Participants with cognitive impairment (CI) were from the memory clinic of the department of neurology, Huashan Hospital, Shanghai, China from December 2020 to May 2021. Participants met the inclusion criteria if they (1) visited the memory clinic due to memory complaints from themselves or the proxy; (2) were diagnosed with AD clinical syndrome or MCI; (3) had the ability to cooperate with neuropsychological tests and physical examinations; (4) completed the olfactory assessment; (5) 60 years or older; and (6) agreed to the blood draw.

The Shanghai Aging Study (SAS) is a community-based cohort in downtown Shanghai, China. SAS aimed to explore the prevalence, incidence, and risk factors for MCI and dementia among older residents. The detailed recruitment procedure and study design have been published elsewhere (Ding et al., 2014). In the SAS, the participants were diagnosed with NC from the third wave of follow-up between June 2020 and November 2021 if they (1) completed the olfactory assessment; (2) had the ability to cooperate with neuropsychological tests and physical examinations; (3) 60 years or older; and (4) agreed to the blood draw.

We excluded participants who (1) had histories of rhinal or paranasal sinuses diseases; (2) underwent maxillofacial surgery; (3) had chronic sinusitis, asthma, chronic obstructive pulmonary disease, or acute upper respiratory tract infection within 7 days before the olfactory assessment; and (4) alcohol or drug abuse.

The lifestyle and demographic characteristics were collected from the participants and/or proxy through questionnaires, including gender, age, and years of education. Hypertension and diabetes mellitus were confirmed by the medical records (Ding et al., 2014).

Olfactory identification was a procedure in which a participant needed to accurately identify olfactory stimuli using alternative choices. Olfactory identification test was assessed by the SSST-12, which includes 12 common odors (orange, cinnamon, leather, banana, peppermint, liquorice, lemon, cloves, coffee, rose, pineapple, and fish) presented on felt-tip sticks (Hummel et al., 2001). The SSST-12 was devised by G. Kobal in Erlangen, Germany, and it is a portable, rapid (approximately 6 min), suited for inexpensive and repetitive screening of olfactory identification. The administrator of SSST-12 was blind for the cognitive diagnosis of each participant. Before the test, participants were reminded to stay away from chewing sweets, cigarettes or gum. Testing was performed in an air-conditioned and quiet room. A brief history was recorded, including questions related to the participant’s previous diseases, olfactory experience, occupation, drug intake and smoking habits. When presenting the odors, the administrator was wearing cotton gloves. The opened odor sticks were positioned about 2 cm in front of both nostrils of each participant. Participants were then asked to sniff for no longer than 3–4 s and to choose one of four answers from a list that described the best odor. An interval of 30s was set between the different sticks. Detailed instructions of SSST-12 were reported elsewhere (Liang et al., 2016).

Comprehensive neuropsychological tests were administered by the certified psychometrists in accordance with the education level of each participant. All tests conducted had been translated, adapted, and validated within the local Chinese population.

Each participant from the memory clinic received a battery of neuropsychological tests including (Xiao et al., 2021): (1) Montreal Cognitive Assessment-Basic (MoCA-B); (2) MMSE; (3) Auditory Verbal Learning Test; (4) Boston Naming Test; (5) Rey-Osterrieth Complex Figure test; (6) Symbol Digit Modalities Test; and (7) Trail-making test A&B. For those who were unable or refused to complete the whole battery of tests, only MoCA-B and MMSE were administered. Because the MMSE is less sensitive for MCI detection (Nasreddine et al., 2005), MMSE was used together with MoCA-B to discriminate MCI and dementia.

As for the participants from SAS, a battery of similar neuropsychological tests was administered due to the study design of SAS (Zhang et al., 1990; Ding et al., 2015): (1) MMSE; (2) Auditory Verbal Learning Test; (3) Conflicting Instructions Task (Go/No Go Task); (4) Modified Fuld Object Memory Evaluation; (5) Modified Common Objects Sorting Test; (6) RMB (Chinese currency) test; (7) Stick Test; and (8) Trail-making test A&B.

Two study neurologists, one neuroepidemiologist and one neuropsychologist reviewed the medical, functional, neuropsychological, neurological, and psychiatric data and reached a consensus regarding the absence or presence of dementia using DSM-IV criteria (American Psychiatric Association, 1994). Probable AD was diagnosed using the NINCDS-ADRDA criteria (McKhann et al., 1984). Participants who met the criterion of probable AD were regarded as having AD clinical syndrome. The diagnosis of MCI was based on Petersen’s criteria (Petersen, 2004).

Blood was collected from the study participants. Serum and plasma samples were centrifuged, aliquoted, and stored at −80°C. Serum TRPA1 were quantified using the Enzyme-Linked Immunosorbent Assay test on the Enzyme labeling apparatus (352 Labsystems Multiskan MS), and serum samples were diluted at a 1:4 ratio following the manufacturer’s instructions. Duplicate measurements were taken for calibrators and quality controls. Sample measurements were conducted in a single run using kits with the identical lot numbers. Operators remained unaware of participants’ characteristics. The statistical median value of serum TRPA1 was used to categorize participants into low or high levels of serum TRPA1.

