Differences in olfactory functional connectivity in early-onset depression and late-onset depression

Abstract Background Late-onset depression (LOD) and early-onset depression (EOD) exhibit different pathological mechanisms and clinical phenotypes, including different extents of olfactory dysfunction. However, the brain abnormalities underlying the differences in olfactory dysfunction between EOD and LOD remain unclear. Objective The aim of this study was to compare the functional connectivity (FC) patterns of olfactory regions between EOD patients and LOD patients and examine their relationship with cognitive function. Methods One hundred and five patients with EOD, 101 patients with LOD and 160 normal controls (NCs) were recruited for the present study. Participants underwent clinical assessment, olfactory testing, cognitive assessments, and magnetic resonance imaging. Eight regions of the primary and secondary olfactory regions were selected to investigate olfactory FC. Results Patients with LOD exhibited decreased odor identification (OI) compared with patients with EOD and NCs. The LOD group exhibited decreased FC compared with the EOD and NC groups when primary and secondary olfactory regions were selected as the regions of interest (the piriform cortex, lateral entorhinal cortex, and orbital-frontal cortex). Additionally, these abnormal olfactory FCs were associated with decreased cognitive function scores and OI, and the FC between the left orbital-frontal cortex and left amygdala was a partial mediator of the relationship between global cognitive scores and OI. Conclusion Overall, patients with LOD exhibited decreased FC in both the primary and secondary olfactory cortices compared with patients with EOD, and abnormal olfactory FC was associated with OI dysfunction and cognitive impairment. The FC between the orbital-frontal cortex and amygdala mediated the relationship between global cognitive function and OI.


