Synaptic density

One component of gray matter is synapses. Synapses are connections between neurons that are formed on protruding dendritic spines and facilitate communication (i.e., excitation or inhibition of the downstream neuron) that ultimately give rise to cognitive function (Glausier & Lewis, Reference Glausier and Lewis2013). Reductions in excitatory glutamatergic synapses and dendritic spine density have been consistently observed in the dorsolateral prefrontal cortex of individuals with chronic schizophrenia, and may also be reduced in the medial prefrontal cortex, superior temporal cortex, and hippocampus (Broadbelt, Byne, & Jones, Reference Broadbelt, Byne and Jones2002; Garey et al., Reference Garey, Ong, Patel, Kanani, Davis, Mortimer and Hirsch1998; Harrison & Eastwood, Reference Harrison and Eastwood2001). In the prefrontal cortex, these deficits are most consistently observed in excitatory pyramidal neurons in layers III and possibly layer V of the dorsolateral and medial prefrontal cortex. These layers are thought to contain synapses connecting cortical pyramidal neurons and GABAergic interneurons as well as synapses connecting cortical and thalamic neurons (Broadbelt et al., Reference Broadbelt, Byne and Jones2002; Glantz & Lewis, Reference Glantz and Lewis2000). Reduced synaptic connectivity in these layers and regions may be associated with functional and cognitive deficits observed in schizophrenia. For example, reductions in layer III pyramidal spine density in cortical regions including the superior frontal cortex, orbitofrontal cortex, and portions of the temporal cortex may be associated with poorer cortical white matter connectivity in individuals with schizophrenia (van den Heuvel et al., Reference van den Heuvel, Scholtens, de Reus and Kahn2016). Broadly, synaptic connections in layer III of the dorsolateral prefrontal cortex form microcircuits of cortical pyramidal neurons that provide the backbone of working memory (Arnsten, Reference Arnsten2011). Though specific cellular mechanisms of other cognitive processes are less clearly determined or tied to specific regions or layers, activity within the medial prefrontal cortex and superior temporal cortex has been associated with cognitive processes such as memory retrieval and source monitoring (Brandt, Bergström, Buda, Henson, & Simons, Reference Brandt, Bergström, Buda, Henson and Simons2014; Sugimori, Mitchell, Raye, Greene, & Johnson, Reference Sugimori, Mitchell, Raye, Greene and Johnson2014). This suggests that reduced synaptic connectivity in these regions and layers could disrupt cognitive processes in ways that contribute to the symptoms of schizophrenia. For example, deficits in memory of whether a phrase was heard or imagined, a process that involves both the medial prefrontal cortex and superior frontal cortex, could contribute to the experience of hallucinations (Brandt et al., Reference Brandt, Bergström, Buda, Henson and Simons2014; Sugimori et al., Reference Sugimori, Mitchell, Raye, Greene and Johnson2014). Reductions in synaptic density and function have also been observed in the hippocampus, and may be more prevalent in CA4 and CA3 relative to CA1 (Harrison & Eastwood, Reference Harrison and Eastwood2001). Though the implications of deficits in these particular regions are not well established, hippocampal dysfunction has been associated primarily with cognitive deficits in schizophrenia, such as verbal memory (Saykin et al., Reference Saykin, Shtasel, Gur, Kester, Mozley, Stafiniak and Gur1994). In sum, reduced synaptic density and functioning in the prefrontal cortex may contribute to reductions in gray matter and cognitive deficits observed in schizophrenia.

Deficits in interneuron functioning

Postmortem studies of schizophrenia cases have also identified reduced numbers of parvalbumin-containing GABAergic interneurons in the gray matter of the hippocampus and across the prefrontal cortex (Bennett, Reference Bennett2011). Interneurons are activated by excitatory pyramidal neurons, and in turn inhibit the activation of pyramidal neurons. The firing of interneurons promotes coordinated firing of cortical pyramidal neurons, which is critical to complex cognitive function (Chung, Fish, & Lewis, Reference Chung, Fish and Lewis2016). Reduced interneuron density could interfere with this coordination, interfering with cognitive processes based in the prefrontal cortex and hippocampus (described above). The contribution of disrupted pyramidal neuron communication to the development of schizophrenia is also supported by evidence implicating NMDA receptor dysfunction (Coyle, Tsai, & Goff, Reference Coyle, Tsai and Goff2003). NMDA receptors are glutamatergic postsynaptic membrane receptors that, when activated, promote firing of the postsynaptic neuron and help strengthen the synapse (Homayoun & Moghaddam, Reference Homayoun and Moghaddam2007). NMDA receptors are present on many neuron types, and their function is particularly important for interneurons. If postsynaptic NMDA receptors on interneurons fail to be activated, the interneuron is less likely to successfully fire on its target pyramidal neurons. This results in poor regulation of pyramidal neuron activity (e.g., excitation/inhibition imbalance), which undermines cognitive functioning (Homayoun & Moghaddam, Reference Homayoun and Moghaddam2007). The influence of NMDA receptor function on symptoms of schizophrenia is also supported by ketamine studies, which have been shown to induce psychotic symptoms by blocking NMDA receptors in the prefrontal cortex (Krystal et al., Reference Krystal, Karper, Seibyl, Freeman, Delaney, Bremner and Charney1994). Thus, reduced interneuron density and function may also contribute to the symptoms of schizophrenia.

