IL-1Ra^high-IL-4^low-IL-13^low: A Novel Plasma Cytokine Signature Associated with Olfactory Dysfunction in Older US Adults

Abstract

Inflammation has been implicated in physical frailty, but its role in sensory impairment is unclear. Given that olfactory impairment predicts dementia and mortality, determining the role of the immune system in olfactory dysfunction would provide insights mechanisms of neurosensory decline. We analyzed data from the National Social Life, Health and Aging Project, a representative sample of home-dwelling older US adults. Plasma levels of 18 cytokines were measured using standard protocols (Luminex xMAP). Olfactory function was assessed with validated tools (n-butanol sensitivity and odor identification, each via Sniffin’ Sticks). We tested the association between cytokine profiles and olfactory function using multivariate ordinal logistic regression, adjusting for age, gender, race/ethnicity, education level, cognitive function, smoking status, and comorbidity. Older adults with the IL-1Ra^high-IL-4^low-IL-13^low cytokine profile had worse n-butanol odor sensitivity (odds ratio [OR] = 1.61, 95% confidence interval [CI] 1.19–2.17) and worse odor identification (OR = 1.42, 95% CI 1.11–1.80). Proinflammatory, Th1, or Th2 cytokine profiles were not associated with olfactory function. Moreover, accounting for physical frailty did not alter the main findings. In conclusion, we identified a plasma cytokine signature—IL-1Ra^high-IL-4^low-IL-13^low—that is associated with olfactory dysfunction in older US adults. These data implicate systemic inflammation in age-related olfactory dysfunction and support a role for immune mechanisms in this process, a concept that warrants additional scrutiny.

Introduction

Olfactory dysfunction affects nearly a quarter of older US adults. This key sensory impairment significantly affects quality of life (Frasnelli and Hummel 2005), as well as social (Wudarczyk et al. 2015), physical, and mental health (Khil et al. 2016). Olfactory impairment has been shown to precede neurodegenerative diseases such as Alzheimer’s (Wilson et al. 2009) and Parkinson’s diseases (Wilson et al. 2008) and dementia (Lafaille-Magnan et al. 2017; Yaffe et al. 2017; Adams et al. 2018). More broadly, several studies have shown that olfactory impairment predicts mortality, with recent data suggesting that this is independent of the development of neurodegenerative diseases (Pinto et al. 2014; Ekstrom et al. 2017).

The precise etiology of age-related olfactory impairment is unclear. Immune mechanisms may play a role in olfaction due to chronic inflammation (Sultan et al. 2011). Indeed, some recent data from animal models suggest that local cytokine expression in the nose may affect olfaction (Turner et al. 2010); human studies testing this hypothesis are challenging due to the difficulty in accessing the olfactory cleft in adequate numbers of subjects, although clinically inflammation in the nose (e.g., chronic rhinosinusitis [CRS], allergic rhinitis [AR]) can affect the sense of smell. Systemic inflammation is associated with frailty, but its role in sensory impairment and olfaction specifically remains unclear.

Here, we focused on the relatively understudied potential connection between systemic immune function (as measured by peripheral blood cytokines) and olfaction. We considered 4 different, preselected profiles of cytokine expression as plausible mechanisms for olfactory dysfunction in older adults: the proinflammatory response, Th1 (Type 1), and Th2 (Type 2) components, and a physical frailty signature.

The proinflammatory response is elevated in chronic inflammatory states and autoimmune disorders, and several systemic diseases (Turner et al. 2014). Recently, induction of proinflammatory cytokine tumor necrosis factor-alpha (TNF-α) expression in the mouse olfactory cleft was shown to lead to progressive infiltration of inflammatory cells and impairment of olfactory function (Lane et al. 2010); such inflammation may also affect olfactory stem cell regeneration (Turner et al. 2010; Chen et al. 2017). Clinical studies have shown higher levels of interleukin-6 (IL-6) in saliva, plasma, and nasal mucus of patients with olfactory dysfunction compared with that of healthy controls (Henkin et al. 2013).

The Th1 (Type 1) immune responses is associated with cell-mediated responses to viral or bacterial infection (Spellberg and Edwards 2001). Olfactory dysfunction can result from such upper respiratory tract infections and therefore could be associated with heightened expression of Th1 cytokines. Indeed, induced secretion of the Th1 cytokine interferon gamma (IFN-γ) in mouse olfactory epithelium results in diminished odorant responsiveness (Pozharskaya et al. 2013). A correlation with this profile could indicate a role for bacterial and viral infections in olfactory dysfunction, either through secondary damage from inflammation or direct damage from pathogens.

