One line of evidence for immune system dysregulation in PPD comes from work examining immune cell status. In an early study by Hucklebridge et al. (83) higher dysphoric mood ratings in parturient women 4 days following birth were associated with lower circulating Th cells, though there was also a weak association showing lower NK cells. More recently, Osborne et al. (84) sought to further define the profile of these T-cell alterations and found that Th1 and Th17 cells were specifically decreased in patients with PPD. These patients had an average onset of symptoms at 7 days postpartum, though blood samples were not collected until approximately 2 months following birth, demonstrating that these peripheral T-cell deficits persist well into the postpartum period.

Other work has examined the link between inflammatory markers and PPD. For example, Maes et al. (85) assessed inflammatory reactivity in the early postpartum period and the relationship to postpartum mood. Along with non-pregnant controls, serum samples were collected from pregnant women the day before delivery and again at 1 and 3 days postpartum (85). In the women whose depressive scores increased postpartum relative to prenatal measures, there was elevated IL-6 and IL-6 receptor (IL6-R) in the serum at all three time points assessed. Though the timeline of this study is more consistent with a “baby blues” phenotype rather than PPD, subsequent work has looked at later postpartum timepoints. In Bränn et al. (86), levels of pro-inflammatory CXCL1, IL-18, and TNF-receptor family proteins were found to be higher in patients with depressive symptoms at 8 weeks postpartum. Notably, these findings did not change when controlling for timing of onset of depressive symptoms which for some women occurred during pregnancy and thus would indicate that there is likely to be persistent inflammation in depressed individuals postpartum. Another study found increased levels of IL-6 and IL-8 along with decreased IL-2 at 8 weeks postpartum, effects that were similarly associated with depressive symptoms (87). It is important to consider that these latter findings were obtained from a population of patients with severe and suicidal PPD (e.g., symptoms that required hospitalization) suggesting that more severe depressive symptoms may be characterized by a different cytokine profile than less severe cases of PPD, as Bränn et al. (86) found no alterations in IL-6 and IL-8.

Some evidence suggests that immune alterations can predict subsequent mood behavior postpartum. Examining the genetic expression of blood samples taken shortly following delivery, Segman et al. (88) reported transcriptional alterations in markers of immune activation, including those associated with the pro-inflammatory chemokine CXCL10 and the TNF-receptor family. These correlated with later development of PPD. At the protein level, Corwin et al. (89) found that elevated peripheral IL-1β 14 days postpartum predicted increased depressive symptoms at 28 days postpartum. IL-6 levels were also measured in this study, but unlike IL-1β, IL-6 levels did not predict depressive symptoms. In contrast to the lack of relationship between postpartum IL-6 and depressive symptoms, Simpson et al. (90) found that lower levels of IL-6 and IL-10 in blood samples collected in the third trimester of pregnancy were associated with increased depressive symptoms 12 weeks postpartum. Individuals with and without depressive symptoms postpartum showed no differences in third-trimester levels of TNF-α or C-reactive protein (CRP) which is often used as a proxy measure of non-specific inflammation. In contrast, Liu et al. (91) showed that in the days following delivery, elevated IL-6, as well as elevated CRP, were predictive of developing PPD within the first 6 months following parturition. These findings are more in line with other depressed populations, such that patients diagnosed with Major Depressive Disorder outside of the postpartum period have been found to have increased levels of peripheral IL-6 (92).

It should be noted that several studies have reported no association between immune function and PPD. Skalkidou et al. (93) did not find any alterations in serum IL-6 collected at 5 days, 6 weeks, and 6 months after delivery that were associated with PPD symptoms. Moreover, Groer et al. (54) did not uncover any link between mood and serum cytokine levels (IL-2, IL-12, IL-10, IFNγ, TNF-α, IL-4, and IL-6) as assessed at various time points from 1 week to 6 months postpartum. In line with this, Osborne et al. (94), despite finding increased IL-6 and CCL3 in the third trimester in those with greater depressive symptoms while pregnant, did not find any persistence of this inflammation postpartum. Whether the inflammation during pregnancy was predictive of later PPD is unclear based on the analyses conducted.