White matter disruption

Disruption to white matter may also influence gray matter volume and contribute to the emergence of these symptoms. The speed and efficiency of communication between neurons is improved with myelination, the wrapping of fatty myelin-containing cells around axons of neurons (Davis et al., Reference Davis, Stewart, Friedman and Buchsbaum2003). Decreased white matter volume in the prefrontal cortex, particularly in the orbitofrontal region, has been associated with increased negative symptoms in individuals with schizophrenia (Sanfilipo et al., Reference Sanfilipo, Lafargue, Rusinek, Arena, Loneragan, Lautin and Wolkin2000). In addition, reduced integrity has been observed in white matter tracts throughout the brain (Lim et al., Reference Lim, Hedehus, Moseley, de Crespigny, Sullivan and Pfefferbaum1999). Deficits in certain tracts have been linked with cognitive deficits in patients, such as poorer attention and working memory associated with alterations in the cingulum bundle (Kubicki et al., Reference Kubicki, Westin, Nestor, Wible, Maier, Kikinis and Shenton2003). Alterations in myelination have been observed in clinically high-risk individuals and first episode patients as well (Witthaus et al., Reference Witthaus, Brüne, Kaufmann, Bohner, Özgürdal, Gudlowski and Juckel2008), though whether these changes are associated with onset of the disorder remains unclear (Cannon et al., Reference Cannon, Chung, He, Sun, Jacobson, van Erp and Heinssen2015).

Connection to dopamine regulation

Deficits in synaptic density, interneuron function, and myelination converge on disruption of communication between cortical and hippocampal pyramidal neurons. Cumulatively, this dysfunction may lead to downstream changes in other areas of the brain that also contribute to the symptoms of schizophrenia. For instance, increased firing of dopaminergic neurons in the striatum has been induced by disruption of pyramidal cell firing in the prefrontal cortex and hippocampus in animal models (Kim et al., Reference Kim, Rossi, Aryal, Racz, Kim, Uezu and Soderling2015; Lodge & Grace, Reference Lodge and Grace2007). Excess dopamine in the striatum is a hallmark finding in schizophrenia, and the efficacy of antipsychotic medication is attributed to countering this neurotransmitter excess (Creese, Burt, & Snyder, Reference Creese, Burt and Snyder1976). Dopamine in the striatum has been implicated in reward and salience-related signaling (Schultz, Reference Schultz1998). Excess dopamine has been associated with salience being attributed to unimportant stimuli in individuals with schizophrenia, and may contribute to the emergence of delusions (Roiser, Howes, Chaddock, Joyce, & McGuire, Reference Roiser, Howes, Chaddock, Joyce and McGuire2013). The role of the prefrontal cortex in regulating subcortical dopamine signaling further supports the theory that that disruption of coordinated communication between prefrontal cortical pyramidal neurons may play an early role in the etiology of schizophrenia.

Onset of schizophrenia

A prominent neurobiological theory of schizophrenia suggests that positive symptoms emerge when a key threshold of prefrontal and temporal cortical synaptic density is crossed during synaptic pruning in adolescence (Cannon, Reference Cannon2015; Feinberg, Reference Feinberg1982; Weinberger, Reference Weinberger1987). This theory integrates the neurobiological findings described above and may help explain the proximal onset of positive symptoms of schizophrenia in adolescence. However, the question remains as to whether psychosis results from a rapid disintegration of synaptic functioning, interneuron coordination, and/or myelination, or if deficits in these systems are present early in development and compound over time. Clarifying the timing of the emergence of these deficits can further clarify etiological pathways to psychosis and potential time points of interaction with exposure to factors such as inflammation to further increase risk for the development of the disorder.