Continuous variables were expressed as the mean (standard deviation) or median (lower quartile[25%], upper quartile[75%]), and categorical variables were expressed as frequencies (%). The Wilcoxon rank-sum test and Pearson chi-square test were used to compare continuous and categorical variables.

Scatter plots were performed to analyze the correlations between SSST-12 and serum TRPA1, between SSST-12 and MMSE, and between serum TRPA1 and MMSE. Their correlations were assessed using the generalized linear model with the adjustment for gender, age, and years of education.

The logistic regression model was used to detect the association between serum TRPA1, the level of serum TRPA1, SSST-12, olfactory dysfunction, or the identification of peppermint and CI after adjusted for gender, age, and years of education. Risk was presented as odds ratio (OR) and 95% confidence interval (95%CI).

Mediation analysis was conducted to decompose the total effect of olfactory function on cognitive function into a natural direct effect and a natural indirect effect through TRPA1 after adjusted for gender, age, and years of education.

Two-tailed tests were used to estimate all p values and 95% confidence intervals. Statistically significant differences were identified at p < 0.05. Mediation analysis was conducted using R packages “lavaan” and “mediation.” Data analysis was conducted in SAS 9.4 (SAS Institute Inc., Cary, NC, United States) and R Software (version 4.1.2).

We recruited 256 participants (164 females and 92 males), including 135 participants with NC and 121 participants with CI. Table 1 displayed the characteristics of the study participants. Significant differences were observed in the level of serum TRPA1 (p = 0.0086), education year (p = 0.0002), MMSE (p < 0.0001), and SSST-12 (p < 0.0001) between the two groups. The CI groups had higher proportion of the high level of serum TRPA1 (58.7%), the lower median education year (median = 11), the lower median MMSE score (median = 27), and the lower median SSST-12 score (median = 7).

SSST-12 were inversely correlated with serum TRPA1 in total participants (β = −8.0848, 95%CI: −12.6423, −3.5272), participants with NC (β = −7.0532, 95%CI: −13.9778, −0.1286) and participants with CI (β = −7.2309, 95%CI: −13.8897, −0.5721) after adjusted for gender, age, and years of education (Figure 1). SSST-12 were positively correlated with MMSE in total participants (β = 0.7217, 95%CI: 0.5454, 0.8980) and participants with CI (β = 0.9544, 95%CI: 0.6528, 1.2559) after adjusted for gender, age, and years of education (Figure 2). The serum TRPA1 level was inversely correlated with MMSE in total participants (β = −0.0088, 95%CI: −0.0138, −0.0037) and participants with CI (β = −0.0093, 95%CI: −0.0182, −0.0003) after adjusted for gender, age, and years of education (Figure 3).

As shown in Table 2, the higher serum TRPA1 level was associated with an increased risk for participants with CI (OR = 1.889, 95%CI: 1.125, 3.172) after adjusted for gender, age, and years of education. Lower SSST-12 score (OR = 0.776, 95%CI: 0.685, 0.878) and olfactory dysfunction (OR = 3.116, 95%CI: 1.592, 6.101) were associated with an increased risk for participants with CI after adjusted for gender, age, and years of education. The correct identification of peppermint may be a protector for participants with CI (OR = 0.858, 95%CI: 0.428, 1.719), although the finding did not reach statistical significance.

Figure 4 showed the total, direct and indirect effects for the mediating role of TRPA1 on the relationship between olfactory function and cognitive function in mediation models. As shown in Figure 4B, after adjusted for gender, age, and years of education, the estimated average causal mediated effect (ACME) (indirect effect estimate β = 0.0402, 95%CI: 0.0023, 0.0906), average direct effect (ADE) (direct effect estimate β = 0.6816, 95%CI: 0.3906, 0.9630), and total effects (total effect estimate β = 0.7217, 95%CI: 0.4193, 1.0015) were all statistically significant, suggesting a partial mediation effect of TRPA1 in the association between olfactory function and MMSE, and TRPA1 had the mediation effect with a proportion of mediation up to 5.57%.After adjusted for gender, age, and years of education, the estimated ADE (direct effect estimate β = −0.0513, 95%CI: −0.0741, −0.0275) and total effects (total effect estimate β = −0.0543, 95%CI: −0.0766, −0.0301) were statistically significantly different from zero, but the estimated ACME were not (Figure 4D). The results suggested that lower SSST-12 may be associated with higher serum TRPA1, which in turn made participants more likely to have worse cognitive function.

Figure 5 showed the total, direct and indirect effects for the mediating role of TRPA1 on the relationship between the identification of peppermint and cognitive function in mediation models. As shown in Figure 5B, after adjusted for gender, age, and years of education, the estimated ACME (indirect effect estimate β = 0.2975, 95%CI: 0.0443, 0.6200), ADE (direct effect estimate β = 2.8716, 95%CI: 0.9727, 4.9200), and total effects (total effect estimate β = 3.1691, 95%CI: 1.2178, 5.3100) were all statistically significant, suggesting a partial mediation effect of TRPA1 in the association between the identification of peppermint and MMSE, and TRPA1 had the mediation effect with a proportion of mediation up to 9.39%. However, the estimated ACME, the estimated ADE and total effects were all not statistically significant (Figure 5D). The results suggested that the incorrect identification of peppermint may be associated with higher serum TRPA1, which in turn made participants more likely to have worse cognitive function.