Some of the inconsistencies across studies is likely attributable to variations in methodology including heterogeneity in timepoints, specific assessments of depression (e.g., clinically diagnosed vs. symptomology), immune markers assessed, sample handling prior to measurement, and the specific measurement techniques used. Demographic differences in subject populations examined could be another moderating factor in observed effects (95, 96). For example, when examining race as a demographic factor, differences in serum cytokine levels were not found but greater depressive symptoms correlated with greater production of IL-6, IL-8, and TNF-α following lipopolysaccharide (LPS) stimulation of blood samples from African American, but not White, women at 7–10 weeks postpartum (97). Because the relationship between depressive symptoms and cytokine markers in peripartum women appears to differ between Black and White women, incorporating racial demographic information as a moderator to future studies is critical. Doing so may also begin to address the underlying causes of racial disparities in PPD prevalence with Black women being more than twice as likely to experience PPD as White women (98). Some work highlights the importance of life history, particularly experiences with the stress of racism, as a contributor to inflammatory changes that drive risk for negative perinatal outcomes, but this has not been well investigated for PPD (99, 100). In one study, higher levels of the inflammatory marker IL-6 were documented in Black women during pregnancy and postpartum (95), although this same study showed neither IL-6 nor TNF-α were associated with depressive symptoms. Other work by Accortt et al. (101) examined pregnant African American women, investigating vitamin D and immune factors during pregnancy as mediators of postpartum depressive symptomology. Their results show that when lower in vitamin D and higher in pro-inflammatory cytokine levels, specifically IL-6 and the relative expression of IL-6 compared to IL-10, self-reported symptoms of PPD were higher. Given the limited number of studies to date, further exploration of the role of inflammation in racial, as well as ethnic, disparities in PPD is warranted (98).

Examining the relative balance or the ratio of both pro-inflammatory and anti-inflammatory factors has gained traction in the literature beyond the single study by Accortt et al. (101) described above. For example, Corwin et al. (102) examined blood cytokine levels during late gestation (third trimester), and again postpartum on days 7 and 14, as well as 1, 2, 3, and 6 months following delivery. Their results show that a significant predictor of developing PPD symptoms was the ratio of IL-8 to IL-10 on postpartum day 14. Furthermore, Bränn et al. (103), when examining 92 markers of inflammation with a multiplex assay from plasma samples collected during late pregnancy (7–23 days before delivery), found that 40 different inflammatory markers were lower in those reporting PPD symptoms at 6–8 weeks postpartum compared with controls. This downregulation, which included decreased IL-8, IL-10, and IL-17, was universal regardless of whether the marker was pro- or anti-inflammatory. These studies indicate that any disruption in the typical cytokine milieu regardless of the direction of the shift, along with the relative pro- vs. anti-inflammatory tone, could be important in mediating later PPD onset. Additionally, the ability of postpartum immune alterations to impact PPD appears to be dynamic and transient, such that certain timepoints may be more influential than others/confer more vulnerability.

All of the studies discussed thus far assessed peripheral inflammatory markers. There are only two studies to date that have looked at the possibility of central cytokine alterations, which may better reflect the inflammatory milieu that the CNS is exposed to during the peripartum period. In a study by Boufidou et al. (81), plasma samples were collected during early labor and cerebrospinal fluid (CSF) samples collected before epidural analgesia was given. In the CSF, positive associations were found between both TNF-α and IL-6 and severity of depressive symptoms, while only plasma TNF-α was positively correlated with depressive outcomes. This work highlights central immune activation in PPD and suggests that neuroimmune changes might not be consistent with what is occurring peripherally. This was further corroborated in a study by Miller et al. (82) which found no correlation between peripheral cytokine levels and cytokine levels in the CSF of pregnant women at the time of delivery. Notably, despite no peripheral alterations in cytokines, Miller et al. (82) did find increased CSF expression of IL-1β, IL-23, and IL-33 and associations with increased prenatal depressive symptoms. Postpartum time points were not examined to determine whether these central alterations persisted or whether they were predictive of later mood disorders. It should be noted, however, that even if peripheral cytokines do not correlate directly with central cytokine levels, this does not diminish the value of peripheral cytokines as potential biomarkers for peripartum mood disorder risk or lessen potential insights into disorder pathophysiology, it simply means that the picture is complex.