Introduction
Late-life depression (LLD) is one of the major risks of disability and dementia in the geriatric population, and it affects 3.68-4.60% of elderly people each year (Invernizzi et al., 2021 ).An increasing number of studies have suggested that people with late-onset depression (LOD) (the first de pressi ve e pisode occurs after age 60) and early-onset depression (EOD) (the first de pressi ve e pisode occurs before age 60) exhibit different responses to antidepressant and electr oconvulsiv e ther a py (Dols et al., 2017 ), and the risk of de v eloping dementia is higher in patients with LOD than in patients with EOD (Lee et al., 2021 ), which may result from the various pathological mechanisms of the tw o disor ders.Specifically, LOD is mor e closel y r elated to neur odegener ation, micr ov ascular d ysfunction, strok e, and other patholog ical ag ing processes, whereas EOD is more strongly associated with genetic susceptibility and adverse life e v ents (Choi et al., 2017 ;P ar anthaman et al., 2012 ).Compared with patients with EOD, patients with LOD exhibited slo w er information pr ocessing speed, poor er executiv e function (Cheng et al., 2020 ), nerve growth factor deficiency (Shi et al., 2010 ), higher inflammatory le v els (Perna et al., 2022 ), increased cortical A β burden (Lour eir o et al., 2020 ), more hippocampal atrophy (Ballmaier et al., 2008, Gao et al., 2022 ), more white matter hypersignal (Choi et al., 2017 ), and different modular organization of the functional brain network (Mai et al., 2021 ).
Aging is a risk factor for olfactory system deterioration, and a meta-analysis indicated that olfactory dysfunction starts in the fifth decade of life in healthy people (Zhang and Wang, 2017 ).Odor identification (OI) is a predictor of Alzheimer's disease (AD) because it r epr esents earl y pathological c hanges in the entorhinal cortex and hippocampus (Lafaille-Magnan et al., 2017 ), and its pr edictiv e effect has been shown in amnestic mild cognitive impairment (MCI) patients and community elderly adults (Murphy, 2019 ).Furthermor e, our pr e vious studies indicate that OI dysfunction may contribute to the risk of de v eloping dementia in patients with LLD, indicating that there are more severe impairments in cognitive function and the ability to perform activities of daily living, and suggesting mor e se v er e structur al and functional br ain abnormalities in LLD patients with OI dysfunction than in those with normal OI (Chen et al., 2018 ;Chen et al., 2021 ).Inter estingl y, our pr e vious study also suggested that the olfactory pr ocessing of patients with EOD and LOD may be different, that LOD patients exhibit worse OI than EOD patients, and that the variations in OI among patients can be attributed to differences in their memory and language function (Liu et al., 2022 ).T herefore , this evidence indicates that there may be an interactive effect between LLD and age of first de pressi ve e pisode on OI, and this effect should be considered when using OI to predict dementia risk in LLD patients.Ho w e v er, the pathological mec hanism underl ying the different OIs between EOD and LOD remains unclear.
OI pr ocesses include peripher al perception of the primary olfactory cortex (such as the piriform cortex) and high-order processing of the secondary olfactory cortex [such as the orbitalfrontal cortex (OFC) and insular cortex] (Rai et al., 2021 ).Because the olfactory pathway str ongl y ov erla ps with the cognitiv e ma p (Murphy, 2019 ), disrupted activity and functional connectivity (FC, the statistical correlation or synchronization of neural activity between br ain r egions that pr ovides insights into their functional relationships and communication) of the olfactory regions are associated with not only OI dysfunction but also cognitive impairment (Chen et al., 2022 ;Lu et al., 2019 ).Additionally, our previous studies suggested that LLD patients exhibited disrupted olfactory FC compared with healthy controls .Furthermore , abnormal FCs of olfactory r egions ar e associated with OI, global cognition, and language function in patients with LLD (Yang et al., 2022 ).Ho w e v er, it remains unclear whether abnormal olfactory FC is involved in the differences in OIs and cognitive functions between EOD and LOD.
Based on this evidence, we hypothesized that patients with LOD exhibit decreased FC in both primary and secondary olfactory regions compared with patients with EOD, and this decreased FC ma y be in volv ed in the differ ence in OI and cognitive function between patients with EOD and those with LOD.Ther efor e, the present study aimed to (i) compare olfactory FC patterns between EOD and LOD patients and (ii) explore the relationship among olfactory FC, OI dysfunction, and cognitive impairment in EOD and LOD patients .T he present study provides an in-depth investigation of the different pathological mechanisms between EOD and LOD and contributes to a r easonable a pplication of OI for pr edicting dementia risk in patients with LLD.

Participants
Two hundred and six patients with LLD were recruited from the Affiliated Brain Hospital of Guangzhou Medical University, and 160 normal controls (NCs) matched for age and gender were recruited from communities in Guangzhou.All participants or rele v ant legal guardians provided written informed consent to participate in the study.The study protocol and assessments were a ppr ov ed by the Ethics Committees of the Affiliated Brain Hospital of Guangzhou Medical University.
Patients with LLD were included in the present study according to the following inclusion criteria: (i) age > 55 years, (ii) major depr ession dia gnosed according to DSM-IV criteria, and (iii) clinical staging and diagnosis made by at least two neurologists with expertise in dementia, a neur opsyc hologist, and a geriatric psyc hiatrist.NCs wer e included if they (1) were right-handed, (ii) exhibited normal cognitive function, and (iii) had no past history of depression.The exclusion criteria for both LLD patients and NCs w ere as follo ws: (i) a history of other major psychiatric disorders, such as bipolar disorder or schizophrenia; (ii) physical illnesses, such as anaemia or h ypoth yroidism, that could lead to depressi ve e pisodes; (iii) neurological disorders, such as brain tumours, P arkinson's disease, m ultiple scler osis, and str ok e; (i v) current or pr e vious psyc hiatric symptoms; (v) head injury with loss of consciousness > 30 min; and (vi) other conditions significantly affecting the sense of smell, including active upper airway/sinus infection or dyspnoea at the time of testing, traumatic or congenital olfactory impairment (known nasal tumours or pol yps), curr ent or recent (in the last 6 months) smoking, and alcohol abuse or substance dependence.For further analyses, patients with LLD were divided into two subgr oups: the EOD gr oup, whose first depr essiv e episode occurred before age 60, and the LOD group, whose depressi ve e pisode occurr ed after the a ge of 60.

