges that were
linked to the macro- or microstructure hub nodes42 (see “Methods”). As
shown in Fig. 3, for the macrostructure, the contributing edges that were
linked to the identified 10 nodes were mainly distributed in the FPN
(2.34 × enrichment). Moreover, for specific brain regions related to
macrostructures, such as the middle temporal gyrus (MTG), angular
gyrus (AG) and medial frontal lobe (MFL), the contributing edges were
enriched mainly in the DMN and FPN (3.26 × enrichment and
4.13 × enrichment) (Fig. 3b). For the microstructure, the contributing
edges that were linked to the identified 10 nodes were mainly distributed
in the DAN (1.40 × enrichment) (Fig. 3a). However, for specific brain
regions related to the microstructure, such as the supramarginal gyrus,
middle cingulate gyrus, and precuneus, the contributing edges were
enriched mainly in the DAN and VAN (3.82 × enrichment and
3.58 × enrichment) (Fig. 3c).
The enrichment of community patterns for macro- and micro-
structures. The enrichment fold of the community was calculated as
described for the hub nodes, that is, the actual number of observed nodes
within a specific network (such as the DMN) divided by the expected
number of nodes. To compare the contributions of various networks to the
communities, we employed a permutation test to calculate the significance of
network enrichment scores within the communities (see “Methods”).
The results indicated that the network enrichment patterns of the main
communities differed between macro- and microstructures. The macro-
structure was more dependent on the FPN and DMN (enrichment
FPN
=1.19,
Fig. 2 | Identification of the brain regions representing the narrative. a The
primary brain regions involved in macrostructure were the middle temporal gyrus,
angular gyrus, medial orbital frontal cortex, and frontal pole. bThe primary regions
involved in microstructure were the supramarginal gyrus, precuneus, superior
temporal gyrus, and middle cingulate cortex (Fig. 2b). cOverlap between the macro
regions and micro regions, including the supramarginal gyrus, precuneus, middle
temporal gyrus, and part of the insula. dThe macrospecific map was obtained by
subtracting the common map from the macro regions. The results were decoded by
Neurosynth and included the medial prefrontal cortex and DMN-related areas.
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 4
Content courtesy of Springer Nature, terms of use apply. Rights reserved

P
permutation
= 0.06; enrichment
DMN
=1.05, P
permutation
=0.09)(Fig. 4d), while
the microstructure was more dependent on the VAN (enrichment
VAN
=1.53,
P
permutation
=0.01; enrichment
FPN
=1.12, P
permutation
=0.07) (Fig. 3e). How-
ever, the network enrichment value of the macrostructure for the main
community did not reach significance, indicating that nodes from various
networks are more likely to cooperate with each other to maintain the cog-
nitive processing of the macrostructure. Therefore, the above results regarding
the DMN and FPN were mainly enriched for the hub nodes and commu-
nities of the macrostructure, which demonstrated that, compared to the
microstructure with more enrichment of the DAN and VAN, the macro-
structure operates at a greater cognitive level.
Integrating the results of 2.3 and 2.4, we found that both the neural
substrates characterizing macrostructure and the enrichment of brain net-
works are situated at higher functional hierarchy, demonstrating that
macrostructure involves more complex cognitive processes compared to
microstructure.Wehaveverified the hypothesis one, which states that
“macrostructure involves more complex cognitive processes compared to
microstructure.”
Using the RSFC pattern to explain the cognitive contribution of
narrative aging
The behavioral data support the second hypothesis that macrostructures are
more sensitive to aging than microstructures due to the complex cognitive
processes involved in episodic memory and spatial constructional abilities.
We used hub nodes from macro- and microstructures to construct the
FC matrix, which was correlated with narrative performance to generate the
narrative-related RSFC pattern. Then, the participants were divided into
continuous age windows, and the RSFC in each window was correlated with
age to determine the aging tendency of each edge to contribute to the
narrative structure. The results showed that the narrative-related functional
connectivity matrix exhibited both positive and negative changes as people
aged (Fig. 4). The narrative-related neural substrates exhibited an age-
related compensatory effect to represent the process of narrative aging; that
is, the RSFC pattern of positive brain areas better represented narrative
ability during aging, while the contribution of negative brain areas
decreased.
Decoding the specific function of regions contributing to narra-
tive aging with three mutually validating methods
To further explain the change in the RSFC pattern during narrative aging
and to verify similar compensatory effects in the behavioral results, we
examined the contribution of increased spatial constructional ability to
macrostructures and decreased episodic memory as people aged. We used
three methods to determine which cognitive function is represented by the
specific brain regions with convertible RSFC patterns during narrative
aging (Fig. 5).
First, we constructed several models for predicting different cogni-
tive functions using the RSFC patterns of specific brain regions. If a
Fig. 3 | The enrichment patterns of contributing hub nodes and communities.
aThe contributing edges that were linked to hub nodes of the macrostructure were
mainly distributed in the FPN. The contributing edges that were linked to the hub
nodes of the microstructure were mainly distributed in the DAN. bThe contribu ting
edges of specific brain regions for macrostructures, such as the MTG, AG, and MFL,
were mainly enriched in the DMN and FPN. cThe contributing edges of specific
regions for microstructures, such as the supramarginal gyrus, middle cingulate gyrus
and precuneus, were mainly enriched in the DAN and VAN. dThe enrichment
pattern of the main community of macrostructures was more dependent on the FPN
and DMN. eThe enrichment pattern of the main community of microstructures was
more dependent on the VAN.
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 5
Content courtesy of Springer Nature, terms of use apply. Rights reserved

model performed well, the corresponding brain region was considered to
represent the cognitive function. In terms of macrostructure, the RSFC
pattern of positive brain regions could predict spatial constructional
ability only (r = 0.42, P
permutation
= 0.015), while that of negative brain
regions could predict executive function and episodic memory (r = 0.38,
P
permutation
= 0.03; r = 0.39, P
permutation
= 0.02). However, for the micro-
structure, the RSFC pattern of positive brain regions could not predict
any cognitive function, while that of negative brain regions could predict
verbal fluency and episodic memory (r= 0.47, P
permutation
= 0.006;
r = 0.62, P
permutation
= 1.04 × e−04). The detailed fitting results for the
abovementioned edges and different cognitive functions are presented in
Supplementary Figs. 7 and 8.
Second, we used meta-analytic data from Neurosynth (https://
www.neurosynth.org) to decode the specificregions.Forthemac-
rostructure, the decoding results for the positive regions included the
intraparietal sulcus, posterior parietal cortex, spatial cortex, and
attention and spatial constructional cortex, while the decoding results
for the negative regions included the anterior cingulate, dorsal
anterior cortex, medial prefrontal lobe, response inhibition, and
memory. For the microstructure, the decoding results of the positive
regions included the posterior cingulate, ventral medial, precuneus,
default, and visual word regions, while the decoding results of the
negative regions included the inferior frontal, language, phonological,
sentence, and word regions.
Third, we used the method CPMto identify the hub nodes associated
with different cognitive functions. These nodes were used to calculate Dice
similarity coefficients with the specific regions of narrative aging. The higher
theDicesimilaritycoefficient is, the more the specific brain regions represent
corresponding cognitive abilities. The results indicated that the overlap
between hub nodes and positive regions of macrostructure was significant for
episodic memory (Dice coefficient = 0.15, P
permutation
=7.8×10
−04), executive
function (Dice coefficient = 0.28, P
permutation
=8.4×10
−04)andspatialcon-
structional ability (Dice coefficient = 0.31, P
permutation
<1×10
−04), while
verbal fluency was not significant (Dice coefficient = 0.29, P
permutation
=0.49).