The Th2 (Type 2) immune response is involved with allergic inflammation, which has also been associated with olfactory dysfunction. For example, one of the most common causes of olfactory loss is sinonasal disease (AR, CRS), likely related to local Th2 (Type 2) inflammation in the olfactory cleft (Sivam et al. 2010; Pozharskaya et al. 2013; Lavin et al. 2017). These inflammatory conditions of the nose commonly show increased expression of Th2 cytokines (Wheatley and Togias 2015; Shah et al. 2016). Additionally, levels of the Th2 cytokine IL-5 in nasal mucus correlate with decreased olfaction in patients with chronic CRS (Schlosser et al. 2016; Lavin et al. 2017). Thus, Th2 cytokines may be involved in olfactory dysfunction.

Lastly, sensory impairment is also linked with physical frailty (Pinto et al. 2017), which has been associated with systemic inflammation (Hubbard et al. 2009) and specific cytokine profiles, for example, the IL-1Ra^high-IL-4^low profile (Perez-Suarez et al. 2016). IL-1Ra is considered an anti-inflammatory cytokine through its action of competing with other IL-1 family member proteins for the IL-1 receptor (Arend and Gabay 2000). IL-4 is a cytokine important in Th2 differentiation of CD4+ T cells; the IL-4 gene is located on chromosome 5 with IL-13. These 2 cytokines share high sequence homology and many regulatory elements, with coordinated expression in CD4+ T cells (Bao and Reinhardt 2015). While these cytokines have overlapping Th2 functions, IL-4 is thought to drive Th2 cell development and humoral immunity while IL-13 is produced by peripherally acting Th2 cells. Additionally, IL-4 plays a significant immunomodulatory role, particularly against IFN-γ producing CD4+ T cells (Nelms et al. 1999). The IL-1Ra^high-IL-4^low profile reported to be associated with physical frailty may represent an “opponent-process” state of dysregulated signal from increased IL-1Ra and decreased IL-4 and IL-13. Furthermore, if olfactory function was associated with this profile, it would suggest that a common immune mechanism underlies both sensory and physical frailty.

In summary, there is strong clinical justification for a relationship between chronic inflammation and olfactory dysfunction. The relative importance of these pathways, however, remains unclear. Additionally, the site(s) of inflammation (peripheral vs. central) and mechanisms (indirect vs. direct damage to olfactory epithelium) remain areas of active investigation. Correlation with a particular immunologic profile would yield information on potential mechanisms that underlie olfactory dysfunction.

In this study, we measured plasma cytokines in a nationally representative cohort of older US adults and determined which profiles were correlated with olfactory dysfunction. Our main goal was to connect systemic immune function and olfaction.

Materials and methods

Subjects

The National Social Life, Health, and Aging Project (NSHAP) is a nationally representative, study of home-dwelling older US adults. Survey data and physiologic biomeasures collected from 3196 adults and their coresident spouses/partners born between 1920 and 1947 and interviewed at home in 2010–2011 as part of NSHAP (O’Muircheartaigh et al. 2014). An additional 181 spouses/partners born outside of 1920–1947 (which was outside our target age range of interest) were not considered in these analyses.

According to predetermined study design, approximately two-thirds of the sample was randomly selected to receive olfactory assessment (n = 2304) and 96% of respondents (n = 2212) consented to participate in the randomized olfactory module. Of these, 2094 were within our target age range (62–90), and 2084 received both olfaction tests. These data are publicly available (Waite et al. 2019). Additional details about this cohort are available elsewhere (Jaszczak et al. 2014; O’Doherty et al. 2014; Pinto et al. 2014). Written informed consent was obtained from all subjects and this study was approved by the Biological Sciences Division/University of Chicago Medical Center (BSD/UCMC) Institutional Review Board at the University of Chicago.

Biomeasure collection

Blood sampling

Blood samples were collected from respondents (92.1% weighted cooperation rate) using a finger-prick protocol (see O’Doherty et al. 2014 for additional details). Briefly, after using hand warmers to increase blood flow, 5 drops (250 µL) of uncoagulated whole blood were collected and stored in an ethylenediaminetetraacetic acid (EDTA) coated microtainer, and kept cool (<4 °C) prior to overnight shipping to the laboratory during the in home survey visit. Plasma isolated from these samples, was aliquoted, and stored at −80 °C prior to assay. Additionally, dried blood spots (DBSs) were collected for measurement of C-reactive protein (CRP) levels using standard protocols (O’Doherty et al. 2014).