Neuropsychological assessments
P articipants underwent compr ehensiv e neur opsyc hological testing to e v aluate v arious domains of cognition after completing a standard clinical assessment.(i) The Mini-Mental State Examination (MMSE) (scor e r ange 0-30) was used to assess global cognitive function (Tombaugh and McIntyre, 1992 ); (ii) the Auditory Verbal Learning Test (AVLT) N5 was used to assess memory (score range 0-12) (Hawkins et al., 2004 ); (iii) the Trail-Making Test B was used to assess executive function with scores depending on the number of seconds it took for the participant to complete the test (Tombaugh, 2004 ); (iv) the Boston Naming Test (BNT) was used to assess language function with scores ranging from 0-30 (Bezdicek et al., 2021 ); and (v) the Symbolic Digit Transformation Test (SDMT) was used to assess information processing speed, with scores depending on the number of symbols r ecorded corr ectl y within 90 s (Fellows and Sc hmitter-Edgecombe, 2019 ).Additionall y, the 17item Hamilton Depression Rating Scale (HAMD-17) was used to e v aluate the se v erity of depr ession (Sc hwab et al., 1967 ).All scale assessments were completed by two trained professional psychiatrists who passed a concordance assessment.

Olfactory assessments
Before the olfactory test, participants were asked to complete a questionnaire that excluded factors that could affect olfactory function (i.e.history of nasal trauma and related surgery, history of radiation or chemotherapy, nasal congestion, etc.).OI was assessed using the Sniffin' Stic ks Scr een 16 test for olfactory assessment (Perna et al., 2022 ).Participants were asked to smell 16 common odors in a specific order and were asked to select one image out of four that best matched the odor smelled; only one of the four images was corr ect.P articipants wer e giv en scor es r anging from 0 to 16.The experiment was conducted in a quiet, wellv entilated r oom fr ee of odors.All participants who completed the neur opsyc hological assessment wer e giv en the OI test on the same day.

MRI data acquisition and processing
P articipants underwent ma gnetic r esonance ima ging (MRI) scans after the neur opsyc hological assessments.A Philips 3.0 T MR system (Ac hie v a, Netherlands) at The Affiliated Br ain Hospital of Guangzhou Medical University was used to acquire the imaging data.For each participant, an anatomical image was obtained with a sagittal 3D gradient echo.Sagittal resting-state functional MRI (fMRI) datasets of the whole brain were obtained in 8 min with a single-shot gradient echo-planar imaging pulse sequence.The resting-state fMRI scanning parameters were as follows: echo time (TE) = 30 ms; repetition time (TR) = 2000 ms; flip angle = 90 • ; number of slices = 33; slice thickness = 4 mm; matrix size = 64 × 64; and field of view = 220 × 220 mm.The pr epr ocessing of r esting-state fMRI datasets was carried out using the Data Processing Assistant for Resting-State 5.1 (Yan et al., 2016 ).The first 10 volumes were discarded to ensure steadystate data.The remaining 230 images were corrected for timing differences and for head motion.A record of participants' head motions was provided after realignment correction.Participants who had > 2 mm maximum displacement in any plane, 2 • of angular motion, and 0.2 mm mean fr ame wise displacement were excluded from further analysis.To minimize the influence of head motion, the mean fr ame wise displacement of each participant was r egr essed out in the gr oup-le v el anal ysis .T hen, the motion-corr ected ima ges wer e spatiall y normalized into the standard Montr eal Neur ological Institute (MNI) Ec ho Planar Ima ging template, resliced to a voxel size of 3 × 3 × 3 mm 3 resolution, and smoothed using a 6 mm full-width at half-maximum Gaussian kernel.Linear tr ends wer e r emov ed fr om the data, and nuisance cov ariates wer e then r egr essed out fr om eac h time series, including signals indicating white matter and cer ebr ospinal fluid as well as the Friston-24 parameters of head motion.To reduce the effect of lo w-frequenc y drifts and high-frequenc y noise, a bandpass filter (0.01 Hz < f < 0.1 Hz) was applied.