Influence of inflammation on the development of implicated cellular mechanisms

Deficits in cognitive and motor development may be more apparent in individuals with schizophrenia who were exposed to prenatal inflammatory challenges compared to cases who did not experience prenatal insults (Brown et al., Reference Brown, Cohen, Harkavy-Friedman, Babulas, Malaspina, Gorman and Susser2001, Reference Brown, Vinogradov, Kremen, Poole, Deicken, Penner and Schaefer2009; Ellman et al., Reference Ellman, Yolken, Buka, Torrey and Cannon2009). This pattern suggests that inflammation may synergize with genetic risk to perturb the development of cellular mechanisms that are implicated in the emergence of schizophrenia. Rodent models of maternal immune activation have identified many changes in neural structure and behavior that parallel those observed in schizophrenia, including anatomical changes such as reductions in synaptic density between layer III pyramidal neurons in the dorsolateral prefrontal cortex (Weir et al., Reference Weir, Forghany, Smith, Patterson, McAllister, Schumann and Bauman2015), reduced gray matter volume across the cortex (Short et al., Reference Short, Lubach, Karasin, Olsen, Styner, Knickmeyer and Coe2010), and reduced myelin (Rodts-Palenik et al., Reference Rodts-Palenik, Wyatt-Ashmead, Pang, Thigpen, Cai, Rhodes and Bennett2004), as well as cognitive changes such as deficits in social interactions (Shi, Fatemi, Sidwell, & Patterson, Reference Shi, Fatemi, Sidwell and Patterson2003) and impairment in associative learning and memory tasks (Golan, Lev, Hallak, Sorokin, & Huleihel, Reference Golan, Lev, Hallak, Sorokin and Huleihel2005). In adolescents at clinical high risk who ultimately converted to psychosis, higher rates of thinning in the superior and medial frontal cortex and medial orbitofrontal cortex were correlated with both the severity of positive symptoms and levels of inflammatory markers in peripheral blood samples (Cannon et al., Reference Cannon, Chung, He, Sun, Jacobson, van Erp and Heinssen2015). Understanding the influence of inflammation on the development of these cellular mechanisms that provide the foundation for subsequent cognition and behavior could lead to more insight into etiological pathways to schizophrenia and ultimately provide novel opportunities for intervention and prevention.

To understand the influence of prenatal inflammation on these cellular mechanisms, it is important to first establish typical neurodevelopmental patterns. Next, we will review neurobiological development from the prenatal period through childhood, adolescence, and early adulthood, and consider how neural changes may relate to cognitive development across this period.

Pyramidal neurons, interneurons, and synapses

See Figure 1 for a visual timeline of neurobiological development. The development of the CNS starts with the formation of the neural tube around the third gestational week, which splits into a series of vesicles that ultimately form the forebrain, midbrain, and hindbrain (Tau & Peterson, Reference Tau and Peterson2010). The forebrain vesicle further splits into the telencephalon (cerebral cortex, basal ganglia, and hippocampus) and diencephalon (including the thalamus and hypothalamus; Catts et al., Reference Catts, Fung, Long, Joshi, Vercammen, Allen and Shannon Weickert2013; Tau & Peterson, Reference Tau and Peterson2010). Proliferation of neural progenitor cells that will later mature into neurons begins in the dorsal and ventral regions of the telencephalon in weeks 5–6 and continues through middle gestation (~23 weeks). Cortical pyramidal neurons are generated in the dorsal (lateral ventrical) region, and GABAergic interneurons originate in the ventral (early basal ganglia/subventricular) region (Catts et al., Reference Catts, Fung, Long, Joshi, Vercammen, Allen and Shannon Weickert2013; Tau & Peterson, Reference Tau and Peterson2010). Neuronal migration peaks between the start of the second trimester (~13 weeks) and about 20 weeks, following a subcortical to cortical pattern (Catts et al., Reference Catts, Fung, Long, Joshi, Vercammen, Allen and Shannon Weickert2013; Tau & Peterson, Reference Tau and Peterson2010). Neurons originating in the dorsal ventricular zone are typically guided by radial glia to their designated location through the developing cortex, whereas interneurons migrate in parallel to the developing cortex before reaching their destination (Tau & Peterson, Reference Tau and Peterson2010). Neuronal migration to the cortex gives rise to the preplate, a transient structure that forms the first identifiable cortical layer (Tau & Peterson, Reference Tau and Peterson2010). By weeks 7–8, the formation of the cortical plate splits the preplate into the subplate and the marginal zone. The subplate grows rapidly in middle gestation (weeks 18–22) and serves as a placeholder for early connections between cortical and subcortical neurons. By the beginning of the third trimester (~26 weeks), the subplate dissolves and the marginal zone (cortical layer I) and cortical plate (cortical layers II–VI) form the basis of the mature organizational structure of the cortex. Neuron proliferation and migration end at about 30 weeks with the completion of association cortices and commissural regions (de Graaf-Peters & Hadders-Algra, Reference de Graaf-Peters and Hadders-Algra2006; Tau & Peterson, Reference Tau and Peterson2010), though accumulating evidence suggests that interneuron proliferation and migration from the ventral subventricular zone may continue at low levels into adulthood (Catts et al., Reference Catts, Fung, Long, Joshi, Vercammen, Allen and Shannon Weickert2013).

Figure 1. (Color online) Timeline of neurodevelopmental processes and specific prenatal cytokine and complement elevations associated with schizophrenia. IL-8, 13–32 weeks; IL–6, 26–32 weeks; TNFa, 26–42 weeks; IL-4, IL5, IL-13, C1q, 37–42 weeks.