Cytokine analytes

Plasma levels of 18 cytokines were measured: IFN-γ, IL-10, IL-1b, IL-6, Monocyte Chemoattractant Protein-1, Transforming Growth Factor alpha, TNF-α, Granulocyte–Macrophage Colony Stimulating Factor, IL-12, IL-13, Interleukin-1 receptor antagonist, IL-2, soluble Interleukin-2 receptor alpha, IL-3, IL-4, IL-5, TNF-β, and Vascular Endothelial Growth Factor. Luminex xMAP assays were performed at the University of Chicago Flow Cytometry Core Facility on standard equipment (Luminex 100; BioRad) and the BioPlex Manager Software (Version 5, BioRad) using protocols from the manufacturer. Mean values were used in analyses presented. This technology has been validated and shown to be reliable and reproducible compared with individual assays (e.g., enzyme-linked immunosorbent assay [ELISA]), especially for low volume samples as were collected here in this population-based study conducted in respondents’ homes (Codorean et al. 2010; Richens et al. 2010).

We examined 4 different cytokine profiles (Table 1). First, cytokine measurements were divided into quintiles after assigning a value of 0 to those determined to be out of range low, with the lowest 3 quintiles designated as “low expression,” and the highest 2 quintiles designated as “high expression,” using standard methods (Sun et al. 2016). The first profile tested was IL-1Ra^high-IL-4^low-IL-13^low. This profile is similar to one reported to be associated with physical frailty (Perez-Suarez et al. 2016) but also includes IL-13, due to its high sequence homology, overlapping function, and interlinked role in immune function with IL-4 (McCormick and Heller 2015). Respondents were required to fit the qualifications of each cytokine within the profile (i.e., a subject with the IL-1Ra^high-IL-4^low-IL-13^low profile must have demonstrated IL-1Ra high and IL-4 low and IL-13 low expression to be included).

Table 1.

Cytokine profile summary

Cytokine profile	Cytokines
Proinflammatory reflex	IL-1b + TNF-α + IL-6
Th1	IFN-γ + IL-2 + IL-12
Th2	IL-4 + IL-13 + IL-5 + IL-10
Frailty	IL-1Rahigh + IL-4low + IL-13low

To be included in proinflammatory, Th1, or Th2 profile, all cytokines had to be high (top 2 quintiles). In the frailty profile, proinflammatory cytokine IL-1Ra had to be high, and anti-inflammatory cytokines IL-4 and IL-13 had to be low (bottom 3 quintiles).

The other 3 profiles were defined by standard categories of immune function: Th1 (Type 1) [IFN-γ ^high, IL-12^high, IL-2^high]; Th2 (Type 2) [IL-4^high, IL-5^high, IL-10^high, IL-13^high]; and the proinflammatory reflex [IL-1b^high, IL-6^high, TNF-α ^high].

Olfactory function

Olfaction was measured in a randomly selected two-thirds of NSHAP respondents by design at baseline using the Olfactory Function Field Exam (OFFE) (Kern et al. 2014). The OFFE includes a 5-item validated assessment of odor identification via Sniffin’ Sticks. For each item, respondents chose between 4 word/picture options to match the odor they were presented with. The OFFE also includes a validated 6-item odor threshold test which assessed sensitivity to the common testing odor n-butanol, also via Sniffin’ Sticks (Kern et al. 2015). The number of respondents with both odor identification and threshold data was 2084; the demographic and health characteristics of this population are described in Table 2.

Table 2.

Olfactory dysfunction and cytokine measurements in older US adults

Characteristic (n = 2084)	Weighted %
n-Butanol sensitivity
Anosmic (4–6 errors)	28.6
Hyposmic (2–3 errors)	42.9
Normosmic (0–1 errors)	28.5
Odor identification
Anosmic (4–5 errors)	5.4
Hyposmic (2–3 errors)	16.9
Normosmic (0–1 errors)	77.7
IL-1Rahigh-IL-4low-IL-13low	15.4
Age (years, weighted mean ± SD)	72.4 ± 7.5
Gender
Men	46.8
Race/ethnicity (n = 2076)
White	81.4
Black	9.9
Hispanic, non-Black	6.3
Other	2.4
Education
<High school	16.0
High school graduate or equivalent	26.4
Some college	30.5
Bachelors or higher	27.1
Cognition (MoCA-SA, weighted mean ± SD)	14.0 ± 3.9
Smoking status
Never	41.0
Former	46.3
Current	12.7
Modified Charlson comorbidity index (weighted mean ± SD)	1.2 ± 1.5
Acute inflammation (CRP >8.6 mg/L) (n = 1904)	11.1

SD, standard deviation.