The increased capability of sequencing technology in recent years has also been able to provide valuable insight beyond cytokine expression at the gene expression and protein levels as measured in most other studies. A recent genome-wide association study (104) sought to identify genetic markers linked to PPD via RNA sequencing of whole blood samples. They found that immune biomarkers, such as genes responsible for encoding TNF-family receptors as well as A disintegrin and metalloproteinase with thrombospondin motifs (ADAMTs), were significantly upregulated in women showing depressive symptoms 2 months postpartum. Further, increased matrix metalloproteinase-8 (MMP8) during late pregnancy was found to be predictive of PPD onset. In addition to strengthening previous evidence pointing to peripheral TNF alterations, imbalances in MMPs and the ADAMTs subfamily of those proteins could be an interesting avenue for future work examining immune-mediated restructuring of the interstitial matrix as well as BBB integrity, both of which are known to be targets for these molecules (105–107), hold relevance to psychiatric symptoms (108, 109), and could point to a potential mechanism of central infiltration of immune cells from the periphery.

Recapitulating postpartum mental illness phenotypes in animal models provides valuable insights and furthers our understanding of these complex disorders in a way that cannot be accomplished at the clinical level. Readers are referred to comprehensive reviews on the topic by Perani and Slattery (110) and Li and Chou (111). In brief, two rodent models that have been used to investigate immune and neuroimmune underpinnings of PPD are stress exposure during pregnancy (i.e., gestational stress) and hormone-simulated pregnancy. Though the specific stress paradigms implemented varies across studies, gestational stress models provide construct validity as stress exposure during human pregnancy is a significant risk factor in later experiencing PPD (112, 113). In hormone-simulated pregnancy models, typically conducted in ovariectomized females, exogenous estrogen and progesterone are administered at high doses for several weeks to mimic the sustained elevations in these hormones experienced during normal rodent pregnancy, after which hormone administration is withdrawn to result in abrupt decreases in hormone levels seen in postpartum females at parturition. In addition to the relevant risk factors of stress reactivity and hormonal fluctuations, these models also provide face validity as animals will display behavioral phenotypes similar to what is seen in human PPD, such as increased behavioral despair, anhedonia, and critically, impairments in maternal care (110, 111).

Only one study, to our knowledge, has used a hormone-simulated pregnancy model to explore immune contributions to PPD by examining T cell alterations in peripheral blood samples (114). This study found that after hormone withdrawal, depressive-like behavior increased and this was accompanied by a shift from immune activation to immune suppression as indicated by relative expression of Th cells. Interestingly, both the behavioral and immune effects were reversed by treatment with the antidepressant, fluoxetine, as well as with the herbal medicine Shen-Qi-Jie-Yu-Fang. Overall, this study points to increased inflammatory profiles as underlying postpartum depressive-like behavior mediated by endocrine influences.