The hub nodes representing spatial constructional ability had the most
significant overlap with the positive regions for macrostructure.
Additionally, the overlap between hub nodes and negative regions for
macrostructure was significant for episodic memory (Dice coefficient = 0.37,
P
permutation
<1×10
−04), executive function (Dice coefficient = 0.15,
P
permutation
= 0.01) and spatial constructional ability (Dice coefficient = 0.20,
P
permutation
=8.2×10
−04), while verbal fluency was not significant (Dice
coefficient = 0.22, P
permutation
= 0.40). The hub nodes representing episodic
memory had the most significant overlap with negative regions for
macrostructure.
Finally, we combined these three methods to identify the cognitive
function that is most appropriately represented by specific brain regions
(Supplementary Table 3). With regard to the macrostructure, spatial con-
structional abilities offer the most compensation for the aging process, while
the contributions of episodic memory and executive functions diminish. In
terms of microstructure, compensated or declined contributions are intri-
cately linked to language-related abilities.
Integrating the results 2.5 and 2.6, we have verified hypothesis 2 which
suggests that “due to its more complex cognitive processes, the mechanisms
underlying aging effects on macrostructure differ significantly from
microstructure,”we found that certain brain regions play an increased role
in the aging process of macrostructure (Fig. 4). To decode the specific
cognitive function associated with these brain regions, we utilized three
methods depicted in Fig. 5for decoding to achieve cross-validation. Ulti-
mately, we discovered that these brain regions correspond to spatial con-
struction abilities.
Investigating the structural basis of cognitive-neural mechan-
isms of narrative aging
By exploring the RSFC pattern associated with the aging of narrative ability,
we found an underlying compensatory effect of spatial construction ability
to macrostructure. However, based on the results in Fig. 4,compensatory
regions also exist at the microstructure as well. To further provide a struc-
tural basis for this effect and verify the hypothesis 3, we verified the
observation using both gray matter and white matter data. We theorized
that the specific brain regions compensate during the narrative process due
to their structural features that support this function, which is demonstrated
through two results. First, the coordinated changes in gray matter structure
between narrative hub nodes and compensatory brain regions will be more
significant. Second, the strength of white matter connectivity between hub
nodes and specific brain regions can predict changes in functional con-
nectivity patterns associated with aging (Fig. 6).
The structural covariance network was used to calculate the
coordinatedchangeingraymattervolume(GMV)betweenhub
Fig. 4 | An age-related compensatory effect represents the process of
narrative aging. We correlated the RSFC matrix with age in each time window to
determine the aging tendency of each edge to contribute to the narrative structure.
The narrative-related neural substrates exhibited an age-related compensator y effect
to represent the process of narrative aging; that is, the RSFC pattern of positive brain
areas could better represent narrative ability during aging, while the contribution of
negative brain areas decreased.
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 6
Content courtesy of Springer Nature, terms of use apply. Rights reserved

nodes and specific regions (see “Methods”). The results indicate that
hub nodes and compensatory regions of macrostructures exhibit
more coordinated changes in gray matter volume (GMV) with age.
Compared to those in the declining regions, the compensatory
regions with coordinated changes were more numerous and exhibited
greater significance (Ms =0.11, P
permutation
= 0.03). However, the
related changes in the hub nodes and compensatory regions for
microstructures were not significantlygreaterthanthoseforthe
declining regions (Ms =0.02, P
permutation
=0.30).
Additionally, the linear regression model using the white matter net-
work to predict the RSFC pattern also validated the compensatory effect of
specific brain regions related to narrative aging (see “Methods”). The results
indicate that the white matter connectivity strengths between macro-
structure hub nodes and specific brain regions (including compensatory
regions and declining regions) can successfully predict age-related changes
characterizing the macrostructure (r = 0.32, p=1.26×e
−04). In other words,
the stronger the compensatory effect on the macrostructure is, the greater
the benefit from the stronger white matter connectivity between the mac-
rostructure hub nodes and compensatory brain regions. However, for
microstructure, the performance of the predictive model was not adequate
(r = 0.13, p= 0.18).
Discussion
We established distinct macro- and microstructures using a wide range of
narrative-related indicators that are theoretically separable. Through the
integration of large-scale multimodal data from brain imaging and com-
prehensive neuropsychological tests, we also investigated the cognitive
mechanisms and neural substrates associated with narrative ability and its
underlying compensated effect in aging. Our results revealed a mechanism
in which macrostructures display an age-related increase in spatial con-
tribution coupled with a decrease in episodic memory contribution. This
mechanism was verified at the neural level, and the associated RSFC pattern
demonstrated a structural relationship.
We obtained convergent evidence from the behavioral, functional, and
structural data supporting the neural substrates of narrative components,
demonstrating the complex cognitive contributions across different narra-
tive levels. The approach used in this study provides a robust and stable
statistical framework for investigating a wide range of cognitive functions.
Consequently, this study has significantly contributed to our understanding
of the cognitive contributions to narrative and their changes with aging, as
well as their neural underpinnings. Furthermore, this study provides insight
into related theories and has potential implications for the development and
clinical application of innovative neuropsychological tests.
Fig. 5 | The schematic diagram of three methods to determine the cognitive
domain represented by specific RSFC patterns during narrative aging. We used
three methods to determine which cognitive function was represented by the specific
brain regions with convertible RSFC patterns during narrative aging. aMethod 1:
We constructed several models predicting different cognitive functions using the
RSFC patterns of specific brain regions. If a model performed well, the corre-
sponding brain region was considered to represent cognitive function. bMethod 2:
we used meta-analytic data from Neurosynth (https://www.neurosynth.org)to
decode the specific regions. cMethod 3: We used the method in Fig. 1to identify the
hub nodes related to different cognitive functions. These nodes were used to cal-
culate Dice similarity coefficients with the specific regions of narrative aging. The
higher the Dice similarity coefficient is, the more the specific brain regions represent
corresponding cognitive abilities.
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 7
Content courtesy of Springer Nature, terms of use apply. Rights reserved

The central focus of this study revolves around the macro- and
microstructure of narratives. To distinguish these two components, pre-
vious research has used the strategy of recording young adults’brain activity
when performing single narrative tasks and using conditional contrasts to
identify the component associated with higher-level cognitive processes28,30.
The activated brain regions included the inferior frontal lobe, which over-
laps with the medial frontal lobe identified in this study. The inferior MFLs
are also key regions for distinguishing macro- and microstructures,
implying that this region plays a vital role in top-down organization for
maintaining the theme and logical flow of narrative processes rather than
supporting basic linguistic units43,44. In addition to the MFL, the angular
gyrus (AG) has been overlooked in previous studies on narrative discourse,
despite its involvement in various high-level cognitive functions, particu-
larly episodic memory and semantic memory45.TheAGplaysarolein
integrating and retrieving the supramodal experience, such as by integrating
individual concepts into larger groups, which is consistent with the concept
of schema46,47. Within the macrostructure, the AG is suggested to engage in
integrative operations generating the conceptual frame (schema) crucial for
organizational representations between conceptual units. The middle
temporalgyrusalsoservesasoneofthehubnodescontributingtothe
macrostructure. A previous study using near-infrared spectroscopy revealed
significant activation in the left middle temporal gyrus (MTG) among older
adults, potentially indicating its involvement in the semantic interpretation
of narrative text32. However, our findings suggest that the right middle
temporal gyrus could serve as the neural basis of the macrostructure. This
finding is reasonable because the right MTG plays an important role in
semantic violation tasks, which requires the detection of conflict arising
from semantic anomalies based on the context48,49. Hence, the MTG appears
to engage in not only the representation of the propositional context within
the macrostructure but also the semantic interpretation within the micro-
structure, which tends to be responsible for the intermediate transition
during the narrative process. However, other studies have shown that the
ability to interpret at the abstract level is more likely to be dependent on a
ventral pathway linking the right MTG with the anterior inferior frontal
lobe50. The interaction of macro hub nodes may increase the involvement of
the MTG.