Data for both olfactory tasks were scored by summing the number of errors (0–5 for odor identification, 0–6 for n-butanol detection) as in prior work. Refusals to answer or responses of “don’t know” were treated as incorrect. Respondents were categorized into olfactory groups for olfactory sensitivity based on number of n-butanol pens correctly detected (anosmic 0–2, hyposmic 3–4, normosmic 5–6). Similarly, respondents were classified into odor identification groups based on number of odor pens correctly identified (anosmic 0–1, hyposmic 2–3, normosmic 4–5) (Table 2).

C-reactive protein

CRP (a general marker of inflammation) was measured from DBS collected from NSHAP respondents concurrently (O’Doherty et al. 2014). CRP levels were categorized as low (<0.76 mg/L), average (0.76–2.5 mg/L), high (2.5–8.6 mg/L), or acutely high (>8.6 mg/L) according to standard criteria (McDade et al. 2011). Of the 2084 respondents with both odor identification and threshold data, 222 had acutely high CRP levels and 180 did not have a CRP value at all (45 due to insufficient sample, 7 due to equipment problems or being lost in transit to the lab, and 128 due to not consenting to the DBS module). Those with acutely high CRP levels were excluded to reduce confounding related to acute infection or inflammation; those without a CRP value were also excluded.

Demographics and health parameters

Olfaction has been shown to vary with age and gender (Murphy et al. 2002; Schubert et al. 2008; Kern et al. 2014). Additionally, race/ethnicity, years of education (as a proxy for socioeconomic status), smoking, and comorbidity could potentially affect olfaction and/or systemic inflammation. Additionally, odor identification, which requires cognitive as well as sensory ability, can vary with cognitive function (Devanand et al. 2010; Pinto et al. 2014). Thus, age, gender, race/ethnicity, years of education, smoking status, comorbidity, and cognitive function were included in our analyses to account for these factors.

We divided our subjects into 3 age groups (62–69, 70–79, 80–90 years of age) as in prior work. Gender and race/ethnicity were recorded using standard National Institutes of Health (NIH) categories. Education level was determined via survey and classified into: did not complete high school, high school degree/equivalent, vocational certification/ some college, and bachelor’s degree or higher. We used a modified Charlson score to measure comorbidity (Vasilopoulos et al. 2014). Cognitive function was measured using the Montreal Cognitive Assessment-Survey Adapted (MoCA-SA) (previously known as the Chicago Cognitive Function Measure [CCFM]) (Shega et al. 2014; Kotwal et al. 2015) which was adapted for field use in NSHAP from the MoCA (Nasreddine et al. 2005). Smoking status was based on 2 questions: “Do you smoke cigarettes, cigars, or a pipe now?” and “Did you ever smoke cigarettes, cigars or a pipe regularly?” A measure of frailty was also constructed (a maximum of 4 points, excluding weight loss) (Huisingh-Scheetz et al. 2014).

Statistical analysis

We tested the association between cytokines/cytokine profiles and olfactory performance using multivariate ordinal logistic regression, adjusted for age group, gender, race/ethnicity, education level, comorbidity, current smoking status, and cognition. For each olfactory function measure, the number of errors was used as the dependent variable in regression models, and the olfactory function group categorization was used for descriptive purposes. For all analyses, those with evidence of acute inflammation or missing data based on CRP were excluded, leaving 1682 respondents with both odor identification and olfactory threshold data. Data were analyzed using Stata 16 (StataCorp.) applying the survey weights provided with the data to account for differential probabilities of selection and differential nonresponse.

Results

In this study, the average age was 72 and slightly less than half were men. Additional demographic and clinical parameters are provided in Table 2. The percentage of older US adults with the IL-1Ra^high-IL-4^low-IL-13^low profile across olfactory function categories was determined. For olfactory sensitivity, 12% of normosmics, 15% of hyposmics, and 23% of anosmics had the profile. For olfactory identification, 16% of normosmics, 15% of hyposmics, and 27% of anosmics had this profile.