Maturation of neurons, including proper alignment of cells and neuronal communication via axon and dendritic outgrowth and the formation of synaptic connections, begins around 16 to 24 weeks as migration is completed (Catts et al., Reference Catts, Fung, Long, Joshi, Vercammen, Allen and Shannon Weickert2013; Tau & Peterson, Reference Tau and Peterson2010). Around this time, cells in regions such as the thalamus, brainstem, and nucleus accumbens send preliminary axonal projections to form synapses with cortical placeholder neurons in the subplate (Tau & Peterson, Reference Tau and Peterson2010). These connections are refined and transferred to enduring neurons in the cortical plate around 24–28 weeks gestation as the subplate dissolves, paving the way for more sophisticated cortical development (Tau & Peterson, Reference Tau and Peterson2010). Neural cells begin to organize into mini columns, and cortical layers composed of different cell types and connections with other regions start to form. Evidence of cortical layering is first present in primary sensory and motor regions starting around week 25 (Tau & Peterson, Reference Tau and Peterson2010). In gestational weeks 24–34, cortical neurons first form synaptic connections nonspecifically with other local cells in the marginal zone. Ultimately, these synaptic connections become cortical layer I, the most superficial cortical layer (Tau & Peterson, Reference Tau and Peterson2010). Cortical pyramidal neurons then begin to send axonal projections to ipsilateral and contralateral cortical regions (layer III) and integrate projections from other regions such as the thalamus and brainstem (layer IV; Tau & Peterson, Reference Tau and Peterson2010). By the late third trimester, pyramidal neurons begin to preferentially form local synaptic connections with other pyramidal neurons in different layers of the same column of cells (layers V and IV) and with other pyramidal neurons in the same layer (layer II; Tau & Peterson, Reference Tau and Peterson2010). By about week 32, the foundation of cortical layers is present throughout the cortex, including integration of projections of all major neurotransmitter systems and a variety of neuronal (i.e., pyramidal and interneuron) and glial cell types (Tau & Peterson, Reference Tau and Peterson2010).

Once the foundation of cortical layering is established, rates of synaptic formation increase rapidly. Rates of growth in synaptic density peak at a rate of almost 40,000 new synapses formed every second from the third trimester through the first neonatal year (de Graaf-Peters & Hadders-Algra, Reference de Graaf-Peters and Hadders-Algra2006; Tau & Peterson, Reference Tau and Peterson2010). In the third trimester and first few neonatal months, synaptic density grows at a particularly rapid rate in primary sensory and motor regions (Huttenlocher & Dabholkar, Reference Huttenlocher and Dabholkar1997), and connections formed are primarily between local neurons (e.g., cortical layers I and II; Tau & Peterson, Reference Tau and Peterson2010). Refinement of synaptic connections via pruning of competing synapses co-occurs at a low rate with synaptic formation in early development, but synaptic outgrowth dominates at this stage of development (de Graaf-Peters & Hadders-Algra, Reference de Graaf-Peters and Hadders-Algra2006).

Glia

Two subtypes of glial cells follow similar patterns of development as neurons. Like neurons, both astroctyes and oligodendrocytes are derived from neural progenitor cells and are generated in that order once neurogenesis for neurons destined for a particular neural region has been completed (Eroglu & Barres, Reference Eroglu and Barres2015). Proliferation of astrocytes and oligodendrocytes is at its greatest rate from about 20 to 40 weeks gestation, but begins as early as 10 weeks and continues after birth for oligodendrocytes (de Graaf-Peters & Hadders-Algra, Reference de Graaf-Peters and Hadders-Algra2006). Astrocytes tend to localize to synapses and play key roles in supporting synaptic formation and plasticity (Eroglu & Barres, Reference Eroglu and Barres2015). The timing of gliogenesis coincides with synaptogenesis, likely due to the important role of astrocytes in this process (Semple, Blomgren, Gimlin, Ferriero, & Noble-Haeusslein, Reference Semple, Blomgren, Gimlin, Ferriero and Noble-Haeusslein2013). Oligodendrocytes form myelin in their mature state and wrap around axons, increasing the speed of signal transmission within neurons (Eroglu & Barres, Reference Eroglu and Barres2015). Myelination follows similar patterns as synaptic growth, beginning in the precursor to the basal ganglia around 14 weeks, in the subcortical striatum around 28 weeks, and in the cortex around 35 weeks (de Graaf-Peters & Hadders-Algra, Reference de Graaf-Peters and Hadders-Algra2006). Rates of myelination peak in the third trimester and first 3 months of life (Catts et al., Reference Catts, Fung, Long, Joshi, Vercammen, Allen and Shannon Weickert2013). For instance, whole-brain volume that contains mature white matter increases from 1% to 5% between gestational weeks 36 and 40, and continues after birth (Catts et al., Reference Catts, Fung, Long, Joshi, Vercammen, Allen and Shannon Weickert2013).