In contrast, the hub node that contributed most to the microstructure
was the supramarginal gyrus. The supramarginal gyrus is regarded as a
particularly important region in phonological processing, which is part of
the initial stage of the microstructure according to the interactive-
construction model51–53. However, the middle cingulate cortex (MCC) is
involved in the recognition of smaller, meaningful events among narrative
speech, which helps to establish local conceptual propositions compared to
the global conceptual frame of macrostructures54. Moreover, the micro-
structure hub node is also attached to a small region of the precuneus.
According to a previous study, the precuneus is a relatively high-order
region with responses that are strongly contextually modulated55,56.Thus,
the MCC-precuneus hub node can also be regarded as an intermediate
region involved in the transition to the macrostructure. In summary, each
region identified to represent narrative ability in the present study could
contribute to a potential subnetwork that can be used to construct the neural
mappings involved in the detailed process of the interactive construction
model, which provides the valid logic and qualities of narrative produc-
tion (Fig. 7).
Based on the hub nodes, we also analyzed the enrichment patterns of
macro- and microstructures. Regardless of the hub nodes or the main
Fig. 6 | The brain structural basis of the cognitive-neural mechanisms of
narrative aging. a The structural covariance network was used to calculate the
coordinated change in GMV between hub nodes and specific regions, indicating that
the coordinated changes in gray matter structure between narrative hub nodes and
compensatory brain regions are more significant. bThe linear regression model
using the white matter network to predict the RSFC pattern also validated the
compensatory effect of specific brain regions related to narrative aging, and the
strength of white matter connectivity between hub nodes and specific brain regions
can predict changes in functional connectivity patterns associated with aging.
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 8
Content courtesy of Springer Nature, terms of use apply. Rights reserved

community, the related contributing edges were enriched in a similar pat-
tern. For the macrostructure, the majority of contributing edges were linked
to the fronto-parietal network (FPN) and default mode network (DMN).
Serving as a key control system, the FPN supports context-dependent
narrative operation and was demonstrated to play a role in resolving
semantic ambiguity when appropriate contextual cues are lacking57,58.In
addition, the FPN is suggested to maintain the coherence and causal rela-
tionships that encourage plot narratives, which are considered to fit well
with the process of semantic connection according to the context of the
macrostructure59,60. The DMN is the highest level in the functional hierarchy
of the human brain; its contribution to various cognitive functions has been
widely studied, and the narrative is no exception. Like the FPN, the DMN is
also sensitive to contextual information61. However, the DMN can process
implicit context, which suggests that the DMN is more adept at information
integration and forming corresponding experiences or schemas62.Inaddi-
tion, previous studies have indicated that the DMN plays a strong role in the
underlying processing of a coherent narrative, and this processing is par-
ticularly related to the encoding and comprehension of conceptual struc-
tures among content63. For example, the sustained activation of the DMN
within its components (especially the angular gyrus) results in integrative
operations and frame generation according to the structural characteristics
of the narrative being processed5,33,64,65. Together with our results, these
studies suggested that the FPN and DMN, as networks of high-level regions,
are responsible for the cognitive processing of macrostructures. The FPN is
involved in the construction of semantic connections in the propositional
context, and the DMN is involved in information integration and the for-
mation of a top-down conceptual framework.
For the microstructure, the majority of contributing edges were linked
to the ventral attention network (VAN) and dorsal attention network
(DAN). The VAN was demonstrated to highly overlap with the language
network and mirror the language network in the right hemisphere of the
brain66. Some of the constituent regions of the VAN and DAN are also the
key nodes contributing to the network that is specific to the reading
function67. This has also resulted in better reading performance when
functional connectivity within the VAN is greater68. More specifically,
previous studies have indicated that various brain regions involved in the
VAN are responsible for phonological processing, syntactic graphs and
semantic interpretation, which are suggested to activate the phonological,
syntactic and semantic components of spoken words using executive or
selective attention69–71.
One notable finding from our study is that brain regions associated with
spatial constructional ability make a greater contribution to macrostructure
narrative processing when regions related to memory exhibit a decreased
contribution with age. This finding suggested a compensatory interaction
between these two cognitive processes. While language and space exist as
distinct entities in our system of representation, they intricately interact.
Even some researchers propose that language is spatial72. In social com-
munication, speakers often anchor their utterances to their spatial
Fig. 7 | Integrated diagram of the neural mappings of two narrative components.
The detailed processes of macro- and micronarrative are composed of four modules
and their associated elements: (1) linguistic component processing; (2) semantic
proposition processing; (3) contextual representation; and (4) formation conceptual
framework. The neural mapping of these processes is depicted as region groupings
highlighted in light yellow and blue (representing the micro- and macrostructure s of
the narrative). The pink groupings indicate the processes of contextual repre-
sentations that relate to both micro- and macrostructures.
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 9
Content courtesy of Springer Nature, terms of use apply. Rights reserved

environment73. This anchoring is referred to as “deixes”. Guided by deixes,
narrators can effectively convey discourse context and enhance narrative
skills, facilitating audience comprehension73,74. In information representa-
tion, theorists posit that various materials manifest diverse conceptual fra-
mework, with language and space acting as quintessential examples75.Both
narrative and spatial representations span the spectrum from concrete to
abstract. At the concrete level, the representation of space and language
differs from each other. Spatial representations are typically perceptual,
geometric, and sensorial, while language tends to be conceptual, algebraic,
and amodal. Nevertheless, at the abstract level, both space and language need
to extract schema from perceptual or event information, potentially leading
to convergence in the abstract realm concerning conceptual structures and
spatial patterns75,76. This is also the reason why, some individuals with spatial
constructional deficitshaveshownnormalscoresinmicrolinguistic
assessments but demonstrate weaknesses in macrostructure narrative
capacity77.Thisfinding suggested that there is some relationship between the
cognitive mechanism of spatial ability and macrostructure narrative ability.
This mutual effect is mainly reflected in the following two aspects.
Also, according to situation models, adults tend to update spatial
location information that is consistent with the protagonist’sperspective
while processing narrative tasks78,79. In other words, this ability to grasp
spatial information serves as a strategy to support the formation of a nar-
rative framework. Corresponding to the compensatory effect observed in
the present study, the aging brain adopts a strategy to increase the con-
tribution of brain regions involved in spatial perception to the narrative
process to delay the rapid decline in narrative ability. Moreover, updating
spatial location information in narrative framework formation requires less
intensive cognitive resources than updating other types of information, such
as temporal series information80. Thus, the compensatory effect of spatial-
related brain regions contributing to the narrative framework is the most
significant and effective.
Furthermore, cortical spatial constructional processing is universally
divided into ventral and dorsal pathways81. According to our results, the
brain regions involved in the dorsal pathway, such as the posterior parietal
lobe and the intraparietal sulcus, tended to be compensatory regions. The
dorsal pathway was more likely to be labeled the pathway that supported
both spatial perception and nonconscious spatial processing and guided
individual behavior and actions81. This ability to organize cognitive units
coincides with the macrostructure of narrative. Therefore, the dorsal
pathway, especially the posterior parietal region, theoretically contributes to
the central mechanism underlying macrostructure narrative ability. It is
reasonable to consider these regions as compensating for the decline in
narrative ability associated with aging.