Older adults with the cytokine profile IL-1Ra^high-IL-4^low-IL-13^low made more errors on n-butanol olfactory sensitivity testing (odds ratio [OR] = 1.61, 95% confidence interval [CI] 1.19–2.17, P = 0.002) after accounting for age group, gender, race/ethnicity, education level, cognition, smoking, and comorbidity (Table 3). Respondents showing this profile also made more odor identification errors (OR = 1.42, 95% CI 1.11–1.80, P = 0.006) after accounting for the same covariates (Table 3). In a sensitivity analysis, the frailty score was added to the multivariable models and the IL-1Ra^high-IL-4^low-IL-13^low profile effect remained strong for both olfactory sensitivity (OR = 1.59, 95% CI 1.21–2.10, P = 0.001) and olfactory identification (OR = 1.38, 95% CI 1.08–1.77, P = 0.01).

Table 3.

Associations between IL-1Ra^high-IL-4^low-IL-13^low and number of errors made on n-butanol and odor identification testing (from separate multivariable ordinal logistic regression models)

Variable	# of errors, n-butanol	# of errors, identification
	OR (95% CI)	OR (95% CI)
IL-1Rahigh-IL-4low-IL-13low (vs. no)
Yes	1.61 (1.19–2.17)	1.42 (1.11–1.80)
Unknowna	0.87 (0.70–1.09)	0.94 (0.68–1.30)
Age group (vs. 62–69)
70–79	1.09 (0.82–1.44)	1.36 (1.04–1.79)
80–90	1.55 (1.09–2.18)	2.89 (2.07–4.04)
Gender (vs. men)
Women	0.87 (0.68–1.11)	0.70 (0.58–0.85)
Ethnic group (vs. White)
Black	1.15 (0.76–1.76)	1.26 (0.89–1.78)
Hispanic	1.19 (0.57–2.49)	1.07 (0.69–1.66)
Other	0.42 (0.20–0.87)	0.83 (0.44–1.57)
Education (vs. <HS)
HS or equivalent	0.96 (0.66–1.40)	1.10 (0.78–1.57)
Some college	1.06 (0.66–1.69)	1.04 (0.70–1.55)
Bachelors or higher	0.94 (0.64–1.39)	1.30 (0.84–2.03)
Cognition (MoCA-SA; per 1 point increase)	0.95 (0.91–0.99)	0.89 (0.85–0.92)
Smoking status (vs. former/never)
Current	0.74 (0.53–1.07)	1.09 (0.72–1.64)
Comorbidity (Charlson index; per 1 point increase)	1.03 (0.95–1.10)	1.04 (0.97–1.12)

HS, high school.

^aRespondent refusal, inability to provide a sample, or insufficient sample (14.7% of analytic sample, see Supplementary Table S1).

Respondents with low IL-4 expression did not make more errors on n-butanol olfactory sensitivity (P = 0.73) or odor identification testing (P = 0.37); similar results were found for those with low IL-13 expression (P = 0.59, P = 0.84 sensitivity and identification errors, respectively) (Table 4). Although respondents with high IL-1Ra expression did not make significantly more errors on n-butanol odor sensitivity testing (P = 0.14), they had more errors on odor identification testing (OR = 1.37, 95% CI 1.09–1.71, P = 0.007). The increased strength of the association between the IL-1Ra^high-IL-4^low-IL-13^low profile and both odor identification and sensitivity supports analyzing cytokines by profile.

Table 4.

Associations between number of errors made on n-butanol and odor identification testing and individual cytokines and other cytokine profiles (from separate multivariable ordinal logistic regression models)

	# of errors, n-butanol	# of errors, identification
	OR (95% CI)a	OR (95% CI)a
IL-4low	1.06 (0.78–1.43)	0.89 (0.68–1.16)
IL-13low	1.08 (0.80–1.46)	1.02 (0.83–1.26)
IL-1Rahigh	1.18 (0.95–1.47)	1.37 (1.09–1.71)
IL-1Rahigh-IL-4low	1.50 (1.10–2.04)	1.37 (1.07–1.76)
IL-1Rahigh-IL-4low-IL-13low	1.61 (1.19–2.17)	1.42 (1.11–1.80)
Proinflammatory profile	1.24 (0.74–2.09)	1.07 (0.67–1.73)
Th2 profile	1.09 (0.82–1.45)	0.86 (0.62–1.20)
Th1 profile	1.02 (0.72–1.45)	0.86 (0.62–1.20)

^aAdjusted for age, gender, race/ethnicity, education, cognition, current smoking status, and comorbidities.

