The Immune Phenotype of Perinatal Anxiety

Abstract

Background:

Immune dysregulation has been linked to both psychiatric illness and pregnancy morbidity, including perinatal depression, but little is known about the immune phenotype of perinatal anxiety. Here, we sought to identify the unique immune profile of antenatal anxiety.

Materials and Methods:

Pregnant women (n=107) were followed prospectively at 2nd and 3rd trimesters (T2, T3) and 6 weeks postpartum (PP6). Each visit included a blood draw and psychological evaluation, with clinical anxiety assessed using the Spielberg State-Trait Anxiety Scale. We enrolled both healthy controls and participants with anxiety alone; those with comorbid depression were excluded. Multiplex cytokine assays and flow cytometry were used to examine the association of anxiety symptoms with secreted immune markers and PBMC-derived immune cells.

Results:

K cluster means revealed three clusters of anxiety symptomatology; due to low numbers in the highest severity anxiety group, these were collapsed into two groups: Non-Anxiety and Anxiety. Principal components analysis revealed two distinct clusters of cytokine secretion including one cluster that consisted of many innate immune cytokines and differed between groups. Compared to women in the Non-Anxiety group, women in the Anxiety group had lower levels of cytokine expression during pregnancy and an increase in levels into the postpartum, whereas Non-Anxiety women experienced a time-dependent decline. Immune cell populations also differed between our two groups, with the Anxiety group showing a decrease in the ratio of B cells to T cells from pregnancy to postpartum, whereas the Non-Anxiety women showed an increase in this ratio over time. Women in the Anxiety group also demonstrated an increased ratio of cytotoxic to helper T cells throughout pregnancy, a modest increase in the Th1:Th2 ratio across pregnancy, and a lower ratio of Th17:T_REG cells in the postpartum as compared with Non-Anxiety women.

Conclusion:

These data suggest that the immune response throughout the antenatal period differs for women with anxiety symptoms compared to those without, suggestive of a unique immune phenotype of perinatal anxiety.

1. Introduction

Anxiety during pregnancy and postpartum is common and impairing. Systematic reviews and meta-analyses have reported prevalence rates from 4% to 39% for the presence of any anxiety disorder during pregnancy (1–3). Self-reported anxiety symptoms that do not meet criteria for a disorder are common as well, with a prevalence in one recent review of 24% in pregnancy and 15% postpartum, (1) with additional evidence of increased rates during the COVID-19 pandemic (4). The burden of these symptoms on women is clear, with one study showing that while only 1.5% of women in a perinatal day hospital met criteria for generalized anxiety disorder (GAD), almost 63% reported believing they had the disorder (5). Even among psychiatric patients, pregnant women with anxiety compared to those without have a 50% higher rate of suicidality, a high rate of obsessions (30%), take more psychiatric medications, and have higher rates of postpartum depression (5–9). Antenatal anxiety symptoms are known risk factors for the health of the pregnancy and the child, with increased risks of pregnancy morbidities (10–12), and risks to the child such as altered immunity, altered HPA axis reactivity, anxiety, and child behavioral outcomes including: increased attention deficit/hyperactivity disorder, conduct problems, and cognitive delays (6, 10, 13–15).

Despite this significant burden of disease on both women and children, our understanding of the biological correlates of perinatal anxiety is poor. Many studies have looked at alterations in HPA axis activity, neuroendocrine changes, and inflammatory activity in women with perinatal depression or mixed anxiety and depression (16–19), and the same biological factors are known to be involved in the neurobiology of anxiety outside of the perinatal period (20–25) – but few have looked at these factors in pregnant and postpartum women with anxiety alone.

One system ripe for additional research in perinatal anxiety is the immune system. There are radical shifts in immune functioning across pregnancy, as the body shifts to allow implantation, keep foreign pathogens out while protecting the one (the fetus) already inside, and allow parturition (19, 26–29). Longitudinal work on immune system changes in perinatal anxiety is lacking. Maes and colleagues found evidence of increased inflammatory activity 1–3 days after delivery (i.e., increased serum IL-6 and IL-1RA and lower LIFR) in women whose anxiety increased in the puerperium (30). In a separate study, the same group found an association between low prenatal zinc and high C-reactive protein (CRP) concentrations and the severity of prenatal anxiety symptoms (31). Leff-Gelman and colleagues found increases in third-trimester pro-inflammatory cytokines and chemokines (i.e., IL-6, TNF-α, IL- 2, IFN-γ, IL-4, IL-10, IL-8/CXCL8, MCP-1/CCL2 and MIP-1β/CCL4) in women with severe anxiety compared to healthy controls (16, 17). Furtado and colleagues found no significant associations between third trimester levels of IL-6, TNF-α, and CRP and worsening anxiety in the postpartum, but the small sample sizes for some markers due to limits of detection make conclusions difficult to draw (32). Our own previous study (the only one of which we are aware with longitudinal measures) showed that women without known psychiatric illness across 5 time points in pregnancy and postpartum had a burst of innate immune activity (i.e., concentrations of IL-6, GM-CSF, CCL-3, and CXCL-8) associated with increased anxiety symptoms in the third trimester and into the postpartum (33).