More frequently, stress models have been employed to investigate immune changes underlying PPD. O'Mahony et al. (115) subjected pregnant rats to restraint stress from gestational days 14–21. Blood and brain samples were taken at one month postpartum. Increases in blood levels of IL-1β, TNF-α, and IL-10 were found in stressed animals, though no changes in PFC, HPC, or hypothalamus were detected. In other work, Poscillo and Schwarz (78) used a sub-chronic stress model in which pregnant rats were subjected to forced swim from gestational days 16–22. They did not find behavioral differences that were dependent on stress exposure, though their postpartum rats did show a depressive-like phenotype. In these same animals, they detected alterations in PFC and HPC cytokine levels, such that there was increased IL-6 and decreased IL-1β in the PFC and increased IL-4 in the HPC. In a follow up study (116), it was found that central administration of an IL-6R antagonist to unstressed postpartum females attenuated anhedonia as measured in the sucrose preference test, suggesting that central inflammatory signaling is involved in mediating the anhedonic component of depressive-like behavior in mothers. Finally, Tan et al. (117) examined the role of inflammation in PPD and the response to antidepressants. Unpredictable chronic mild stress (UCMS) was applied from gestational days 7–16 to control mice (i.e., C57 wildtype) and microglia-specific autophagy deficient mice as autophagy is a cellular process that can modulate the inflammatory response. Mice were then treated postpartum with the antidepressant, fluoxetine, beginning one week following delivery. Results showed that fluoxetine ameliorated stress-induced depressive-like behavior, but only partially in autophagy deficient mice. Further, it was found that BDNF, autophagy-associated proteins, and IL-10 were upregulated in the HPC of fluoxetine-treated control mice exposed to UCMS who also showed a reduction in inflammatory factors. In contrast, autophagy-deficient mice displayed diminished autophagy, greater inflammation, and reduced BDNF expression when compared with control mice. Given that reduced BDNF expression has also been reported in women with PPD (118), this study suggests that inflammatory processes might be regulating this change, while also providing evidence that fluoxetine may mediate its antidepressant effect in PPD by influencing the inflammatory response.

Preliminary evidence from our lab also implicates altered microglia activity in PPD as we have found increased expression of the phagocytic markers ITGAM and C1q in the PFC at gestational day 21 following UCMS occurring from gestational days 7–20 (119). Moreover, in our model we found increased expression of IL-1β and IFNγ in the PFC at gestational day 21, while there were no differences in peripheral cytokines at this time point or any changes in samples taken at postpartum day 8 (119). Along with previous work from our lab that has shown reduced dendritic spines in the postpartum PFC following gestational stress (120), and other evidence implicating microglia in engulfment of synaptic elements (121–123), these studies suggest that central inflammation and microglia-mediated synaptic remodeling may be resulting from gestational stress. Further exploration of the connections between central immune and neural changes would thus be critical in understanding how stress exposure during pregnancy is contributing to PPD-like behavioral phenotypes.

Despite evidence suggesting immune links to anxiety outside of the peripartum period (124–126), relatively little work has examined how immune factors relate to PPA. This may be at least in part because PPA in generally not well studied and because most perinatal immune studies focus on depression.

Maes et al. (85) were the first to investigate an association between postpartum anxiety and peripheral immune markers. They found that parturient women with greater anxiety scores had lower leukemia inhibitory factor receptor (LIFR) prenatally (1 day before delivery) as well as increased IL-6 and IL-1 receptor antagonist postpartum (1 and 3 days after delivery). The lower prepartum anti-inflammatory capacity could distinguish the onset of PPA from that of PPD since, as previously discussed, this same study found that there was an increase in IL-6R at the late stages of pregnancy and no differences in LIFR in women with higher depression scores. The short time frame of this study leaves the remaining questions of whether these immune alterations persist into the postpartum period and whether on their own they are sufficient in giving rise to postpartum mental illness beyond the state measures assessed here.

A more recent study by Osborne et al. (94) also studied possible immune system dysregulation in PPA and the extent to which these might change over time. A broad range of cytokines were examined at 5 timepoints across the peripartum period, including once during each trimester and at 6 and 24 weeks postpartum. During the first and third trimester, both blood IL-6 and CCL3 were significantly higher in those who were anxious than those who were not. Significant differences were not detected at any of the postpartum time points. However, given that antenatal anxiety is a strong predictor of subsequently developing PPA, it would be of interest to know whether any of these immune changes were predictive of later developing anxiety, but this was not examined.