Older adults with the proinflammatory profile did not make more errors on n-butanol olfactory sensitivity testing (P = 0.41) or odor identification (P = 0.77) (Table 4). There were also no significant associations with Th1 or Th2 profiles with either sensitivity or identification (Th1: n-butanol number of errors, P = 0.90; odor identification errors, P = 0.38; Th2: n-butanol number of errors, P = 0.55; odor identification errors, P = 0.37).

Finally, we tested whether elevated CRP levels versus average or low levels, as a general measure of chronic inflammation, correlated with olfactory function; they did not (n-butanol olfactory sensitivity, P = 0.40; odor identification, P = 0.31; data not shown).

Older respondents demonstrated worse n-butanol sensitivity and odor identification (Table 3). Women made a similar number of errors compared with men on odor sensitivity testing (OR = 0.87, 95% CI 0.68–1.11), but did have fewer odor identification errors (OR = 0.70, 95% CI 0.58–0.85) (Table 3), consistent with prior work (Doty et al. 1984). There were no racial/ethnic differences after adjustment for the other covariates, except for the category of “other” (n = 48), which was associated with fewer errors on odor sensitivity compared with whites (OR = 0.42, 95% CI 0.20–0.87) (Table 3). We did not find effects of education level in our models. Comorbidity scores did not significantly affect either olfactory modality (Table 3). Respondents with better cognitive function made fewer odor sensitivity errors (OR = 0.95, 95% CI 0.91–0.99) and odor identification errors (OR = 0.89, 95% CI 0.85–0.92) (Table 3).

Discussion

We identified IL-1Ra^high-IL-4^low-IL-13^low as a novel peripheral cytokine profile that correlates with olfactory dysfunction. This work supports the role of immune system in chemosensory impairment in aging and highlights the role of IL-1Ra, IL-4, and IL-13 in these physiologic processes. To the best of our knowledge, this is the first study connecting olfaction and systemic inflammation (as assessed by plasma cytokines) in a nationally representative population of older adults. Our results may indicate that chronic inflammation affects sensory function, consistent with the relationship between immune responses (measured by peripheral cytokines) and aging across numerous physiological systems (Fougere et al. 2017).

Divining specific mechanisms underlying these findings is challenging and remains outside the scope of this epidemiologic study. Thus, we may only speculate on how these cytokines may affect olfaction based on functions of individual cytokines within the group. First, IL-1β is thought to be an initiator of the inflammatory response; IL-1Ra is a natural inhibitor of the proinflammatory effect of IL-1β, and is more accurately measured (Biasucci et al. 1999). Thus, elevated IL-1Ra levels signify a response to increased inflammation in the setting of olfactory dysfunction. IL-4 and IL-13 have been implicated in neurodegeneration, a process that may begin first in the olfactory system (Mori et al. 2016). In some settings, they have been shown to generate protective neuroimmune responses (Mori et al. 2016). Thus, the IL-1Ra^high-IL-4^low-IL-13^low profile may signify elevated proinflammatory processes combined with a deficiency of mitigating anti-inflammatory signal, leading to damaged olfactory epithelium and decline in olfaction. This reasoning is consistent with recent mouse models showing that cytokine-mediated inflammation can directly damage olfactory epithelium (Lane et al. 2010; Chen et al. 2017). Given that IL-1Ra potentially serves as a proxy measure of proinflammatory IL-1 activity, our findings may represent imbalance of pro- and anti-inflammatory signals. This imbalance reflects dysregulated anti-inflammatory signaling and results in olfactory impairment. Clearly, this is a complex area; further mechanistic studies are needed to investigate the precise immune mechanisms underlying our findings.

Other work has identified associations between sensory dysfunction and physical frailty (Linard et al. 2016; Pinto et al. 2017; Arnadottir et al. 2020). We approached the study here with the hypothesis that chronic inflammation would be a common process that would connect these 2 phenomena mechanistically. However, at least regarding the specific cytokine profiles tested here, this was not the case. For example, olfactory dysfunction did not track with general markers of systemic inflammation (CRP, proinflammatory cytokine group) (as has physical frailty; Velissaris et al. 2017) and the associated IL-1Ra^high-IL-4^low-IL-13^low profile was not confounded by physical frailty (Perez-Suarez et al. 2016). We speculate that other specific types of chronic systemic inflammation related to different forms of aging may underlie physical and sensory frailty, a question that remains open for subsequent study.