In the present study, we expanded our previous work by including women with and without clinically significant anxiety without comorbid depression, and measured peripheral cytokines and immune cell populations longitudinally. We hypothesized that immune activation would be positively associated with increased anxiety symptoms in late pregnancy and continuing into the postpartum.

2. Material & Methods

2.1. Participants and inclusion criteria

This study was approved by the institutional review board at the Johns Hopkins University School of Medicine. Pregnant women with and without clinically significant anxiety (N=107) were recruited within the Johns Hopkins Health System via Facebook ads, a listing on our website, physician letters, and flyers. They were enrolled in the first or second trimester of pregnancy and data were collected at four study visits. The present study includes data from the second and third trimesters of pregnancy (T2 and T3) and 6 weeks postpartum (PP6). Data were also collected at six months postpartum but were not included in the present study due to smaller numbers of participants with biological data at this time point. Psychological scales and a blood draw were completed at each visit. Women were recruited to the study between August 2016 and February 2020; recruitment was shut down prematurely due to our hospital’s temporary suspension of clinical research at the outset of the COVID-19 pandemic. At baseline, psychiatric diagnosis was determined by the Structured Clinical Interview for DSM-V Diagnoses (SCID), conducted by a trained research coordinator, and confirmed by psychiatrist interview. Any discrepancies between SCID and clinician interview were reconciled by discussion among research staff, study clinician, and the principal investigator immediately after the assessment visit. Data were collected using the REDCap (Research Electronic Data Capture) tool hosted at Johns Hopkins University. REDCap is a secure, web-based software platform designed to support data capture for research studies, providing 1) an intuitive interface for validated data capture; 2) audit trails for tracking data manipulation and export procedures; 3) automated export procedures for seamless data downloads to common statistical packages; and 4) procedures for data integration and interoperability with external sources (34, 35).

Inclusion criteria for all subjects were as follows: pregnant and <27 weeks’ gestation, age 18 or above, able to provide written consent, healthy pregnancy. Exclusion criteria for all subjects included: multifetal pregnancy, autoimmune or endocrine disease, meeting criteria for a major depressive episode at study entry, active suicidal ideation at study entry, bipolar disorder or primary psychotic disorder, or current or recent (within 90 days) substance use disorder. In order to ensure that our sample included both women with and without clinically significant anxiety, half of our sample had additional inclusion criteria as follows: significant anxiety symptoms as measured by a score ≥ 21 on the Perinatal Anxiety Screening Scale (PASS) (36), a validated screening tool for anxiety in pregnancy and the postpartum, and/or a diagnosis of a current, active anxiety disorder by SCID or clinician using DSM-V criteria. (Women with current psychotropic medication use and those with a prior history of major depressive episodes were not excluded, as such criteria would have limited both the severity of symptoms we could include and our ability to recruit.) The other half had additional exclusion criteria as follows: no history of a current or past psychiatric disorder as determined by Structured Clinical Interview for DSM-V Diagnoses (SCID) and no current use of an antidepressant. Women who met at least one exclusion criterion were removed during the screening process and did not move forward in the study.

Participants completed the following scales at each study visit: the Perinatal Anxiety Screening Scale (PASS) (36), the Edinburgh Postnatal Depression Scale (EPDS) (37), the Penn State Worry Questionnaire (PSWQ) (38), the Spielberg State-Trait Anxiety Inventory (STAI) (39), The Pittsburgh Sleep Quality Index (PSQI) (40), the Perceived Stress Scale (PSS) (41), the Obsessive Compulsive Inventory-Revised (OCI-R) (42), and the Life Events Checklist (LEC) (43). Scales were completed online using a secure REDCap link. Participants also completed the Major Depressive Episode module from SCID and were evaluated by a member of the research team to assess for current depressive episodes. At baseline, participants also completed a Childhood Traumatic Events Scale (44), which was scored into total number of events and sum of severity scores. Because anxiety symptoms shifted across time, because our psychological scales were closely correlated to one another, and because the majority of our participants with anxiety had impairing symptoms without meeting criteria for a disorder on the SCID, we elected to use a K-means cluster analysis to define two groups of participants, Anxiety and Non-Anxiety (see below in statistical methods for details of this approach).