The rodent postpartum period is characterized by a reduction in anxiety-like behavior (127). This altered emotionality is thought to be adaptive and necessary for providing proper maternal care (127, 128). Manipulations (i.e., certain gestational stress paradigms, maternal separation) which disrupt typical anxiety-like and fear responses during this period and concurrently impact maternal behavior, may provide insights into PPA. However, distinguishing between depressive- and anxiety-like phenotypes remains a challenge, as manipulations which induce anxiety-like behavior in rodents also often affect depressive-like behavior.

Little work in rodent models has examined immune function as an etiological factor contributing to PPA. A study conducted by Wallace et al. (129) explored the relationship between prenatal inflammation and the development of PPA-like behavior, although the primary focus of the study was on hypertension. Using a rat model of severe preeclampsia that produced upregulated splenic CD4⁺ and CD8⁺ T-cells, they assessed both depressive- and anxiety-like behavior between 2 and 4 weeks postpartum as well as BBB integrity at postpartum week 6. Rats with severe preeclampsia displayed greater anxiety-like behavior as demonstrated by significantly more time spent in the closed arms of an elevated plus maze compared to healthy postpartum controls and more marble burying in the marble burying task. Anhedonia, a component of depressive-like behavior as measured with a sucrose preference test, was not impacted. The results of this study also show that T-cell suppression during pregnancy prevented the PPA-related behaviors as well as the increased BBB permeability that occurred. Whether these mechanisms are translatable to onset of PPA in the absence of preeclampsia remains to be investigated. Nonetheless, as clinical data does point to positive associations between preeclampsia and PPA, this model could also provide insights into any causality underlying this relationship that is currently lacking in human data.

A role for immune dysregulation in PP is supported by evidence showing that women who experience PP have higher rates of other complications related to immune dysfunction during their pregnancy (130). For example, immune imbalance occurs in preeclampsia and preeclampsia is a significant risk factor in later developing first onset PP (131). Similarly, other work has found that ~20% of those with PP also had autoimmune thyroid disorder compared to 5% in postpartum controls (132). Interestingly, anti-N-methyl-D-aspartate (NMDA) receptor encephalitis is an autoimmune disorder that has been associated with onset of psychosis symptoms (133, 134). Thus, Bergink et al. (135) sought to investigate whether undiagnosed autoimmune encephalitis may be occurring in patients with PP by conducting an immunohistochemistry-based screening for antibodies. Results from this study showed that in a large cohort of patients with PP, a subset (~4%) had anti-NMDA receptor antibodies suggesting this mechanism might hold relevance to at least some cases of PP.

A growing literature points to links between immune alterations and the development of PP in otherwise healthy mothers (130). One study assessed circulating immune cell types in a population of women who were experiencing the first onset of PP, along with healthy postpartum participants and non-pregnant age-matched healthy controls (53). Those who had PP were reported to have significantly decreased amounts of circulating lymphocytes relative to healthy postpartum participants. Concurrently, those with PP had significantly higher percentages of circulating blood monocytes compared to both healthy postpartum and non-pregnant control participants (53). Another group (136) sought to immunophenotype cells associated with first-onset PP. While there was not a difference associated with PP in overall T_reg cells or naïve T_reg cells, there was an increase in memory T_reg cells compared to healthy postpartum controls. In addition, women with PP had decreased cytotoxic NK cells and elevated regulatory NK cells. Together, these two studies point to altered peripheral immune cells associated with PP, although looking beyond broad classes of cell types (such as distinguishing between cytotoxic vs. regulatory NK cells) will likely provide greater insight into disease etiology.