Seed-based FC analysis
Regions of interest (ROI) with a radius of 6 mm centred around eight MNI coor dinates w er e selected.The r egions included the left and right piriform cortices [( −22, 0, 14), (22, 2, 12)], the left and right OFC [( −24, 30, 10), (28, 34, 12)], and the left and right insula [( −30, 18, 6), (28, 16, 8)]; these r egions wer e obtained fr om a meta-analysis that identified these three left and right brain regions as those most likely to be activated by olfactory stimulation (Seubert et al., 2013 ).In addition, the lateral entorhinal cortex (LEC) w as selected accor ding to the mask reported in the study by Maass et al.(Maass et al., 2015 ).The time courses of all voxels within each R OI w ere extracted and averaged as the seed point r efer ence time courses.Individual-le v el FC ma ps of the eight ROI w ere obtained b y calculating P earson's correlation coefficients between the mean time courses of the ROI and the time courses of each voxel in the whole brain.Then, Fisher's r -to-z transformation was performed to impr ov e normality.The z scor e FC ma ps wer e used for gr oup-le v el statistical analyses.

Statistics
Statistical P ac ka ge for the Social Sciences version 25.0 (IBM SPSS v.25.0) was used to perform the statistical analyses .T he results were visualized with GraphPad Prism v.9.0.0 and the BrainNet View er ( http://www.nitr c.org/ projects/ bnv/ ) (Xia et al., 2013 ).The neur opsyc hological differ ences among participants wer e e v aluated by one-way analysis of covariance (ANCOVA).Post hoc com-parisons were made using the post hoc least significant difference test.ANCOVA was used to assess the z scor e ma ps for each ROI to detect significant differences among the three groups.Gaussian random field theory was used for multiple comparison corr ection (voxel P v alue < 0.001; cluster P < 0.05).The mean z scores of significantly different regions among groups in ANCOVA were extracted for post hoc analysis (significance level 0.05, least significant difference correction).Age, gender, HAMD scores, and framewise displacement were chosen as co variates .T he associations between OI and cognitive test scores were evaluated by partial corr elation anal yses.Stepwise m ultiple linear r egr ession was used to analyse the effects of different FCs on OI scores and the effects of v arious cognitiv e scor es on differ ent FCs.Mediation anal yses wer e performed to investigate the potential relationship between cognitiv e scor e (independent v ariable) and OI (dependent v ariable).FC v alues wer e r egarded as mediators, and we calculated the mediation model in PROCESS v.3.4 running in an SPSS v.25 environment.The le v el of confidence for all confidence intervals in the output was 95% with 5000 bootstr a p samples.Effects wer e considered significant if 0 was not contained in the 95% confidence intervals .Age , gender, and HAMD scores were chosen as covariates in the partial corr elation, r egr ession, and mediation anal yses.
When LLD patients were divided into the EOD group and LOD gr oup, ther e wer e significant differ ences in the HAMD, OI, MMSE, A VLT N5, TMTB, BNT , and SDMT scores between the EOD group and the NC group and between the LOD group and the NC group and the NC group ( P < 0.05); no significant difference was found in TMTB scores among the three groups ( P > 0.05).In the post hoc anal yses, both the EOD gr oup and LOD gr oup exhibited higher HAMD scores and lo w er MMSE, AVLT N5, and SDMT scores than the NC gr oup.Additionall y, the LOD gr oup exhibited lo w er OI and BNT scores than the NC and EOD groups.No significant difference in HAMD scores was found between the EOD group and LOD group ( P > 0.05) (Table 1 ).