2.2. Blood draw methods and PBMC separation

Blood was collected in sodium-heparin tubes (30mL) in the morning and transported to the laboratory at room temperature. Tubes collected for plasma were spun down within 1 hour of blood collection by low-density gradient centrifugation via an Eppendorf centrifuge #5702 R and transferred to 1.8 mL aliquots, which were frozen at −80°C until analysis. Separate tubes were collected for peripheral blood mononuclear cell (PBMC) suspensions, which were isolated within 8 hours of blood collection by low-density gradient centrifugation via a Thermo Scientific^™ Sorvall^™ Legend^™ XTR Centrifuge, to avoid erythrophagy-related activation of the monocytes. PBMCs were frozen and stored at −80° C using 10% dimethylsulfoxide until processed.

Plasma Cytokine and Chemokine Measurement

Plasma levels of IFN-γ, IL-10, IL-12p70, IL-13, IL-1β, IL-2, IL-4, IL-6, IL-8, TNF-α, IL-6, TNF-α, GM-CSF, IL-12/IL-23p40, IL-15, IL-16, IL-17A, IL-1α, IL-5, IL-7, TNF-β, VEGF-A, MCP-1, MCP-4, Eotaxin, IP-10, MDC, Eotaxin-3, TARC, MIP-1α, and MIP-1β were assayed in duplicate using multiplex kits (V-PLEX Proinflammatory Panel 1 Human Kit, V-PLEX Cytokine Panel 1 Human Kit, and V-PLEX Chemokine Panel 1 Human Kit respectively) from Meso Scale Discovery (MSD, Meso Scale Discovery, 1601 Research Blvd, Rockville, MD). Plates were read by an MSD SECTOR Imager 2400 measuring electrochemiluminescence. Sample concentrations were extrapolated from a standard curve calculated using a four-parameter logistic fit using MSD Workbench 3.0 software.

2.3. Flow cytometric analysis

PBMCs were defrosted with 1mL FBS and washed twice with medium. The total numbers of live cells were determined using a hemocytometer and trypan blue (Invitrogen) and were resuspended at 1 × 10⁶ cells/ml in medium. Two different staining procedures were used: Stain A was used to determine percentages of immune cell phenotypes, and Stain B was used to specifically characterize T helper subsets.

Stain A: To determine the percentage of CD3+ T cells, CD19+ B cells, CD3+CD8+ T cytotoxic lymphocytes, and CD3+CD4+ T helper lymphocytes, 1×10⁶ PBMCs were stained for 30 min at room temperature containing staining buffer (BD Biosciences, San Jose, CA, USA) and CD3-BV-786 (BD Biosciences) 1:24, CD4-PE-Cy7 (BD Biosciences) 1:24 CD8-BV605 (BD Biosciences) 1:24, and CD19-BV421 (BD Biosciences) 1:24. Percentages of cells were calculated as frequencies of PBMCs from parent population.

Stain B: To determine T helper cell subsets (T helper memory cells and Th1, Th2, Th17, T regulatory cells), 1 × 10⁶ PBMCs were cultured for 4 hours at 37 °C stimulated in the RPMI-1640 culture medium (as was used for freezing the cells down, see before) with 50 ng/ml phorbol 12-myristrate 13-acetate (PMA; Sigma Aldrich, St. Louis, MO, USA) and 1.0 μg/ml ionomycin (Sigma) in the presence of Golgistop (BD Biosciences). Cells were harvested and stained for membrane markers with CD45-RO-BV605 (BD Biosciences) 1:24, CD3-BV786 (BD Biosciences) 1:24, CD4-PE-Cy7 (BD Biosciences) 1:24, and CD25-PE (BD Biosciences) 1:6 antibodies for 20 min at room temperature. After washing with PBS, cells were fixed and permeabilized according to the manufacturer’s instructions (eBioscience, San Diego, CA, USA) for 45 min at 4 °C. Cells were washed twice and stained with intracellular IFN-y-APC (BD Biosciences) 1:12, IL-4-FITC (BD Biosciences) 1:12, IL-17A-BV421 (BD Biosciences) 1:24 and Fox-P3-PerCP-Cy5.5 (BD bioscience) 1:24 in permeabilization buffer (eBioscience) for 45 min at 4 °C. Stained samples were analyzed by 14-color flow cytometry using a BD LSRII machine (BD biosciences) and analyzed by FlowJo software (Tree Star, Ashland, OR, USA). Gating strategy for Stains A&B is given in Supplemental Fig. 1 and evidence of cellular stimulation is provided in Supplemental Fig 2.