One class of glia, microglia, does not originate from neural progenitor cells but instead from immune progenitors outside of the brain (Deverman & Patterson, Reference Deverman and Patterson2009). Immature microglia are thought to migrate into the CNS very early in neurodevelopment, possibly in response to markers of programmed cell death that begin starting at 7 weeks gestation and continue through the first trimester (de Graaf-Peters & Hadders-Algra, Reference de Graaf-Peters and Hadders-Algra2006). Microglia seem to be in an activated state during the late second and early third trimester through the early postnatal period, and play a critical role in promoting survival of neurons as well as shaping synaptic strength and density (Bilbo & Schwarz, Reference Bilbo and Schwarz2009). Microglia are thought to mature into their quiescent adult form within the early neonatal period and only take an active form in response to later immune challenges (Bilbo & Schwarz, Reference Bilbo and Schwarz2009).

Neonatal neural structure

By the end of prenatal development, much of the cellular and structural framework of the brain has been established. This progress can also be observed with noninvasive structural and functional scans of the developing human brain. For instance, fetal magnetic resonance imaging scans have shown that gyri and sulci appear in the thickening developing cortex as synaptogenesis accelerates during the third trimester, and total brain volume increases nearly threefold during this period (Tau & Peterson, Reference Tau and Peterson2010). Functionally, it seems that basic visual, sensorimotor, and auditory processing circuits are functional at term birth (39–44 weeks; Fransson, Ulrika, Blennow, & Lagercrantz, Reference Fransson, Ulrika, Blennow and Lagercrantz2011), and possibly a preliminary default mode network connecting prefrontal and parietal regions as well (Tau & Peterson, Reference Tau and Peterson2010). This provides the foundation for the emergence of basic sensorimotor and visual functioning at the time of birth (Tau & Peterson, Reference Tau and Peterson2010). Continued growth and refinement of the neural framework during the postnatal period gives rise to increasingly complex cognition and behavior.

Prenatal cytokine exposure and risk for schizophrenia

A variety of prenatal risk factors for schizophrenia have been linked with increases in maternal and fetal cytokine levels, such as viral infections with IL-6 (Cunningham, Campion, Teeling, Felton, & Perry, Reference Cunningham, Campion, Teeling, Felton and Perry2007), bacterial infections with IL-1β and TNFα (Cai et al., Reference Cai, Pan, Pang, Evans and Rhodes2000), and hypoxic/ischemic birth insults with IL-1β and IL-2 (Girard, Kadhim, et al., Reference Girard, Kadhim, Larouche, Roy, Gobeil and Sébire2008; Girard, Larouche, et al., Reference Girard, Larouche, Kadhim, Rola-Pleszczynski, Gobeil and Sébire2008). Only a handful of studies have examined whether elevations in individual cytokines during pregnancy are associated with risk for schizophrenia in human samples. Epidemiologic studies examining alterations in maternal cytokine levels during pregnancy in association with risk for schizophrenia have identified second/third trimester elevations in maternal IL-8 (and not IL-1β, IL-6, or TNFα; Brown, Hooton, et al., Reference Brown, Hooton, Schaefer, Zhang, Petkova, Babulas and Susser2004; Ellman et al., Reference Ellman, Deicken, Vinogradov, Kremen, Poole, Kern and Brown2010), third trimester elevations in IL-6 for males and low levels of TNFα for females (and not IL-1β, IL-8, or IL-10; Goldstein et al., Reference Goldstein, Cherkerzian, Seidman, Donatelli, Remington, Tsuang and Buka2014), and elevations in TNFα (and not IL-1β, IL-2, IL-6, or IL-8) at the time of delivery (Buka, Tsuang, Torrey, Klebanoff, Wagner, et al., Reference Buka, Tsuang, Torrey, Klebanoff, Wagner and Yolken2001; Figure 1). In addition, elevations in a combined measure of maternal IL-4, IL-5, and IL-13 at birth have been identified as a protective factor (Allswede, Buka, Yolken, Torrey, & Cannon, Reference Allswede, Buka, Yolken, Torrey and Cannon2016). Levels of cytokines from neonatal blood samples have not consistently been associated with later risk for schizophrenia. This suggests that earlier prenatal development may be a particularly vulnerable time frame for cytokine-mediated influences on neurodevelopment that increase risk for schizophrenia (Nielsen et al., Reference Nielsen, Agerbo, Skogstrand, Hougaard, Meyer and Mortensen2014).