Differences in olfactory FC among the EOD , LOD , and NC groups
In total, 88 health NC participants, 82 LOD patients, and 50 EOD patients a gr eed to under go the neur oima ging assessments.In the ANCOVA of FC, significant differ ences wer e found among the EOD,  In the post hoc comparison, the LOD group exhibited significantl y decr eased FC v alues compar ed with the EOD gr oup and NC group when the bilateral piriform cortex, bilateral LEC, and left OFC were selected as the seeds.Additionally, both the EOD and LOD gr oups exhibited significantl y decr eased FC v alues compar ed with the NC group when the left and right insula were selected as the seeds (Fig. 2 ).

Associa tion betw een olfactory FC and OI
P artial corr elation anal yses suggested that OI was associated with FC between the left piriform cortex and left middle cingulum ( r = 0.182, P = 0.012), FC between the left piriform cortex and right superior temporal gyrus ( r = 0.153, P = 0.035), FC between the right piriform cortex and right anterior cingulum ( r = 0.147, P = 0.043), FC between the left OFC and left amygdala ( r = 0.239, P = 0.001), FC between the right insula and left medial superior frontal gyrus ( r = 0.238, P = 0.001), FC between the right insula and left inferior temporal gyrus ( r = 0.182, P = 0.012), FC between the left LEC and right inferior cerebellum ( r = 0.156, P = 0.032), and FC between the left LEC and left middle temporal gyrus ( r = 0.149, P = 0.040).

Associa tion betw een olfactory FC and cogniti v e function
When various FCs w ere regar ded as the dependent variables and v arious cognitiv e scor es wer e designated as the independent variables, the SDMT score was the most highly associated variable in the model of the FC between the left piriform cortex and left middle cingulum, FC between the left piriform cortex and right praecuneus, FC between the left piriform cortex and right superior temporal gyrus, FC between the right piriform cortex and right anterior cingulum, FC between the left LEC and left middle temporal gyrus, and FC between the left LEC and left temporal pole.Additionall y, the BNT scor e was most str ongl y associated with FC between the right piriform cortex and right temporal pole, and the MMSE score was most str ongl y associated with FC between the left OFC and left amygdala (Table 3 , Fig. 3 ).

Mediating effect of olfactory FC on the rela tionship betw een OI and cognition
The total effect of MMSE scores on OI was β = 0.379 ( P < 0.001), and the indirect effect of MMSE scores on OI mediated by the FC between the left OFC and left amygdala was β = 0.034 (BootLLCI = 0.002, BootULCI = 0.045).Furthermore, the remaining direct effect of MMSE scores on OI was β = 0.345 ( P < 0.001), with the effect of MMSE scores on FC between the left OFC and left amygdala being β = 0.210 ( P = 0.012) and the effect of FC between the left OFC and left amygdala on OI being β = 0.161 ( P = 0.015).In summary, these results revealed that the FC between the left OFC and left amygdala was a mediator of the relationship between MMSE scores and OI.No significant mediating effect was found in FC values on the association between OI and the other cognitiv e scor es (Fig. 4 ).