3. Statistical methods

Data were measured at three time points, in the second and third trimesters of pregnancy and again at 6 weeks postpartum. To investigate the association of different patterns of anxiety symptoms in women across pregnancy and the early postpartum with their concurrent immune environment we utilized the R-package KmL3D to employ k-means clustering of joint-trajectories longitudinally, a commonly used machine learning algorithm for identifying distinct subgroups of women with similar trajectories of anxiety across time (i.e., high intra-class similarity) (45–47). Anxiety symptom clusters were informed by women’s State-Trait Anxiety Inventory Trait scores obtained at the three measurement occasions. We elected to use only one scale score as all were closely correlated and we wanted to avoid multiple comparisons. We chose Trait because of prior literature indicating that prevalence of anxiety is highest when measured as trait, and that trait anxiety is associated with a host of negative outcomes, including preterm birth, poor maternal bonding, poor infant cognition, and increased difficult infant temperament (as reviewed in (1)). Moreover, we reasoned that trait, as a durable characteristic, might be most closely connected to long-term effects in both mother and child; indeed, trait anxiety is known to have different neuroanatomical correlates so may be the most distinct of the measures we could have chosen (48). This approach supported inclusion of women with missing data at one timepoint; those with missing data at two or more visits were not retained for cluster classification.

Demographic and clinical characteristics were described for the total sample (n=107) and compared between the cluster of Non-Anxiety women (n=51) and the cluster of women with moderate to high symptoms of anxiety (n=56). Student’s t-tests and cross-tabulations using Chi-square or Fisher’s Exact test, where appropriate, were used to examine group differences. Associations of demographic and clinical characteristics with immune outcomes were also examined via t-test or Pearson correlations. Trauma and sleep were examined as potential covariates but ultimately were not retained in the final models.

Principal components analysis was utilized as a data reduction technique for peripheral cytokine outcome data measured in maternal plasma at each timepoint. Immune cell outcomes similarly indexed in maternal PBMCs were retained as raw frequencies of PBMCs. In the evaluation of outcome variable distributions, Panel B: flow cytometry data from one woman were excluded for biological implausibility (i.e., values exceeded 3 standard deviations at all time points). Specific to the third trimester, Panel B: flow cytometry data were omitted at single time points for two additional women, one with an active allergic state in the third trimester and one with an infection at six weeks postpartum; both had values exceeding 3 standard deviations at those time points. Mixed effect models were used to estimate the association of anxiety cluster and women’s cytokine and immune cell activity over time, accounting for within-person correlations (i.e., dependency in immune measurement associated with repeated measures data of the same individual over time) with adjustment of relevant covariates. Apart from the k-means cluster analysis conducted using R software (49), analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC). Because this was an exploratory study designed to reveal potential relationships worthy of future study in larger samples, we did not perform corrections for multiple comparisons.

4. Results

Sample demographic and clinical characteristics are presented in Table 1. All participants in this study were pregnant women between the ages of 23 and 45. Women in the Anxiety group were older, had higher levels of trauma, took more psychiatric medications during pregnancy, and reported poorer sleep quality as compared to the Non-Anxiety women. A small number of women (6 in all) developed depressive symptoms by the 6-week postpartum visit; this number was too small to analyze separately and these participants were considered within the Anxiety group.

Table.1.

	Participant Demographics
	All (n= 107)	Anxious group (n= 56)	Healthy group (n= 51)	Group differences
Age	32.7	33.5	31.8	p = 0.027
Race - n (%)				p = 0.633
White	94 (87.85)	50 (89.29)	44 (86.27)
Non-White	13 (12.15)	6 (10.71)	7 (13.73)
Pre-pregnancy BMI - n (%)				p = 0.969
Normal Weight	51 (47.66)	27 (48.21)	24 (47.06)
Overweight and Obese	56 (52.34)	29 (51.79)	27 (52.94)
Education- n (%)				p = 0.675
Associates degree/Some college or less	9 (8.49)	6 (10.71)	3 (6)
Bachelors degree	38 (35.85)	20 (35.71)	18 (36)
Graduate or Professional Degree	59 (55.66)	30 (53.57)	29 (58)
Relationship Status - n (%)				p = 0.338
Married/Relationship	106 (99.07)	51 (100)	55 (98.21)
Single	1 (0.93)	0 (0)	1 (1.79)
Psychiatric medication use during pregnancy	13 (12.15)	12 (21.43)	1 (1.96)	p = 0.021
Average STAI Trait Score
T2	35.5	43	27.3	p < 0.001
T3	33.1	40	26	p < 0.001
W6	32.1	38.6	25.25	p < 0.001
Average pregnancy PSQI	46.1	56.4	34.7	p < 0.001
Average Trauma Severity Score	9.4	10.9	7.9	p = 0.039