Role of cytokines in typical neurodevelopment

Rodent studies suggest a role for cytokines in most aspects of prenatal neurodevelopment throughout the brain, including neuronal and glial migration and differentiation, synaptic maturation, and myelination (Depino, Reference Depino2013; Ratnayake et al., Reference Ratnayake, Quinn, Walker and Dickinson2013). Both pro- and anti-inflammatory cytokines have been observed in the brain as early as 5 weeks gestation and exhibit different patterns of elevation across early gestation (Dziegielewska et al., Reference Dziegielewska, Moller, Potter, Ek, Lane and Saunders2000; Mousa, Seiger, Kjaeldgaard, & Bakhiet, Reference Mousa, Seiger, Kjaeldgaard and Bakhiet1999). Certain cytokines appear to play specific roles in distinct neurodevelopmental processes. For example, IL-6, but not IFNγ, has been shown to influence the generation of neural progenitor cells in the adult subventricular zone, and may play a similar role in early stages of neurogenesis (Gallagher et al., Reference Gallagher, Norman, Woodard, Yang, Gauthier-Fisher, Fujitani and Miller2013). IL-6 has also been implicated in mediating the transition from neurogenesis to astrogenesis (Semple et al., Reference Semple, Blomgren, Gimlin, Ferriero and Noble-Haeusslein2013). IL-1β has been shown to guide the migration of cortical neurons (Ma et al., Reference Ma, Li, Zhang, Yang, Zhu, Yuan and Jiang2014), and may be particularly effective at promoting the maturation of midbrain neural progenitor cells to a dopaminergic phenotype and regulating their survival (Meyer et al., Reference Meyer, Feldon and Yee2009). IL-1β, IL-6, TNFα, and IFNγ have been implicated in regulation of cortical neuron dendrite development (Barish, Mansdorf, & Raissdana, Reference Barish, Mansdorf and Raissdana1991; Gilmore et al., Reference Gilmore, Fredrik Jarskog, Vadlamudi and Lauder2004), and TNFα has been shown to regulate expression of AMPA receptors at synapses (Eroglu & Barres, Reference Eroglu and Barres2015). IL-1β has also been implicated in regulating proliferation and maturation of oligodendrocytes and myelin levels (Cai, Lin, Pang, & Rhodes, Reference Cai, Lin, Pang and Rhodes2004; Cammer & Zhang, Reference Cammer and Zhang1999; Favrais et al., Reference Favrais, van de Looij, Fleiss, Ramanantsoa, Bonnin, Stoltenburg-Didinger and Gressens2011). Though certain cytokines, such as IL-6, appear to be implicated in many of these early neurodevelopmental processes, other cytokines are present during this period as well and have been studied less extensively. For example, similar elevation patterns in the first 10 weeks of gestation to IL-6 have been observed in IL-1β, TNFα, IFNγ, IL-4, and IL-10 (Mousa et al., Reference Mousa, Seiger, Kjaeldgaard and Bakhiet1999), suggesting that these cytokines could also potentially play a role in early neurodevelopment. These studies suggest that cytokines play specific roles in the maturing brain and that disruption of individual cytokines may have different effects on development.

Potential influence of cytokine elevations linked with schizophrenia risk on neurodevelopment

Potential interactions between cytokines and genetic vulnerability to schizophrenia

Elevations in maternal cytokines have the potential to influence neurobiological systems that are implicated in schizophrenia. However, as most individuals exposed to prenatal inflammation do not develop schizophrenia, these exposures are likely not sufficient to cause schizophrenia (Mittal et al., Reference Mittal, Ellman and Cannon2008). In addition, the broad role for cytokines in neurodevelopment throughout the brain is inconsistent with patterns of specific deficits observed in the disorder, such as lower synaptic density in specifically layer III of the dorsolateral prefrontal cortex (Garey et al., Reference Garey, Ong, Patel, Kanani, Davis, Mortimer and Hirsch1998). This pattern suggests that elevations in cytokines may increase risk for schizophrenia by interacting with a variety of other vulnerability factors, such as genetic variants, that in aggregate influence certain aspects of neurodevelopment that increase risk for the development of psychosis later in life. One such interaction may be with genetic variants that have been implicated in schizophrenia, including disrupted-in-schizophrenia-1 (DISC1). DISC1 was originally associated with schizophrenia, bipolar disorder, and major depression in a Scottish pedigree (St. Clair et al., Reference St. Clair, Blackwood, Muir, Carothers, Walker, Spowart and Evans1990) and has since been associated with schizophrenia in many studies, though findings are mixed (Brandon & Sawa, Reference Brandon and Sawa2011). DISC1 is thought to be involved in a variety of neurodevelopmental processes, including neuronal proliferation and migration as well as dendritic spine regulation and synaptic maintenance (Brandon & Sawa, Reference Brandon and Sawa2011). In rodents, fetuses exposed to maternal viral infection who carried a mutant DISC1 variant exhibited more extreme changes in pro- and anti-inflammatory cytokines in the brain compared to groups exposed to only one risk factor (Abazyan et al., Reference Abazyan, Nomura, Kannan, Ishizuka, Tamashiro, Nucifora and Pletnikov2010). Neural changes were observed in adulthood, including poorer hypothalamic–pituitary–adrenal axis reactivity and decreased serotonin neurotransmission and dendritic spine density in the hippocampus, lower amygdala volumes, and decreased lateral ventricle volume. Behaviorally, these offspring exhibited anxiety and depressive-like responses and altered social behavior. However, these behavioral deficits were only observed if the DISC1 variant was expressed across the life span. Subsequent work identified that similar effects are observed in fetuses exposed to IL-6 during the prenatal period with a variant of DISC1 that produces schizophrenia-related biological and behavioral phenotypes (Lipina, Zai, Hlousek, Roder, & Wong, Reference Lipina, Zai, Hlousek, Roder and Wong2013). Similar interactions between genetic variants and prenatal cytokine exposure on neurodevelopment and behavior have been observed for Nurr1, implicated in dopaminergic system development (Vuillermot et al., Reference Vuillermot, Joodmardi, Perlmann, Ove Ogren, Feldon and Meyer2012), and CHRNA7, a component of nicotinic acetylcholine receptors that are important for cognitive function (Wu, Adams, Stevens, Chow, & Patterson, Reference Wu, Adams, Stevens, Chow and Patterson2015).