Discussion and conclusion
The present study compared the OI and olfactory FC of patients with EOD and LOD and explored the relationship among olfactory FC, OI d ysfunction, and cogniti v e impairment.The main r esults are as follows: (i) the LOD group exhibited worse OI than the EOD group; and (ii) the LOD group exhibited decreased FC compared with the EOD group and NC group when the piriform cortex, LEC, and left OFC were selected as the seeds.Additionally, both the EOD and LOD groups exhibited significantly decreased FC compared with the NC group when the insula cortex was selected as the seed.(iii) Olfactory FCs were associated with OI and cognitive function (especially information processing speed), and FC between the left OFC and left amygdala mediated the relationship between global cognitive function and OI.
LLD accompanied by OI dysfunction and olfactory impairment is associated with more severe cognitive deficits, impairments in activities of dail y living, and structur al and functional brain dama ge compar ed to LLD with intact olfaction, suggesting that individuals with LLD and olfactory impairment may exhibit a higher risk of de v eloping dementia (Chen et al., 2018 ;Chen et al., 2021 ).Ho w e v er, whether the age of the first episode of depression influences OI in patients with LLD has not been fully explained.Consistent with our pr e vious studies, the present study with a larger sample size demonstrated that the OI score was lower in patients with LOD than in patients with EOD and controls, suggesting that abnormalities in olfactory pr ocessing ar e mor e obvious in patients with LOD (Liu et al., 2022 ).In the cognitive function comparison, both EOD and LOD patients exhibited impaired global cognitive function, information processing speed, and memory, which was consistent with pr e vious r esults in LLD patients.Inter estingl y, the langua ge deficit exhibited a similar tendency to OI: it was found in LOD patients but not EOD patients.OI pr ocesses not onl y r el y on odor perception but also involv e man y higher cognitiv e pr ocesses (Dulay et al., 2008 ), particularly in naming (Liu et al., 2022 ).Ther efor e, the worse OI and naming performance in LOD patients may reflect a more se v er e dysfunction in ov erla pping r egion r esponses for olfaction and language .T herefore , the current results indicate that there ar e inter activ e effects of age and depression on OI, and the age of the first episode of depression needs to be considered when a ppl ying OI dysfunction to predict dementia risk in LLD patients.
Similar to the psychophysical results of OI, the LOD group exhibited decreased FC in the primary and secondary olfactory regions compared with the EOD and HC groups, and the olfactory FCs were relatively intact in the EOD group.Additionally, these decreased olfactory FCs were associated with OI dysfunction and cognitiv e impairment.OI r equir es not onl y peripher al perception of odor molecules but also high-order processing of olfactory information (Rai et al., 2021 ), and damage to both the primary and secondary olfactory cortices could lead to OI dysfunction (Chen et al., 2022 ).Considering that the function of many olfactory regions is also responsible for cognitive and emotional processing (Murphy, 2019 ), the worse OI and olfactory FC in the LOD group suggested that these participants suffered more severe and widespr ead br ain abnormalities associated with cognitiv e impairment and affective disorders, which is consistent with previous evidence (Ballmaier et al., 2008 ;Choi et al., 2017 ;Mai et al., 2021 ).Additionally, in AD patients, abnormalities in olfactory FC have been found, including disrupted FC between the olfactory network and the right hippocampus (Lu et al., 2019 ) and lo w er FC betw een the left OFC and right frontal area and between the right OFC and right tempor al ar ea (Lee et al., 2020 ).Inter estingl y, the pr esent r esults indicated that the pattern of disrupted olfactory FC in LOD patients was closer to that in AD patients, which is consistent with pr e vious opinions that LOD patients exhibit higher AD risk than EOD patients (P a gni et al., 2022(P a gni et al., ) (van Reekum et al., 1999 ) ).
Among various abnormal olfactory FCs, the disconnection between the OFC and amygdala seems to be the most special oc-   Both the OFC and amygdala are not only highorder olfactory regions but also centres of emotional processing (Zald and P ardo, 1997 ).A gr owing body of fundamental studies has indicated that there are strong anatomical connections between the OFC and the amygdala (Stein et al., 2007 ), and disrupted FC of the OFC-amygdala has been found both in resting-state and task-fMRI studies (Liao et al., 2010 ;Sladky et al., 2015 ).Furthermore, changes in brain activity within the left amygdala and the left OFC when receiving odor stimulation were highly intercorrelated (Zald and Pardo, 1997 ), and the disconnection of the OFC-amygdala was found to be