Future studies can be conducted to address the following aspects. First,
it would be valuable to investigate whether the compensatory effect asso-
ciated with spatial constructional ability can be generalized to longitudinal
data, allowing for the examination of changes over time. Second, because the
participants in the present study were all healthy older adults, it remains
unclear whether the compensatory mechanism is specifictonon-
pathological aging. It is important to ascertain whether individuals experi-
encing pathological aging are capable of compensating for the decline in
narrative ability through similar neural substrates. Third, understanding the
factors that influence the extent of this compensatory response and whether
it effectively delays the decline of narrative ability is essential.
In summary, this study offers insights into the cognitive and
neural factors underlying narrative abilities and proposes a potential
compensatory mechanism involving brain regions associated with
spatial constructional processing. This perspective offers a novel
approach to comprehending the nature of narratives and their
changes during the aging process.
Methods
Participants
Seven hundred and forty participants (533 females) aged 50 to 90 years
(mean age 68.26 ± 8.20 years) participated in the first behavioral part of the
study. The inclusion criteria were as follows: native Mandarin speaker, score
≥24 on the Chinese version of the Mini-Mental State Examination (MMSE),
and the ability to complete a battery of neuropsychological tests. Individuals
who hada history of major neurological or psychiatric illness were excluded.
The above participants also underwent fMRI scans. The participants
had high-quality resting-state fMRI and T1 MRI data.
Stimuli and procedures
Narrative. A four-panel comic was used in the study as the narrative
material (Supplementary Fig. 1). The comic depicts a story about an old
woman who fell down accidentally when she got off the bus and received
help from the surrounding people. The instructions required the subjects
to familiarize themselves with the pictures first; after the experimenter
confirmed that the subjects were ready to tell the story, they were asked to
start the recording. The experimenter also asked the participants to
confirm the completion of the story and stop the recording. All the text
materials were transcribed by an iFlytek machine and then manually
proofread by a psychological undergraduate, who is a native Mandarin
speaker. The undergraduate student was not aware of the status of the
participants. The first author, also a native Mandarin speaker, coded the
samples; at the time of coding, the first author was also blinded to the
demographic information of the participants.
Cognition. As described in our previous study, all participants under-
went a battery of neuropsychological tests at baseline82. The assessment
involved general cognitive ability and cognitive function across five
domains, namely, memory, language, attention, visuospatial abilities, and
executive function. General cognitive ability was tested using the Chinese
version of the Mini-Mental State Examination (MMSE)83; memory was
tested using the Auditory Verbal Learning Test (AVLT)84; executive
function was tested using the Trail Making Test (TMT)85; spatial con-
structional ability was tested using the ROCF-Copy test86; and language
was tested using the Verbal Fluency Test (VFT)87.
Coding
Macrostructure measures. Deconstructing the overall organization of
the story allowed us to create corresponding scoring standards for seven
macrostructure elements, namely, character, setting, initiating, event,
internal response, plan, action series, and consequence. Each element was
scored on a scale of 0–3 points88,89.
The detailed grading criteria are presented in the supplementary
materials (Supplementary Table 1). According to the rules, the frequency of
each macrostructure element described by the participants, as well as the
connection of the element to the main storyline, was taken into account. The
scoring of different elements focused on the subjects’grasp of the causality of
materials, which can better reflect the narrative modes of different subjects.
Microstructure measures. The microstructures used in this study
included the total number of characters or words (TNC/TNW), total
number of different characters or words (NDC/NDW), longest sentence
length (LSL), average sentence length (ASL), average character frequency
(ACF) and average word frequency (AWF), as well as grammatical errors,
fluency problems and content errors. TNC and TNW referred to the
original number of words in the narrative recording transcribed into text;
NDC and NDW referred to the number of characters and words that are
not repeated in the text; and LSL was defined as the number of words used
in the sentence with the most words in the text. The subjects described
picture 1 and picture 2 as one sentence and picture 3 and picture 4 as
another sentence; ASL was calculated by adding and averaging on this
basis. ACF and AWF were defined as the sum of the frequency of each
character or word used in the text. The richer and more diverse words
used, the lower the average frequency was. Finally, grammatical errors,
fluency problems and content errors were the statistics of the errors in the
narrative text of the participants. If an error in a sentence appeared once,
the number of grammatical errors was recorded as 1, with a maximum of
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 10
Content courtesy of Springer Nature, terms of use apply. Rights reserved

3 points. Similarly, fluency problems were defined as the number of
pauses during the narrative, and content errors were defined as
descriptions inconsistent with the picture contents. The above scores
were generated automatically via Python.
Image acquisition and data preprocessing
MRI data acquisition. MRI data, including T1-weighted MRI and resting-
state functional magnetic resonance (fMRI) scans, were acquired via a Sie-
mens Trio 3T scanner at the Imaging Center for Brain Research at Beijing
Normal University. During the scans, participants lay supine with their head
snugly fixed by straps and foam pads to minimize head movement and were
instructed to stay awake and relax with their eyes closed. High-resolution T1-
weighted, sagittal 3D magnetization-prepared rapid gradient echo sequences
were acquired and covered the entire brain (176 sagittal slices, repetition
time = 1900 ms, echo time = 3.44 ms, slice thickness = 1 mm, flip angle = 9°,
inversion time = 900 ms, field of view = 256 mm × 256 mm, and acquisition
matrix = 256 × 256). Resting-fMRI data were acquired using a gradient echo
EPI sequence (TE=30ms, TR=2000 ms, flip angle = 90°, 33 slices, slice
thickness = 4 mm, in-plane matrix = 64 × 64, field of view = 256 × 256 mm2).
The resting scans lasted for approximately 8 min, and 240 image volumes
were obtained.
T1-weighted MRI preprocessing.TheSPM12 and CAT12 (http://www.
neuro.uni-jena.de/cat) toolboxes were used to preprocess the anatomical
images. Preprocessing was performed with the default settings, including
high-dimensional diffeomorphic anatomical registration through expo-
nentiated lie algebra (DARTEL) normalization algorithms and modulation
of nonlinear components. Preprocessing steps also included the segmenta-
tion of whole-brain images into GM, WM, and cerebrospinal fluid and the
normalization of the DARTEL template to the Montreal Neurological
Institute space (Template_1_IXI555_MNI152.nii). Finally, we estimated the
total intracranial volume (TIV) and the absolute volume of GM/WM using
the “Estimate Total Intracranial Volume”module of CAT12. All GM and
WM maps were smoothed by an 8 mm full width at half-maximum
(FWHM) kernel.
Resting-fMRI preprocessing. For each participant, the first 10
volumes were discarded to allow for adaptation to the magnetic field.
Resting data were preprocessed using Statistical Parametric Mapping
(SPM; http://www.fil.ion.ucl.ac.uk/spm/), including slice timing,
within-subject interscan realignment to correct possible movements,
spatial normalization to a standard brain template in the Montreal
Neurological Institute coordinate space, resampling to 3 × 3 × 3 mm3,
and smoothing with an 8 mm full-width at half-maximum Gaussian
kernel. In addition, resting-fMRI data were processed with linear
detrending and 0.01–0.08 Hz bandpass filtering and regression cor-
rection for nuisance covariates, which included six motion parameters,
the global mean signal, the white matter signal, and the cerebrospinal
fluid signal.