4.1. Anxiety Group Clusters

K-means cluster analysis of joint-trajectories was used to identify distinct anxiety subgroups of women based on longitudinal STAI Trait scores at two time points during pregnancy and one time point in the early postpartum. (See methods for a discussion of why we used Trait scores.). As shown in Figure 1, cluster analysis revealed three groups: Group 1: Non-Anxiety (i.e., low anxiety) (n=56, 52.3%), Group 2: moderate and fluctuating anxiety (n=39, 36.4%), and Group 3: high and stable anxiety (n=12, 11.2%)). Due to small sample size of women with high anxiety (e.g., STAI Trait, 2nd trimester M = 50.16, SD = 9.16), this cluster was collapsed with the adjacent cluster of women with moderate anxiety symptoms, creating our two groups for subsequent comparison, hereafter referred to as Non-Anxiety and the Anxiety group (n=56) and (n=51).

Figure 1:

Group Cluster Analysis

Group cluster analysis revealed three groups based on STAI Trait anxiety score: Group 1: Non-Anxiety: Red (i.e., low anxiety) (n=56), Group 2: moderate and fluctuating anxiety: Green (n=39), and Group 3: high and stable anxiety: Blue (n=12).

4.2. Plasma Cytokines and Chemokines

Principal components analysis was used to examine the associations and clustering of the 32 plasma-derived cytokines and chemokines at the second trimester of pregnancy. Two functional clusters were extracted: Cluster A containing CXCL8, IFN-y, TNF-a, IL-10, and IL-6, which explained 39% of the variance with factor loadings ranging from .382 to .470; followed by Cluster B containing IL-17, IL-12p40, CCL3, CCL11, CCL13, CCL17, which explained 25% of the variance with factor loadings from .284 to .448, together these two components explained 64% of the cumulative variance (Figure 2A). Cytokines that did not load onto these components were excluded to reduce the number of estimated parameters. A factor-based composite Z-score was computed for each cluster to compare groups in longitudinal mixed effect models across pregnancy and postpartum. As depicted in Figure 2B, women within the Anxiety group had significantly reduced Cluster A activity in trimester 2 (β = 0.371, SE = 0.173, t = 2.14, p=0.035), and at a trend level in trimester 3 (β = 0.287, SE = 0.148, t = 1.94, p = .054) as compared with Non-Anxiety women. Cluster A activity in women in the Anxiety group increased from trimester 3 to postpartum week 6, a pattern that differed at a trend level from Non-Anxiety controls, who showed declining activity over the same period, (β = 0.381, SE = 0.226, t = 1.69, p = .096)) (Figure 2B). Cluster B activity did not differ between groups (Figure 2C).

Figure 2:

Principal Component Analysis and Multiplex Results

(A) Two functional clusters were extracted in the second trimester of pregnancy: Cluster A (bottom right quadrant) containing: IFN-y, IL6, IL8, IL10, and TNFα that explained 39% of the variance and Cluster B (top-right quadrant) containing: Eotaxin, IL12p40, IL17, MCP4, MIP1α, and TARC that explained 25% of the variance.

(B) Cluster activity between groups: women within the Anxiety group (Red) had significantly reduced Cluster A activity at trimester 2 but demonstrated a trending increase in innate immune activity from trimester 3 to week 6, compared to Non-Anxiety controls (Blue) who had reduced innate immune activity at this time. (C) Cluster B activity did not differ between groups. *p<0.05

4.3. Flow Cytometry

Figure 3 depicts proportions of broad categories of immune cells within the total immune cell population. Non-Anxiety women had a significantly higher proportion of B-cells at PP6 compared to Anxiety group women (Figure 3A: β = 4.289, SE = 1.955, t = 2.19, p=0.031). There was a marginally steeper increase in B-cell population from T3 to PP6 for the Non-Anxiety women as compared to the Anxiety women, but this slope difference did not reach statistical significance (Figure 3A: β = −3.505, SE = 2.068, t = −1.69, p=0.098). The overall T-cell population was largely similar between groups during pregnancy, though, at a trend level, women in the Anxiety group had a higher proportion of T-cells at PP6 relative to Non-Anxiety women (Figure 3B: β = −8.146, SE = 4.799, t = −1.70, p =0.093). The ratio between the B- and T-cell populations revealed a significant slope difference between T3 and PP6. The Non-Anxiety group’s B:T ratio increased over this period, whereas this ratio decreased among women in the Anxiety group (Figure 3C: β = −0.146, SE = 0.057, t = −2.56, p =0.014 for the slope and β = 0.162, SE = 0.080, t = 2.03, p=0.046 for the difference at PP6).