associated with more severe depression and anxiety, suggesting that inefficient top-down modulation from the OFC to the amygdala leads to hyperactivity of fear-processing circuits to w ar d negative or threat cues (Mao et al., 2020 ).Consistently, the current study indicated that the LOD group exhibited subtl y higher depr essiv e scor es than the EOD gr oup.Mor eov er, the positive association between this FC and between OI and global cognition suggested that OFC downregulation in the amygdala is also important in olfactory and cognitiv e pr ocessing, and the results of mediation analyses suggested that cognitive impairment may facilitate OI dysfunction by interrupting FC between the OFC and amygdala.Ho w e v er, their causal relationship still needs to be verified by task-fMRI and longitudinal studies.The piriform cortex is an important part of the primary olfactory cortex, which is responsible for olfactory perception, valence, and action (Chen et al., 2022 ).It r eceiv es neur onal pr ojections fr om the olfactory bulb and is widely connected to secondary olfactory br ain r egions.Additionall y, the piriform cortex is vulner able to the neurofibrillary tangles and amyloid β seen in patients with AD (Bathini et al., 2019 ;De v anand, 2016 ).Inter estingl y, patients with mild AD and amnestic MCI exhibited r educed activ ation in the right anterior piriform cortex when receiving olfactory stimulation compared to healthy elderly individuals, and odor discrimination was associated with the activation of the right piriform cortex (Kjelvik et al., 2020 ).Furthermore, our pr e vious study suggested that FC between the right piriform cortex and right superior parietal lobule was positiv el y associated with OI scores in patients with LLD, suggesting that decreased FC of the piriform cortex was involved in OI dysfunction in LLD (Yang et al., 2022 ).In the olfactory pathway, the LEC r eceiv es olfactory information from the piriform cortex and interacts with the hippocampus to encode and r etrie v e olfactory memory (Chapuis et al., 2013 ).Notably, the LEC is among the first regions to be affected by tau during AD de v elopment, and structur al and functional abnormalities in the LEC have been found to be associated with OI dysfunction (Murphy, 2019 ).In patients with LLD, the disruption of the amygdalaentorhinal-hippocampal network in LLD was found to be involved in emotional and memory processing (Leal et al., 2017 ), but its relationship with OI has not yet been explored.In the present study, both EOD patients and LOD patients exhibited decreased FC of the piriform cortex and LEC, and these disconnections were more obvious in LOD patients .T hese results indicate the mor e se v er e disconnection of the olfactory pathway with other regions in LOD patients and provide evidence of the higher risk of developing AD in LOD patients compared with that of EOD patients.
The insular cortex plays a k e y role in the integration of multimodal information and in interoce pti ve and exteroce pti ve pro-cessing (Koeppel et al., 2020 ).Olfactory input is dir ectl y tr ansmitted to the insular cortex without passing the thalamus first, and the mid-dorsal insula may be an integr ated or al sensory region that plays a critical role in flav our per ception (Mazzola et al., 2017 ).Insular dysfunction has been found in patients with hyposmia, and odor-induced responses in the insular cortex were associated with olfactory function (Han et al., 2018 ).Consistently, decreased insular FCs were associated with OI dysfunction in the present study.Ne v ertheless, ther e was no significant difference in insular FC between patients with EOD and those with LOD, although both groups exhibited decreased FC compared with the NC group.Howe v er, considering that the function of the insula varies among subr egions, futur e studies a ppl ying a mor e detailed atlas could pr ovide a clearer understanding of insular FC in patients with EOD and LOD.
Ther e wer e se v er al limitations in the pr esent study.First, the ROI for FC analyses were obtained from a meta-analysis and the LEC mask with high-solution, and these R OI w er e r egarded as most likely to be activated by olfactory stimulation.Howe v er, ther e may be other olfactory regions exhibiting different FCs between patients with EOD and those with LOD, and future studies including more ROI need to be further explored.Second, odor threshold and odor discrimination were not assessed in the present study, and future studies are needed to explore their relationships with olfactory FC.Thir d, FC w as calculated b y the association betw een tw o differ ent r egions, and futur e studies using analyses of dynamic causal models and Granger causality analysis or using olfactory task-fMRI could further clarify the direction of the abnormal connectivity in olfactory r egions.Finall y, LLD patients take various types of antidepressant in variable doses, which may be a confounding factor in the FC analyses.