Cortical parcellation and region of interest (ROI) definition. To esti-
mate functional connectivity from resting-state fMRI data, we first par-
cellated the brain into 400 parcels (i.e., nodes) according to the Schaefer
2018 parcellation atlas matched to Yeo 7 networks90. Behavioral data
suggest a strong correlation between narrative ability and various cog-
nitive processes, particularly episodic memory, executive function, and
spatial constructional abilities. Additionally, prior research has demon-
strated that the ventromedial prefrontal cortex, orbital frontal cortex,
anterior cingulate cortex, and default mode network are the main regions
involved in narrative ability. Therefore, we selected the default mode
network (DMN), frontoparietal network (FPN), dorsal attention network
(DAN) and ventral attention network (VAN) as regions of interest and
excluded networks that do not represent high-level cognitive features of
narrative ability, such as the visual network91–94. For each participant, the
degree of connectivity was estimated by calculating the Pearson
correlation coefficient of the 236 nodes’BOLD time series, which resulted
in 27,730 edges.
Data reliability. The majority of microstructures were generated auto-
matically by the same code. However, the macrostructure scores and the
three kinds of errors needed to be validated. Following the same rubric,
another undergraduate majoring in psychology independently scored
20% of the text materials that were randomly selected from all narratives.
Cohen’s k (a statistic measuring interrater agreement for qualitative data)
indicated that agreement between the two coders was substantial
(k= 0.611, p< 0.0001)95.
Statistical analysis
Behavioral data. The relationship between the two kinds of narrative
indicators and cognitive functions was investigated by a multiple linear
regression model. Additionally, with macrostructure and microstructure
as dependent variables, a seemingly uncorrelated regression (SUR) model
was used to test the significant difference between the coefficients of these
two regression models to detect dissimilar patterns of cognitive
interpretation96. Furthermore, we used the Sharpley value for the pre-
sentation of multidomain cognitive functions, which contributed to the
explanatory power of the regression model97.
In addition, regarding the trajectory of narrative aging, a generalized
additive model (GAM) was used to estimate the nonlinear trend to find the
inflection point at the group level. We used the SUR model to evaluate the
differences in the coefficients between two subgroups divided according to
age (older age and middle-aged patients).
Neuroimaging data
Linking macro- and micro-structures of narratives with functional con-
nectivity patterns. We described an algorithm inspired and modified from
the CPM protocol41 for selecting narrative correlated edges based on a set of
single-subject connectivity matrices, constraining these edges using cross-
validation of the testing set and finally constructing a predictive model to test
the generalization of these edges (schematic shown in Fig. 1). Through this
approach, the specific brain regions associated with narrative components
were identified. Then, these contributing edges could be constructed as an
undirected network. We ranked nodes according to the authority value of
network topological measurements98 and extracted the top 50 nodes as the
primary brain regions representing the narrative components. Additionally,
to evaluate the neural disassociation between different narrative structures,
we calculated the degree of overlap of the primary nodes contributing to the
narrative performance, which was quantified by Dice similarity
coefficients99. In addition, we used conjunction analysis to visualize com-
mon or specific nodes that represented macro- and microstructure narrative
components.
In addition, we performed community detection on the undirected
weighted network using the Louvain algorithm100,101. Based on the principle
of maximizing modularity, the Louvain algorithm calculates a greedy
algorithm to minimize the number of edges within communities and
maximize the number of edges between communities. During each itera-
tion, the algorithm assigns nodes to communities, optimizing the mod-
ularity of each community and obtaining the optimal community allocation
result. The community assignment results were subsequently subjected to
hierarchical clustering via the linkage function, with Ward’s method serving
as the distance measure index. In the present study, four communities were
identified.
Validating the higher functional hierarchy of macrostructures mapped
according to RSFC. In the hub brain regions (nodes) representing the
narrative components, we estimated the enrichment patterns of the con-
tributing edges of the corresponding network connected to these nodes. The
enrichment fold was calculated as the ratio of the actual observed number of
selected edges within the network (Ai) to the expected number of selected
edges(Ei).Aiwasdefined as the number of edges by which the node was
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 11
Content courtesy of Springer Nature, terms of use apply. Rights reserved

actually connected to the target network. Ei was calculated by multiplying
the total number of edges of a node connected to all networks by the number
of nodes in the target network and dividing by the maximum number of
edges that can connect to that node (i.e., 236)42,102. We selected the top 10
hub nodes for the macro- and microstructures and calculated the enrich-
ment fold using the contributing edges associated with them. We also
independently calculated the enrichment fold using the contributing edges
of representative nodes located in different brain regions.
Similarly, the enrichment fold of the community was calculated as
described for the hub nodes, the actual observed nodes within the specific
network divided by the expected number of nodes. Here, Ai refers to the
number of nodes in a specific network actually included in the target
community, and Ei refers to the total number of nodes in the target com-
munity multiplied by the actual number of nodes in the specificnetwork
divided by the total number of nodes.
The permutation test for calculating enrichment significance involved
the randomization of the labels of nodes within a community and their
assignment to different communities while maintaining the size of the
original community. For example, if the main community comprises 90
nodes, it would still contain the same number of nodes.
We used the BrainSpace toolbox to identify, visualize, and analyze the
large-scale gradients of brain organization among different narrative per-
formance groups103.Inthefirst stage, we identified the top third of parti-
cipants according to their macro- and microstructure scores while excluding
those who overlapped. Next, the RSFC patterns of these two groups were
used as the input matrix to generate gradients. High-dimensional RSFC data
were mapped onto a low-dimensional manifold using the default PCA
linear dimensionality reduction method, and the gradients of all participants
served as templates for gradient alignment via Procrustes analysis. Finally,
we conducted independent sample T tests on the gradient values of each
node of the two aligned groups of participants and identified the brain
regions with significant differences.
Investigating changes in the RSFC patterns associated with narrative aging
using sliding age windows. The analysis used the hub brain regions repre-
senting the narrative components as seeds to construct the functional
connectivity network of each subject. First, participants were arranged in
ascending order according to their age and divided into consecutive age
windows of 30 people each. Second, we constructed the RSFC matrix of each
participant using the macro- and micro-structure hub nodes as seeds. Third,
we computed the correlation matrix between the RSFC and narrative scores
among participants in each age window. Finally, we related the RSFC-
narrative correlation matrix to the mean value of the age window and used a
permutationtesttotestforsignificance (false discovery rate [FDR] BH
corrected, P< 0.05), which yielded the trend of the contribution of the
related edges to narrative changes with age.
We did not use a fixed age range as a time window to avoid unnecessary
errors caused by a significant difference in the number of participants in
different age windows (because the age distribution of the subjects was close
to normal). However, we verified that there was no difference between these
two methods in exploring the neural correlation of narrative aging (Sup-
plementary Fig. 2).
Validating three methods for investigating cognitive function represented
by narrative aging-related brain regions. Fit linear regression model: Based
on the RSFC pattern of narrative aging, a functional connectivity matrix was
constructed between each subject’s narrative hub nodes and specificbrain
regions (such as positive regions of macrostructures). Then, cross-validation
was performed by randomly selecting 74 subjects as the test dataset and the
remaining subjects as the training dataset. Next, linear regression model was
established with the functional connectivity matrix as the independent
variable and various cognitive functions as the dependent variables. In
addition, the model used cross-validation to evaluate the performance. The
model was fitted using the training dataset and used to predict the test
dataset. Finally, the correlation coefficient between the predicted cognitive
scores and actual cognitive scores was calculated as the evaluation index for
the model.
Neurosynth decoding: We decoded the specific regions using meta-
analyses in Neurosynth104 (https://www.neurosynth.org/) and selected
the first 30 related terms. Only one entry with the same meaning was
retained (e.g., inferior frontal and inferior, keeping the former), and a
word cloud of the remaining entries was generated based on their
correlation value.