Broadly, these findings suggest that genetic variants implicated in psychiatric illness and cytokine exposure during the prenatal period may interact in a way that potentiates prenatal exposure to pro-inflammatory cytokines and downstream effects on neurodevelopment and behavior. Though the effect of a given interaction between a genetic variant and a particular cytokine is likely small, the aggregate influence of many such interactions could be sufficient to alter neurodevelopment in ways that increase risk for the development of schizophrenia. Many of the 108 schizophrenia-associated genetic loci identified in the largest genome-wide significance study to date (Ripke et al., Reference Ripke, Neale, Corvin, Walters, Farh, Holmans and O'Donovan2014) may be preferentially expressed during middle to late fetal development (Ohi et al., Reference Ohi, Shimada, Nitta, Kihara, Okubo, Uehara and Kawasaki2016), and inflammation tends to induce changes in many cytokines at once (Saito, Nakashima, Shima, & Ito, Reference Saito, Nakashima, Shima and Ito2010). There is potential for huge numbers of these interactions to take place in response to maternal inflammation during pregnancy, but only for individuals with variants in many of these genes. This possibility may explain why most individuals exposed to infection during prenatal development or birth complications do not exhibit signs of adverse neurodevelopment, yet for certain individuals these exposures alter neurodevelopment in ways that increase risk for schizophrenia. The potentiating effect of these interactions may also explain why effect sizes tend to be small for variants of genes implicated in schizophrenia in genetic studies. Further investigation of interactions between genetic variants associated with schizophrenia and cytokine exposure on neurodevelopment may yield more insight into mechanistic pathways that can lead to schizophrenia, and could ultimately provide novel platforms for prevention.

Summary

In sum, cytokines may mediate the association between prenatal inflammation and risk for schizophrenia by affecting neurodevelopment. Elevations in maternal cytokines during pregnancy have been associated with risk for schizophrenia. Rodent models suggest these cytokines could affect the development of systems implicated in the disorder such as synaptic formation, oligodendrocyte maturation, and interneuron function. It is plausible that the composite effect of many interactions between genetic variants implicated in schizophrenia and cytokines leads to specific changes in neurodevelopment that increase risk for the disorder. Further research to assess these potential interactions could produce more precise insight into the etiology of schizophrenia.

Potential links between complement exposure and risk for schizophrenia

Complement gene expression and protein levels are altered following immune challenge with bacterial (Gewurz, Shin, & Mergenhagen, Reference Gewurz, Shin and Mergenhagen1968) and viral components (Sahu, Isaacs, Soulika, & Lambris, Reference Sahu, Isaacs, Soulika and Lambris1998). There is increasing recognition that the complement system can be activated by noninfectious antigens as well, such as hypoxic-ischemic birth complications (Cowell, Plane, & Silverstein, Reference Cowell, Plane and Silverstein2003), broad markers of inflammation such as C-reactive protein (Stephan et al., Reference Stephan, Barres and Stevens2012), and antibodies to gluten or dairy (Severance et al., Reference Severance, Alaedini, Yang, Halling, Gressitt, Stallings and Yolken2012). This pattern suggests that it is plausible that the associations observed between prenatal inflammatory insults and risk for schizophrenia may be mediated in part by complement system activation. However, there is a paucity of studies examining complement proteins during pregnancy and their association with risk for schizophrenia. One human study has identified an association between maternal C1q levels at delivery and risk for schizophrenia among offspring (Severance et al., Reference Severance, Gressitt, Buka, Cannon and Yolken2014; Figure 1). More rodent and human research is needed to assess whether levels of other complement proteins at specific time points during development are associated with risk for schizophrenia.