Conclusion
Patients with LOD exhibited decreased FC in both the primary and secondary olfactory cortex compared with patients with EOD.Ab-normal olfactory FC was associated with OI dysfunction and cognitive impairment, and the FC between the OFC and amygdala mediated the relationship between global cognitive function and OI.The present study provides an in-depth understanding of the different pathological mechanisms between EOD and LOD, and emphasizes that the age of the first episode of depression needs to be considered when using OI dysfunction as a predictor of dementia risk in LLD patients.

Figure 1 :
Figure 1: Differences in FC among the EOD , LOD , and NC gr oups.Differ ences in FC were found ( A ) between the left piriform cortex and right superior temporal gyrus, the left piriform cortex and left middle cingulum, and the left piriform cortex and right praecuneus; ( B ) between the right piriform cortex and right temporal pole, and between the right piriform cortex and right anterior cingulum; ( C ) between the right LEC FC and left temporal pole; ( D ) between the left LEC and right inferior cerebellum, left LEC and left temporal pole, left LEC and left middle temporal gyrus, and left LEC and left supplementary motor area; ( E ) between the left OFC and left amygdala, and between the left OFC and left Heschl's gyrus; ( F ) between the left insula and left praecuneus; and ( G ) between the right insula and left inferior temporal gyrus, between the right insula and left medial superior frontal gyrus, and between the right insula and left superior frontal gyrus.The color bar r epr esents F values in ANCOVA.

Figure 2 :
Figure 2: Comparison of FC among the EOD , LOD , and NC groups.* P < 0.05; * * P < 0.01; * * * P < 0.001.( A ) The LOD group exhibited significantly decreased FC values compared with the EOD group and NC group when the bilateral piriform cortex was selected as the seed.( B ) The LOD group exhibited significantl y decr eased FC v alues compar ed with the EOD gr oup and NC gr oup when the bilater al LEC was selected as the seed.( C ) The LOD gr oup exhibited significantl y decr eased FC v alues compar ed with the EOD gr oup and NC gr oup when the left OFC was selected as the seed, and both the EOD and LOD groups exhibited significantly decreased FC values compared with the NC group when the left and right insula were selected as the seeds.Abbr e viations: Pir: piriform gyrus.Ant: anterior, Inf: inferior, Mid: middle, Sup: superior, Supp: supplementary, L: left, R: right.

Figur e 3 :
Figur e 3: T he corr elations between FC v alues and neur opsyc hological v ariables in all participants .( A ) T he correlation betw een OI and FC betw een the left OFC and left amygdala.( B ) The correlation between MMSE scores and FC between the left OFC and left amygdala.( C ) The correlation between SMDT scores and FC between the left piriform and left middle cingulum.( D ) The correlation between the SMDT scores and FC between the left piriform and right praecuneus.( E ) The correlation between SMDT scores and FC between the left piriform and right superior temporal gyrus.( F ) The correlation between SMDT scores and FC between the right piriform and right anterior cingulum.( G ) The correlation of FC between the right piriform and right temporal poles.( H ) The correlation between SMDT scores and FC between the left LEC and left middle temporal gyrus.( I ) The correlation between SMDT scores and FC between the left LEC and left temporal pole.

F igure 4 :
FC betw een the left OFC and left amygdala mediated the relationship betw een MMSE scores and OI. a sho ws the effect of MMSE scores on FC between the left OFC and left amygdala.b shows the effect of FC between the left OFC and left amygdala on OI scores.c means the direct effect of MMSE scores on OI scores.c is the total effect of MMSE scores on OI scores.

Table 1 :
Differences in OI and neuropsychological scores between EOD and LOD groups.

Table 2 :
Differences in FC among the EOD , LOD , and NC groups.

Table 3 :
Association between olfactory FC and cognitive function.