CPM analysis: The top 50 hub nodes for each cognitive function were
identified using the flowchart in Supplementary Fig. 5. The degree of overlap
between these nodes and specific brain regions was calculated using the Dice
similarity coefficient. A permutation test was used to calculate the sig-
nificance of similarity in the Dice coefficient.
Using multimodal structural data to validate the RSFC pattern in
narrative aging. Furthermore, we aimed to provide structural brain vali-
dation for the functional neural associations in narrative aging. We
extracted the gray matter volume in each node and constructed the struc-
tural covariant network of all the subjects (with the hub cluster as the seed
point and the nodes associated with narrative aging as the mask). The
structural covariant network was correlated with the mean age in each
window (successive age windows as above), and the development trend of
each edge with increasing age and the structural connectivity of the seed
points was obtained. The significant edges after correction were included in
the subsequent analysis.
In addition, we selected fiber number as an indicator to construct a
white matter network. The edges of the white matter network linked the
narrative hub and the nodes involved in narrative aging as the independent
variation, while the correlation between these edges in the resting-state
functional connectivity network and age was used as the dependent varia-
tion. We constructed a high-dimensional regression model to predict
functional connectivity patterns associated wit h narrative aging using white
matter structures. First, we normalized the data and divided it into training
and testing sets. Second, the independent variation and dependent variation
were calculated in both the training and test sets. We used the variations in
the training set to construct a linear regression model using the cross-
validation method to tune the hyperparameters. Finally, we used the cor-
relation between the predicted and actual results to evaluate the model
performance. The predicted response variable was calculated on the basis of
the test set data by using the regression coefficient and intercepts of the
linear model.
Reporting summary
Further information on research design is available in the Nature Portfolio
Reporting Summary linked to this article.
Data availability
The Neurosynth database is available at https://neurosynth.org/.Rawdata
of the older adults with their completed narrative and extensive neu-
ropsychological tests and MRI scans are available from the corresponding
author on reasonable request. Restriction of raw data is to protect the privacy
of participants. Source data are provided with this paper.
Code availability
Custom codes are variable at https://github.com/Rainmon2020/Narrative-
Aging2022.
Received: 17 April 2024; Accepted: 8 January 2025;
References
1. Kintsch, W. The Psychology of Discourse Processing (1994).
2. Hasson, U., Chen, J. & Honey, C. J. Hierarchical process memory:
memory as an integral component of information processing. Trends
Cogn. Sci. 19, 304–313 (2015).
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 12
Content courtesy of Springer Nature, terms of use apply. Rights reserved

3. Zacks, J. M. & Tversky, B. Event structure in perception and
conception. Psychol. Bull. 127, 3 (2001).
4. Schank, R. C. & Abelson, R. P. Knowledge and memory: The real
story (s. 1–85) In: (Hillsdale: NJ. Lawrence Erlbaum Associates,
1995).
5. Baldassano, C. et al. Discovering event structure in continuous
narrative perception and memory. Neuron 95, 709–721. e705 (2017).
6. Verma, M. & Howard, R. J. Semantic memory and language
dysfunction in early Alzheimer’s disease: a review. Int. J. Geriatr.
Psychiatry 27, 1209–1217 (2012).
7. Taler, V. & Phillips, N. A. Language performance in Alzheimer’s
disease and mild cognitive impairment: a comparative review. J.
Clin. Exp. Neuropsychol. 30, 501–556 (2008).
8. Mueller, K. D., Hermann, B., Mecollari, J. & Turkstra, L. S. Connected
speech and language in mild cognitive impairment and Alzheimer’s
disease: a review of picture description tasks. J. Clin. Exp.
Neuropsychol. 40, 917–939 (2018).
9. Ahmed, S., Haigh, A. M., de Jager, C. A. & Garrard, P. Connected
speech as a marker of disease progression in autopsy-proven
Alzheimer’s disease. Brain 136, 3727–3737 (2013).
10. Marini, A., Boewe, A., Caltagirone, C. & Carlomagno, S. Age-related
differences in the production of textual descriptions. J.
Psycholinguist. Res. 34, 439–463 (2005).
11. Wright, H. H. & Capilouto, G. J. Manipulating task instructions to
change narrative discourse performance. Aphasiology 23,
1295–1308 (2009).
12. Fergadiotis, G. & Wright, H. H. Lexical diversity for adults with and
without aphasia across discourse elicitation tasks. Aphasiology 25,
1414–1430 (2011).
13. Kemper, S. & Sumner, A. The structure of verbal abilities in young
and older adults. Psychol. Aging 16, 312 (2001).
14. Kemper, S., Thompson, M. & Marquis, J. Longitudinal change in
language production: effects of aging and dementia on grammatical
complexity and propositional content. Psychol. Aging16, 600 (2001).
15. Kintsch, W. & Van Dijk, T. A. Toward a model of text comprehension
and production. Psychol. Rev. 85, 363 (1978).
16. Kintsch, W. The use of knowledge in discourse processing: a
construction-incrementation model. Psychol. Rev. 85,363
–394 (1988).
17. Shtyrov, Y. Automaticity and attentional control in spoken language
processing: neurophysiological evidence. Ment. Lex. 5, 255–276
(2010).
18. Caplan, D. & Waters, G. The relationship between age, processing
speed, working memory capacity, and language comprehension.
Memory 13, 403–413 (2005).
19. Waters, G. S. & Caplan, D. Age, working memory, and on-line
syntactic processing in sentence comprehension. Psychol. Aging
16, 128 (2001).
20. Shafto, M. A. & Tyler, L. K. Language in the aging brain: the network
dynamics of cognitive decline and preservation. Science 346,
583–587 (2014).
21. Qun, Y. & Qing-Fang, Z. Aging of word frequency, syllable frequency
and phonological facilitation effects in Chinese speech production.
J. Psychol. Sci. 6, 1303 (2015).
22. Wu, H., Yu, Z., Wang, X. & Zhang, Q. Language processing in normal
aging: contributions of information-universal and information-
specific factors. Acta Psychol. Sin. 52, 541 (2020).
23. Salthouse, T. A. The processing-speed theory of adult age
differences in cognition. Psychol. Rev. 103, 403–428 (1996).
24. Adams, A. M. & Gathercole, S. E. Limitations in working memory:
implications for language development. Int. J. Lang. Commun.
Disord. 35,95–116 (2000).
25. Burke, D. M. & College, P. Language, aging, and inhibitory deficits:
evaluation of a theory. J. Gerontol. Ser. B: Psychol. Sci. Soc. Sci. 52,
P254–P264 (1997).
26. Morrison, C. M., Hirsh, K. W., Chappell, T. & Ellis, A. W. Age and age
of acquisition: an evaluation of the cumulative frequency hypothesis.
Eur. J. Cogn. Psychol. 14, 435–459 (2002).
27. Bortfeld, H., Leon, S. D., Bloom, J. E., Schober, M. F. & Brennan, S. E.
Disfluency rates in conversation: effects of age, relationship, topic,
role, and gender. Lang. Speech 44, 123–147 (2001).
28. Kuperberg, G. R., Lakshmanan, B. M., Caplan, D. N. & Holcomb, P. J.
Making sense of discourse: an fMRI study of causal inferencing
across sentences. Neuroimage 33, 343–361 (2006).
29. Mason, R. A. & Just, M. A. How the brain processes causal
inferences in text: A theoretical account of generation and
integration component processes utilizing both cerebral
hemispheres. Psychol. Sci. 15,1–7 (2004).
30. Troiani, V. et al. Narrative speech production: an fMRI study using
continuous arterial spin labeling. Neuroimage 40, 932–939 (2008).
31. Xu, J., Kemeny, S., Park, G., Frattali, C. & Braun, A. Language in
context: emergent features of word, sentence, and narrative
comprehension. Neuroimage 25, 1002–1015 (2005).