Figure 3:

Immune Cell Populations between Groups

Average percent of cells from parent gate (y-axis) across pregnancy and into the postpartum (x-axis). Women in the Anxiety group (red line) had fewer B-cells and more T-cells at week 6 postpartum compared to Non-Anxiety group (blue line). The Anxiety group and Non-Anxiety group did not differ in helper T-cell population (G), however cytotoxic T-cells were higher in the anxiety group at trimester 3 and week 6 postpartum (D). There was a steeper rise in Th1:Th2 ratio from T2 to T3 within the Anxiety group compared to Non-Anxiety (I). Within the Anxiety group, Th17 cells significantly reduced postpartum compared to the Non-Anxiety group (J). *p<0.05; #<0.10

Also shown in Figure 3, examination of differences within the T-cell compartment showed that the cytotoxic T-cell population (CD8+) was higher in women in the Anxiety group at T3 and PP6 compared Non-Anxiety (Figure 3D: T3: β = −7.058, SE = 3.231, t = −2.18, p =0.032; PP6: β = −6.617, SE = 3.359, t = −1.97, p =0.052). There were no differences in helper T-cells (CD4+) between Non-Anxiety and Anxiety women. The cytotoxic: helper T-cell ratio, however, was greater among Anxiety women than Non-Anxiety women at T3 (Figure 3F: β = −0.140, SE = 0.070, t = −2.00, p =0.049). Within the helper T-cell population, we detected no significant differences between groups for either Th1 or Th2 populations. The Th1:T_h2 ratio, however, increased between T2 and T3 and decreased from T3 to PP6 for the Anxiety Group, a pattern not seen among Non-Anxiety women. This pattern reached trend level significance for the slope differences from T2 to T3 (Figure 3I: β = 11.076, SE = 6.363, t = 1.74, p =0.087; T3 to PP6, β = −8.154, SE = 6.138, t = −1.33, p= 0.188). Regulatory T-cells (T_REG) did not differ between groups. The Th17 population was significantly higher at PP6 among Non-Anxiety women compared to the Anxiety group (Figure 3L: β = 1.656, SE = 0.711, t = 2.33, p =0.022), and the Th17:T_REG ratio was significantly greater among Non-Anxiety women compared to the Anxiety Group at PP6 (β = 0.537, SE = 0.243, t = 2.21, p =0.029).

5. Discussion

Over the last 30 years, a wide body of research has made it increasingly clear that immune system factors, chiefly inflammation, are associated with a number of psychiatric disorders, with the bulk of the evidence centering on major depressive disorder, post-traumatic stress disorder, and schizophrenia (50–57). Inflammation, in particular, may play a role in subtypes of mental health diseases or symptom clusters, rather than being a factor for all who suffer a particular disorder (58). The literature supporting a role for altered immune functioning in anxiety is less abundant. Heightened anxiety has been noted in those with inflammatory diseases (59–61). Exposure to stressful stimuli has been shown to activate immune cells to release cytokines and other peripheral inflammatory mediators. This increased inflammation can be related to HPA axis dysregulation in either direction (up or down), depending on the timing of the stressful stimuli – and this dysregulation in turn can be linked to behavioral manifestations and clinical symptoms of anxiety (62–71).

Pregnant and postpartum women, whose immune activity is altered and who suffer from heightened symptoms of anxiety, are an obvious subtype who could be used to explore these likely bidirectional relationships. Our chief aim in this study was to examine the longitudinal relationship between immune functioning and perinatal anxiety without comorbid depression. We identified clusters of cytokines and chemokines in early pregnancy and examined them in a combined score across the peripartum, rather than merely statistically comparing individual cytokines associated with anxiety. By including immune cell populations as well as peripheral cytokines, we identified the immunophenotype of perinatal anxiety. This goal was inspired by recent calls to move “away from simplistic assumptions about immunological changes during pregnancy and toward greater analysis and consideration of the kinetics of immune cells and responses during pregnancy” (72). Finally, another central goal was to study a population with clinically significant symptoms, and to isolate anxiety symptoms by enrolling only women without comorbid depression. We expected that our findings would be consistent with those in the limited prior literature, which by and large shows an increase in inflammatory activity associated with anxiety in the third trimester or shortly after delivery (17, 18, 30, 73).