Another line of work has aimed to establish the profile of circulating cytokines in women with PP. An examination of IL-6 levels by Sathyanarayanan et al. (137) found no difference in women with first onset PP as compared with healthy postpartum women. This finding was confirmed in another study which showed no difference in IL-6 levels of women with PP, those at risk (because of bipolar disorder, schizoaffective disorder or a history of PP), and a group of healthy postpartum controls (138). However, there are data showing higher peripheral serum IL-8 (137) and elevated high sensitivity CRP (hsCRP) in PP as compared to healthy postpartum women (138). In addition to IL-6, IL-8 and CRP, CCL2 (also known as monocyte chemoattractant protein [MCP]-1) has been examined as a possible marker for PP. Although one study found evidence for increased expression of CCL2 in women with first-onset PP (53), this finding was not replicated (137, 138) and thus further investigation is needed. Interestingly, when examining a population that included those with a prior history of PP, BPD, or schizophrenia rather than first onset, Hazelgrove et al. (139) found that inflammatory markers (IL-1β, IL-6, IL-8, TNFα or hsCRP) measured during the third trimester of pregnancy were not predictive of an additional PP episode. Thus, while evidence suggests that dysregulated immune function is associated with PP, more work is needed to parse out the exact role of altered cytokine profiles and how they may differ across individuals with varying predispositions to the disorder.

Case studies of obstetric patients during the COVID-19 pandemic also point toward altered immunity contributing to the onset of PP. Indeed, there is evidence suggesting a possible link between PP and SARS-Cov-2 infection (140, 141). Although in some women who developed COVID-related PP there was a personal or family history of psychiatric illness, this was not universal. Critically, none of the case reports took direct measures of immune activity but increased pro-inflammatory cytokines, including IL-6, TNFα, IL-8, and IL-10, have been observed in the plasma of those with severe COVID-19 infections (142). The additional stress and social isolation of the pandemic when occurring during an already vulnerable period could also be influential environmental triggers despite the varying degrees of predisposition mentioned here. Further exploration of PP in the context of two-hit models (immune perturbation plus environmental stressor) is likely to be a fruitful approach for determining precise etiology of the disorder.

The use of animal models to investigate immune underpinnings of PP is almost completely unexplored. Largely, this can be attributed to the lack of well-validated preclinical models of PP. One of two studies that have attempted to model PP in animals was conducted in 2007 by Quilter et al. (143) who profiled genetic expression in a porcine model and found distinct signatures in the hypothalamus of mothers that committed infanticide compared to those who did not. Moreover, cross-species analysis with human tissue showed promise for several candidate genes. The pig provides many advantages to immunology research due to its similarity to the human immune system (144) and thus additional work in a porcine model could have potential to advance understanding of role of the immune system in PP.

Another more recent study (145) which could lend mechanistic insight to the development of PP employed a pharmacological rodent model involving inhibition of steroid sulfatase (SS) in maternal mice. SS is an enzyme that is responsible for the conversion of dehydroepiandrosterone (DHEA) from its sulfated form (DHEAS). Decreased DHEA has been associated with increased pro-inflammatory cytokine expression in the postpartum period (146) while PP and DHEAS levels are positively correlated (147). Inhibition of the SS enzyme in maternal mice increased anxiety-like behavior in an elevated plus maze and resulted in sensorimotor gating deficits as indicated by a decrease in the startle response; both of these behavioral abnormalities were alleviated by treatment with an antipsychotic (145). Reductions in the startle response have been implicated in other neuropsychiatric disorders that serve as the biggest risk factors for developing PP, namely schizophrenia and BPD (148). In rodent models, attenuated startle responsiveness has been found to be partially mediated by peripheral IL-1β signaling (149), but whether IL-1β is involved in the reduced startle response of maternal rodents after SS inhibition was not examined. Indeed, immune measures were not the primary focus of the study although expression of three immune-related genes (CCL2, CXCL1, and IL-33) were assessed and a significant increase in CCL2 following SS inhibition was reported (145). It should be noted that although SS deficiency has been hypothesized to be a risk factor for PP (150), experimental evidence to date is lacking (151, 152). Moreover, as noted above, the results in the human literature regarding CCL2 levels in PP also appear contradictory (53, 137, 138). Thus, possible links among SS, CCL2, and PP onset require further study. Overall, further validation and development of novel preclinical models would be a valuable step toward advancing immune-centered approaches and translatability in PP.