32. Martin, C. O. et al. Narrative discourse in young and older adults:
behavioral and NIRS analyses. Front Aging Neurosci. 10, 69 (2018).
33. Baldassano, C., Hasson, U. & Norman, K. A. Representation of real-
world event schemas during narrative perception. J. Neurosci. 38,
9689–9699 (2018).
34. Buckner, R. L. & DiNicola, L. M. The brain’s default network: updated
anatomy, physiology and evolving insights. Nat. Rev. Neurosci. 20,
593–608 (2019).
35. Simony, E. et al. Dynamic reconfiguration of the default mode
network during narrative comprehension. Nat. Commun. 7, 12141
(2016).
36. Reagh, Z. M., Delarazan, A. I., Garber, A. & Ranganath, C. Aging
alters neural activity at event boundaries in the hippocampus and
Posterior Medial network. Nat. Commun. 11, 3980 (2020).
37. Scherer, L. C. et al. Neurofunctional (re)organization underlying
narrative discourse processing in aging: evidence from fNIRS. Brain
Lang. 121, 174–184 (2012).
38. Cuevas, P., He, Y., Billino, J., Kozasa, E. & Straube, B. Age-related
effects on the neural processing of semantic complexity in a
continuous narrative: modulation by gestures already present in
young to middle-aged adults. Neuropsychologia 151, 107725
(2021).
39. Deco, G., Jirsa, V. K. & McIntosh, A. R. Emerging concepts for the
dynamical organization of resting-state activity in the brain. Nat. Rev.
Neurosci. 12,43–56 (2011).
40. Kanai, R. & Rees, G. The structural basis of inter-individual
differences in human behaviour and cognition. Nat. Rev. Neurosci.
12, 231–242 (2011).
41. Shen, X. et al. Using connectome-based predictive modeling to
predict individual behavior from brain connectivity. Nat. Protoc. 12,
506–518 (2017).
42. Cole, M. W. et al. Multi-task connectivity reveals flexible hubs for
adaptive task control. Nat. Neurosci. 16, 1348–1355 (2013).
43. Ramnani, N. & Owen, A. M. Anterior prefrontal cortex: insights into
function from anatomy and neuroimaging. Nat. Rev. Neurosci. 5,
184–194 (2004).
44. Cosentino, E., Adornetti, I. & Ferretti, F. in 2013 Workshop on
Computational Models of Narrative (Schloss Dagstuhl-Leibniz-
Zentrum fuer Informatik, 2013).
45. Humphreys, G. F., Lambon Ralph, M. A. & Simons, J. S. A unifying
account of angular gyrus contributions to episodic and semantic
cognition. Trends Neurosci. 44, 452–463 (2021).
46. Gilboa, A. & Marlatte, H. Neurobiology of schemas and schema-
mediated memory. Trends Cogn. Sci. 21, 618–631 (2017).
47. Ritchey, M. & Cooper, R. A. Deconstructing the posterior medial
episodic network. Trends Cogn. Sci. 24, 451–465 (2020).
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 13
Content courtesy of Springer Nature, terms of use apply. Rights reserved

48. Newman, A. J., Pancheva, R., Ozawa, K., Neville, H. J. & Ullman, M.
T. An event-related fMRI study of syntactic and semantic violations.
J. Psycholinguist. Res. 30, 339–364 (2001).
49. Ullman, M. T. A neurocognitive perspective on language: the
declarative/procedural model. Nat. Rev. Neurosci. 2, 717–726
(2001).
50. Kilner, J. M. More than one pathway to action understanding. Trends
Cogn. Sci. 15, 352–357 (2011).
51. MacSweeney, M., Capek, C. M., Campbell, R. & Woll, B. The signing
brain: the neurobiology of sign language. Trends Cogn. Sci. 12,
432–440 (2008).
52. Sliwinska, M. W., Khadilkar, M., Campbell-Ratcliffe, J., Quevenco, F.
& Devlin, J. T. Early and sustained supramarginal gyrus contributions
to phonological processing. Front Psychol. 3, 161 (2012).
53. Lazard, D. S. et al. Phonological processing in post-lingual deafness
and cochlear implant outcome. Neuroimage 49, 3443–3451 (2010).
54. Whitney, C. et al. Neural correlates of narrative shifts during auditory
story comprehension. Neuroimage 47, 360–366 (2009).
55. Regev, M., Honey, C. J., Simony, E. & Hasson, U. Selective and
invariant neural responses to spoken and written narratives. J.
Neurosci. 33, 15978–15988 (2013).
56. Ben-Yakov, A., Honey, C. J., Lerner, Y. & Hasson, U. Loss of reliable
temporal structure in event-related averaging of naturalistic stimuli.
Neuroimage 63, 501–506 (2012).
57. Smirnov, D. et al. Fronto-parietal network supports context-dependent
speech comprehension. Neuropsychologia 63, 293–303 (2014).
58. Assouline, A. & Mendelsohn, A. Weaving a story: narrative formation
over prolonged time scales engages social cognition and
frontoparietal networks. Eur. J. Neurosci. 57, 809–823 (2023).
59. Naci, L., Cusack, R., Anello, M. & Owen, A. M. A commonneural code
for similar conscious experiences in different individuals. Proc. Natl
Acad. Sci. USA 111, 14277–14282 (2014).
60. Assouline, A. & Mendelsohn, A. Weaving a story: Narrative formation
over prolonged time scales engages social cognition and
frontoparietal networks. Eur J Neurosci 57, 809–823 (2019).
61. Ames, D. L., Honey, C. J., Chow, M. A., Todorov, A. & Hasson, U.
Contextual alignment of cognitive and neural dynamics. J. Cogn.
Neurosci. 27, 655–664 (2015).
62. Yeshurun, Y. et al. Same story, different story: the neural
representation of interpretive frameworks. Psychol. Sci. 28, 307–319
(2017).
63. Anderson, D. R. & Davidson, M. C. Receptive versus interactive
video screens: a role for the brain’s default mode network in learning
from media. Comput. Hum. Behav. 99, 168–180 (2019).
64. Nakano, T., Kato, M., Morito, Y., Itoi, S. & Kitazawa, S. Blink-related
momentary activation of the default mode network while viewing
videos. Proc. Natl Acad. Sci. USA 110, 702–706 (2013).
65. Nakano, T. The right angular gyrus controls spontaneous eyeblink
rate: a combined structural MRI and TMS study. Cortex 88, 186–191
(2017).
66. Bernard, F. et al. The ventral attention network: the mirror of the
language network in the right brain hemisphere. J. Anat. 237,
632–642 (2020).
67. Igelstrom, K. M. & Graziano, M. S. A. The inferior parietal lobule and
temporoparietal junction: a network perspective. Neuropsychologia
105,70–83 (2017).
68. Freedman, L., Zivan, M., Farah, R. & Horowitz-Kraus, T. Greater
functional connectivity within the cingulo-opercular and ventral
attention networks is related to better fluent reading: a resting-state
functional connectivity study. Neuroimage Clin. 26, 102214 (2020).
69. Wang, Y. et al. Learning to read may help promote attention by
increasing the volume of the left middle frontal gyrus and enhancing
its connectivity to the ventral attention network. Cereb. Cortex 33,
2260–2272 (2023).
70. Sturm, W., Schnitker, R., Grande, M., Huber, W. & Willmes, K.
Common networks for selective auditory attention for sounds and
words? An fMRI study with implications for attention rehabilitation.
Restor. Neurol. Neurosci. 29,73–83 (2011).
71. Mort, D. J. et al. The anatomy of visual neglect. Brain 126, 1986–1997
(2003).