Our results did not confirm this hypothesis. We identified two clusters of cytokines and chemokines that together explained 64% of the variance. Neither cluster divided neatly into pro-inflammatory or anti-inflammatory functions. Cluster B, which consisted largely of chemokines, did not differ between groups. Cluster A, which included pro-inflammatory cytokines and chemokines in addition to IL-10, was largely reflective of innate immune activity. Cluster A was consistently lower for our Anxiety group than for our Non-Anxiety group during pregnancy. This cluster underwent a sharp increase from pregnancy to postpartum for the Anxiety group, while it declined for the Non-Anxiety group. We also observed alterations in adaptive immune cells, with women in the Non-Anxiety group demonstrating a reduction in B-cell population during pregnancy that rebounded postpartum, while the Anxiety group showed relatively low and stable B cell populations throughout pregnancy and into the postpartum. Further, the anxiety group showed a steep decrease in the ratio of B cells to T cells from pregnancy to postpartum, while the Non-Anxiety women showed a concomitant increase. Women in the Anxiety group also presented with an increased ratio of cytotoxic to helper T cells throughout, a modest increase in the Th1: Th2 ratio across pregnancy (not significant), and a lower proportion of Th17 cells and lower ratio of Th17:T_REG cells in the postpartum when compared Non-Anxiety women. These findings are not consistent with the broadly increased inflammatory activity described previously; (17, 18, 30, 73) if anything, we have also found evidence of decreased activity in pregnancy. Perhaps most important, our two groups presented with sharply different patterns across the pregnant and postpartum periods, with slopes that are often in opposite directions for the two groups.

How can we understand these results in the context of prior literature? One key difference may have to do with the population studied. Compared to our own previous work, the population studied here had three key differences: this population was largely white and of high socioeconomic status; was largely of normal weight; and contained both women with known active psychiatric illness and healthy controls. In our previous study, by contrast, participants were largely Black and Hispanic, of low socioeconomic status, majority overweight or obese, and without known psychiatric illness, resulting in lower levels of anxiety (73). Compared to the work of Leff-Gelman et al., population and measurement tools may explain the different findings. Leff-Gelman’s population of pregnant Mexican women included many with quite severe symptoms; our population, on the other hand, while including women with active illness, had few women with very severe symptoms. Moreover, the different scales used by our two groups may further explain our differences. The Hamilton Anxiety Rating Scale, used by the Leff-Gelman group, contains many somatic items and is designed to capture women with high levels of current anxiety symptoms. The STAI Trait scale, by contrast, identifies those with longstanding personality traits associated with anxiety, rather than those who are necessarily suffering high acute symptoms. Neuroanatomical correlates of the two scales appear to be different (48). In fact, when we further subdivided our population and looked separately at women experiencing a large increase in anxiety symptoms from T2 to T3 (bringing them from the healthy to the clinically significant range), we saw a pattern of increased pro-inflammatory activity. The small number of women included in this analysis did not allow us to draw any conclusions from these data, but they do indicate that women suffering chronic symptoms may be different from those with acute new-onset symptoms, and it will be important for future studies to tease out those differences.

It may also be important to consider our findings in the context of literature concerning immune function in other perinatal morbidities, including perinatal depression. Initial data in perinatal depression seemed to imply that it was the “fourth inflammatory morbidity of pregnancy” (alongside preeclampsia, preterm birth, and gestational diabetes) (19). Considerable literature in the last several years in both pregnancy and postpartum confirms that point (74, 75). Some studies, however, have found the opposite, with decreased inflammatory activity in pregnancy and postpartum associated with depressive symptoms (76–78). Few studies have considered shifts in immune cell populations in perinatal mental illness, and we know of no other studies that have considered such shifts with perinatal anxiety. A reduction in B-cell populations in healthy pregnancy, with rebound postpartum, has been previously demonstrated (79); our findings indicate that women with anxiety may be failing to mount that B cell response in the postpartum. This calls to mind our previous work in postpartum depression, where we found that women with severe postpartum depression failed to mount a rebound in T cells in the postpartum (80)– something we do NOT see with our anxiety population here. Further larger studies incorporating additional B-cell subsets might help to clarify these differences. We also found a lower Th17:Treg ratio in our Anxiety group, whereas other pregnancy morbidities (preeclampsia, recurrent spontaneous abortion) and depression outside of pregnancy have been associated with a higher Th17:Treg ratio (26). One of the few studies that looked at a marker of Th17 activity (IL17C) in perinatal mental illness, however, did find a lower level associated with perinatal depressive symptoms (76). Further studies looking directly at immune cell populations may help to clarify this as well.