Prior studies on cytokines and olfactory dysfunction have found evidence for involvement of IL-5, IL-6, TNF-α, and other cytokines in olfactory function (Sultan et al. 2011; Henkin et al. 2013; Schlosser et al. 2016). Here, we did not identify an association between IL-6 or TNF-α and olfactory dysfunction (data not shown). One potential explanation for this difference is that the general inflammatory signal captured by the IL-6 and TNF-α is often comorbid with processes that affect olfaction (and which our models account for), but olfaction is not causally linked to the chronic inflammatory or autoimmune processes associated with these cytokines. Alternatively, TNF-α and other proinflammatory cytokines may be necessary for stimulating regeneration of adult neural stem cells in the olfactory epithelium after injury (Chen et al. 2017). We found no association between olfactory function and Th1 cytokines, suggesting that immune responses that are involved in fighting microbial infections do not result in olfactory dysfunction. These issues are complex and require additional study.

There were several limitations in our study. Our study was cross-sectional and therefore we cannot make conclusions on causality, though we note that reverse causality is not likely here. Other limitations stem from the inability to determine mechanism, which is difficult in human immunological studies, especially when not performed in a laboratory or clinical setting. We acknowledge that logistical issues surrounding measurement of cytokines using multiplex versus single assays in population studies is challenging, although the methods we employed are standard (Codorean et al. 2010; Richens et al. 2010). We also note that this study was done in a large representative population (see Supplementary Table S1, showing the comparability of the analytic sample to the rest) and therefore provide normative data on the general population of older US adults.

Our results reflect systemic immune function, which is directly related to the study design regarding the feasibility of collecting peripheral blood cytokines. We were unable to study local immunity in the nasal cavity, which is an entirely different scientific question. Although understanding local immune effects in the olfactory cleft may provide additional information and be desirable, such studies are not feasible in home-based studies. Similarly, assessing immune responses in the central nervous system is not feasible in humans outside of autopsy studies or neuroimaging. However, peripheral inflammation can produce immune responses in the central nervous system mediated by microglial cells and astrocytes (Doursout et al. 2013) in animal models and a contributory role of systemic inflammation in neurodegeneration and cognitive aging has been observed in longitudinal studies in humans (Walker et al. 2017). For example, Bettcher found differences in the cognition/memory measures and brain volumes by imaging in those with elevated CRP versus undetectable levels (Bettcher et al. 2012). Bettcher also reported that the relationship between peripheral blood-obtained IL-6 levels and white matter integrity is strongest in older ages (Bettcher and Kramer 2014; Bettcher et al. 2014). These studies and others indicate that inflammation, assessed in the systemic circulation, may be correlated with changes in the central nervous system. Thus, systemic immune function may affect olfaction via effects on central olfactory pathways in addition to any local effects of the immune system in the peripheral olfactory system. Interestingly, Herbert et al. have shown that Staphylococcus aureus creates a localized IL-6-mediated immune response and can penetrate olfactory epithelium and enter the olfactory bulb (Herbert et al. 2012). Through this mechanism, local inflammation in the nasal passages could potentially affect olfactory sensory function via central effects. Although the precise mechanisms remain unknown, there is evidence for systemic inflammation to affect olfactory function, both by peripheral and central mechanisms.

This is the first study to use a nationally representative sample of older adults to analyze the relationship between plasma cytokines and olfactory dysfunction. Future work will focus on validating this in other cohorts and testing whether this profile is specific to olfaction or is related to other sensory modalities (e.g., decreased vision, hearing).

In conclusion, we have identified a peripheral cytokine profile that is associated with olfactory dysfunction in older adults. Further studies are needed to investigate the precise molecular implications involved.

Funding

The National Social Life, Health and Aging Project is supported by the National Institute on Aging (NIA) (R37AG030481, R01AG033903, R01AG043538, R01AG048511). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. ED was supported by the Summer Research Program (NIA T35AG029795) and the Scholarship and Discovery Program at the Pritzker School of Medicine at The University of Chicago.

Conflicts of interest

The authors declare no conflicts of interest.