72. Richardson, D. Spatial representations activated during real-time
comprehension of verbs. Cognitive Science 27, 767–780 (2003).
73. Garnham, A. A unified theory of the meaning of some spatial
relational terms. Cognition 31,45–60 (1989).
74. Levelt, W. Speaking: From Intention to Articulation (MIT Press, 1989).
75. Jackendoff, R. The Architecture of the Language Faculty (MIT Press,
1997).
76. Chatterjee, A. Language and space: some interactions. Trends
Cogn. Sci. 5,55–61 (2001).
77. Marini, A., Martelli, S., Gagliardi, C., Fabbro, F. & Borgatti, R.
Narrative language in Williams syndrome and its neuropsychological
correlates. J. Neurolinguist. 23,97–111 (2010).
78. Fischer, M. H. & Zwaan, R. A. Embodied language: a review of the
role of the motor system in language comprehension. Q J. Exp.
Psychol. 61, 825–850 (2008).
79. Zwaan, R. A. Embodied cognition, perceptual symbols, and situation
models. Discourse Process. 28,81–88 (1999).
80. Barnes, M. A., Raghubar, K. P., Faulkner, H. & Denton, C. A. The
construction of visual-spatial situation models in children’s reading
and their relation to reading comprehension. J. Exp. Child Psychol.
119, 101–111 (2014).
81. Kravitz, D. J., Saleem, K. S., Baker, C. I. & Mishkin, M. A new neural
framework for visuospatial processing. Nat. Rev. Neurosci. 12,
217–230 (2011).
82. Yang, C. et al. Early prevention of cognitive impairment in the
community population: The Beijing Aging Brain Rejuvenation
Initiative. Alzheimers Dement 17, 1610–1618 (2021).
83. Bradley, K. M. et al. Cerebral perfusion SPET correlated with Braak
pathological stage in Alzheimer’s disease. Brain A J. Neurol. 125,
1772 (2002).
84. Guo, Q., Lu, C. & Hong, Z. Auditory verbal memory test in Chinese
elderly. Chin. Ment. Health J. 15,13–15 (2001).
85. Reitan, R. M. Validity of the Trail Making Test as an indicator of
organic brain damage. Percept. Mot. Skills 8, 271–276 (1958).
86. Berry, D. T., Allen, R. S. & Schmitt, F. A. Rey-Osterrieth Complex
Figure: psychometric characteristics in a geriatric sample. Clin.
Neuropsychol. 5, 143–153 (1991).
87. Mok, E. H. L., Lam, L. C. W. & Chiu, H. F. K. Category verbal fluency
test performance in Chinese elderly with Alzheimer’s disease.
Dement. Geriatr. Cogn. Disord. 18, 120–124 (2004).
88. Gillam, S. L., Gillam, R. B., Fargo, J. D., Olszewski, A. & Segura, H.
Monitoring indicators of scholarly language: a progress-monitoring
instrument for measuring narrative discourse skills. Commun.
Disord. Q. 38,96–106 (2017).
89. Hao, Y. et al. A narrative evaluation of mandarin-speaking children
with language impairment. J. Speech Lang. Hear Res. 61, 345–359
(2018).
90. Schaefer, A. et al. Local-Global parcellation of the human cerebral
cortex from intrinsic functional connectivity MRI. Cerebral Cortex 29,
3095–3114 (2018).
91. Alves, P. N., Forkel, S. J., Corbetta, M. & Thiebaut de Schotten, M.
The subcortical and neurochemical organization of the ventral and
dorsal attention networks. Commun. Biol. 5, 1343 (2022).
92. Wen, X., Yao, L., Liu, Y. & Ding, M. Causal interactions in attention
networks predict behavioral performance. J. Neurosci. 32,
1284–1292 (2012).
93. Brown, C. A., Schmitt, F. A., Smith, C. D. & Gold, B. T. Distinct
patterns of default mode and executive control network circuitry
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 14
Content courtesy of Springer Nature, terms of use apply. Rights reserved

contribute to present and future executive function in older adults.
Neuroimage 195, 320–332 (2019).
94. Sestieri, C., Shulman, G. L. & Corbetta, M. The contribution of the
human posterior parietal cortex to episodic memory. Nat. Rev.
Neurosci. 18, 183–192 (2017).
95. Landis, J. R. & Koch, G. G. An application of hierarchical kappa-type
statistics in the assessment of majority agreement among multiple
observers. Biometrics 33, 363–374 (1977).
96. McDowell, A. From the help desk: seemingly unrelated regression
with unbalanced equations. STATA J. 4, 442–448 (2004).
97. Vadas, P., Kleinman, P., Sharpley, A. & Turner, B. Relating soil
phosphorus to dissolved phosphorus in runoff: a single extraction
coefficient for water quality modeling. J. Environ. Qual. 34, 572–580
(2005).
98. Kleinberg, J. M. Authoritative sources in a hyperlinked environment.
J. ACM 46, 604–632 (1999).
99. Dice, L. R. Measures of the amount of ecologic association between
species. Ecology 26, 297–302 (1945).
100. Whitfield-Gabrieli, S. & Nieto-Castanon, A. Conn: a functional
connectivity toolbox for correlated and anticorrelated brain
networks. Brain Connect 2, 125–141 (2012).
101. Nicolini, C., Bordier, C. & Bifone, A. Community detection in
weighted brain connectivity networks beyond the resolution limit.
Neuroimage 146,28–39 (2017).
102. Feng, J. et al. A cognitive neurogenetic approach to uncovering the
structure of executive functions. Nat. Commun. 13, 4588 (2022).
103. Vos de Wael, R. et al. BrainSpace: a toolbox for the analysis of
macroscale gradients in neuroimaging and connectomics datasets.
Commun. Biol. 3, 103 (2020).
104. Yarkoni, T., Poldrack, R. A., Nichols, T. E., Van Essen, D. C. & Wager,
T. D. Large-scale automated synthesis of human functional
neuroimaging data. Nat. Methods 8, 665–670 (2011).
Author contributions
L.Y.M. and L.X. conceived the study. L.Y.M. participated in the data
collection. L.Y.M. carried out statistical analysis. L.Y.M. analyzed and
interpreted the data. L.Y.M. wrote the manuscript. L.X. revised the
manuscript. Z.Z.J. and Z.J.Y. supervised and coordinated the study. All
authorscontributed to the improvement of this manuscript and approvedthe
final version for submission.
Competing interests
The authors declare no competing interests.
Additional information
Supplementary information The online version contains
supplementary material available at
https://doi.org/10.1038/s42003-025-07501-5.
Correspondence and requests for materials should be addressed to Xin Li
or Zhanjun Zhang.
Peer reviewinformation Communicat ions Biology thanks Sahba Besharati
and the other, anonymous, reviewer(s) for their contribution to the peer
review of this work. A peer review file is available.
Reprints and permissions information is available at
http://www.nature.com/reprints
Publisher’s note Springer Nature remains neutral with regard to
jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as long
as you give appropriate credit to the original author(s) and the source,
provide a link to the Creative Commons licence, and indicate if changes
were made. The images or other third party material in this article are
included in the article’s Creative Commons licence, unless indicated
otherwise in a credit line to the material. If material is not included in the
article’s Creative Commons licence and your intended use is not permitted
by statutory regulation or exceeds the permitted use, you will need to
obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://creativecommons.org/licenses/by/4.0/.
© The Author(s) 2025
https://doi.org/10.1038/s42003-025-07501-5 Article
Communications Biology | (2025) 8:104 15
Content courtesy of Springer Nature, terms of use apply. Rights reserved

1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com