Another key to understanding the differences between our findings and others may be timing – indeed, one recent study that found no relationship between inflammatory markers and concurrent depressive symptoms in pregnancy nevertheless found that pregnancy IL-6 and IL-10 were strongly predictive of postpartum depression (74). More studies that carefully distinguish between anxiety and depression, and between illness that begins antenatally and postnatally, are crucially needed to understand these differences. This exploratory study has both strengths and limitations. Strengths include the longitudinal design, the careful characterization of participants and exclusion of those with co-morbid depression, the use of clustering to produce a combined immune score, and the use of immune cells in addition to peripheral cytokines to measure immune functioning. This is the first study of its kind to measure immune cell phenotype across pregnancy and postpartum in women with pure anxiety symptoms, and is one of few studies to do this in a healthy population (81, 82) – as it is still largely unclear how immune cell populations shift even in a healthy pregnancy. Limitations include the small sample size, which fell short of our original target due to COVID-19 restrictions, and the fact that we have no direct measurement of central nervous system immune functioning. Further, we used K cluster analysis to form our two comparison groups. This technique allowed us to capture women’s anxiety scores at all three time points, and thus to include in our Anxiety group women who may have entered the study as “healthy controls” but later exhibited anxiety. It may also mean, however, that our Anxiety group included women whose anxiety decreased over the course of the study. This inability to account for individual variance is a limitation, and in future studies we hope to have sample sizes large enough to allow is to examine additional trajectories. Moreover, while we were striving for a “pure” anxiety phenotype, a small number of subjects (6) did develop depressive symptoms during the course of the study. We elected to include them because our sample size was small to begin with, but it is possible that results would have been different without these women. Both the strengths and the limitations of this study can help set the stage for future research. Additional large studies should separate anxiety from depression; should create additional, more precise, anxiety trajectory groups; should include further B-cell subsets; should test cellular immune response to stress (i.e., stimulation of cells with cortisol); and should carefully distinguish between chronic and acute symptoms, and between symptoms that begin in pregnancy vs. those that begin postpartum. Moreover, we will make real advances in the psychoneuroimmunology of pregnancy and postpartum only when the entire research community can agree on common measures of both psychiatric symptoms and immune functioning – a tall order, but one that will be necessary to advance the field.

6. Conclusions

For the first time, we have longitudinally characterized the immune profile of perinatal anxiety through the assessment of cytokine levels and immune-cell phenotype. In this study, we found that our Anxiety group had a significantly altered immune environment compared to our Non-Anxiety group, suggesting that immune profile differs between women with differing levels of trait anxiety. It is well known that the immune system undergoes drastic alterations during the perinatal period, but it is less known what implications this has on the development of specific psychiatric disorders. Further research is necessary to understand the broader implications of this altered immune environment, such as risk for infection, vaccination response, and autoimmune disorders. We must continue to investigate these differences so that we may provide better therapies for patients suffering from perinatal anxiety.

Supplementary Material

1

Supplementary Figure 1: Panel A and Panel B Gating Strategies

Panel A: Examples of gating in flow cytometry to detect the different cell subsets in a participant. (A) Selection of lymphocytes by forward scatter (FSC: proportional to cell size) and side scatter (SSC: proportional to cellular granularity). (B) Single cell events were selected by gating out cell aggregates on the basis of event area versus event height. (C) Selection of T cells and B cells within the single cell gate. T cells: CD3+CD19- were selected, B cells: CD19+CD3- were selected. (D) Selection of T helper cells and cytotoxic T cells within the T cell gate. T helper: CD4+CD8- cells were selected, cytotoxic T cells: CD8+CD4- were selected.

Panel B: Examples of gating in flow cytometry to detect T cell subsets in a participant. First, live cells were selected by gating out dead cell aggregates on the basis of event area versus event height (data not shown). (A) Selection of lymphocytes by forward scatter (FSC: proportional to cell size) and side scatter (SSC: proportional to cellular granularity). (B) Single cell events were selected by gating out cell aggregates on the basis of event area versus event height. (C) Selection of T helper cells within the single cell gate. CD4+CD3+ cells were selected. (D) Selection of FoxP3+CD25high regulatory T cells (Treg) within the T helper gate. Within the helper T cells, CD25high cells were selected and in this subset FOXP3+ were designated regulatory T cells. (E) Selection of T helper 1 (Th1) and T helper 17 (Th17) cells within the T helper cells. Cells expressing intracellular IFNγ (upper left gate) represent Th1 cells, cells expressing IL-17A (lower right gate) represent Th17 cells. (F) Selection of T helper 2 (Th2) cells within the T helper gate. The gated IL4+IFNγ- cells represent Th2 cells.

Supplementary Figure 2: In vitro PBMC stimulation

PBMCs were cultured and stimulated PMA and ionomycin in the presence of Golgistop to induce T cell activation. (A) Th1, Th17, and Th2 flow cytometry output depicts cell population increase in stimulated cells. (B-D) graphical representation of cellular increase in stimulated cells (red dots) compared to unstimulated cells (black dots) across all three timepoints.