Complement proteins in neurodevelopment

Preliminary evidence suggests that the complement pathway is involved in typical neurodevelopment (Stephan et al., Reference Stephan, Barres and Stevens2012). The complement proteins are clearly implicated in normative experience-dependent synaptic pruning (Sekar et al., Reference Sekar, Bialas, de Rivera, Davis, Hammond, Kamitaki and McCarroll2016; Stephan et al., Reference Stephan, Barres and Stevens2012). Glial-produced C1q and C3, and increasingly C4, have been localized to immature neurons and have been shown to be critical for synaptic pruning in rodent models of retinal ganglion neuron maturation (Chu et al., Reference Chu, Jin, Parada, Pesic, Stevens, Barres and Prince2010; Sekar et al., Reference Sekar, Bialas, de Rivera, Davis, Hammond, Kamitaki and McCarroll2016; Stephan et al., Reference Stephan, Barres and Stevens2012). In contrast, the complement system does not seem to be implicated in synaptic formation (Stephan et al., Reference Stephan, Madison, Mateos, Fraser, Lovelett, Coutellier and Barres2013). Complement proteins C3 and C5 may promote the proliferation and migration of neural precursor cells in young and adult rodents in many regions of the brain, including the hippocampus and cerebellum (Bénard et al., Reference Bénard, Raoult, Vaudry, Leprince, Falluel-Morel, Gonzalez and Fontaine2008; Moriyama et al., Reference Moriyama, Fukuhara, Britschgi, He, Villeda, Molina and Holers2011; Shinjyo, Ståhlberg, Dragunow, Pekny, & Pekna, Reference Shinjyo, Ståhlberg, Dragunow, Pekny and Pekna2009). Initial evidence suggests that C1q may be associated with a subset of inhibitory interneurons, though the specific role played by C1q is unclear (Stephan et al., Reference Stephan, Madison, Mateos, Fraser, Lovelett, Coutellier and Barres2013). In addition, complement proteins have been implicated in priming helper T cells, peripheral immune cells that are similar to microglia, toward pro- or anti-inflammatory phenotypes (Yamamoto, Fara, Dasgupta, & Kemper, Reference Yamamoto, Fara, Dasgupta and Kemper2013). This set of findings suggests that the complement system is involved in a variety of neurodevelopmental processes, but the details of this involvement remain to be elucidated.

Potential influence of complement elevations on neurodevelopment of schizophrenia

The clearest potential etiological connection between the complement system and schizophrenia is via synaptic pruning (Sekar et al., Reference Sekar, Bialas, de Rivera, Davis, Hammond, Kamitaki and McCarroll2016). Most synaptic pruning takes place after birth (Figure 1), so complement elevations during the prenatal period may not influence synaptic pruning specifically. However, elevations in complement in the late second to third trimester, when glial programming may be particularly vulnerable (Bilbo & Schwarz, Reference Bilbo and Schwarz2009), could sensitize microglia to future immune challenges and have implications for future neurodevelopment. For example, reactive astrocytes respond to inflammatory markers by expressing molecules including C1q that affect synapse formation (Eroglu & Barres, Reference Eroglu and Barres2015). It is possible that sensitized microglia may release more C1q in response to inflammation, which could contribute to overpruning of synapses in childhood or adolescence (i.e., multiple hits hypothesis; Bilbo & Schwarz, Reference Bilbo and Schwarz2009; Meyer, Reference Meyer2013; Monji et al., Reference Monji, Kato and Kanba2009).

The complement system has also been associated with proliferation and migration of pyramidal neurons and interneurons (Bénard et al., Reference Bénard, Raoult, Vaudry, Leprince, Falluel-Morel, Gonzalez and Fontaine2008; Stephan et al., Reference Stephan, Madison, Mateos, Fraser, Lovelett, Coutellier and Barres2013). This suggests that elevations in complement proteins during the prenatal period may interfere with the development of these systems. However, not enough is known about the typical role of complement proteins in these processes or about exposures associated with risk for schizophrenia to make clear predictions about how this relates to the etiology of schizophrenia. Additional research identifying the role of complement proteins in typical neurodevelopment and the consequences of exposure to altered levels of these proteins in rodents, and studies examining prenatal complement exposure and risk for schizophrenia, are needed to elucidate the potential role of complement proteins in the development of schizophrenia.

Potential interactions between complement and genetic vulnerability to schizophrenia

Like cytokines, elevations in complement proteins likely contribute to a neurodevelopmental risk trajectory for schizophrenia specifically via interactions with genetic variants implicated in the disease. Investigation of potential interactions in rodent models is needed to explore this possibility. An avenue that may be fruitful is exploration of interactions between genetic variants that affect microglial function and prenatal complement exposure. Genetic variants that interfere with microglial regulation or prime microglia toward a pro-inflammatory phenotype could plausibly be associated with more microglial-mediated neural damage and/or pruning abnormalities in childhood and adolescence. Individuals with such variants who are exposed to elevated complement in the late prenatal period could undergo microglial sensitization, leading to exacerbated neural damage in response to inflammatory challenges in later childhood and adolescence. Microglial activity may be particularly important in the early development of psychosis in adolescence, as elevated microglial activation as assessed using positron emission tomography of TSPO radioligands has been observed in ultra-high risk prodromes (Bloomfield et al., Reference Bloomfield, Selvaraj, Veronese, Rizzo, Bertoldo, Owen and Howes2016) and findings with this technique are mixed in first episode (Hafizi et al., Reference Hafizi, Tseng, Rao, Selvanathan, Kenk, Bazinet and Mizrahi2017; van Berckel et al., Reference van Berckel, Bossong, Boellaard, Kloet, Schuitemaker, Caspers and Kahn2008) and chronic schizophrenia (Doorduin et al., Reference Doorduin, de Vries, Willemsen, de Groot, Dierckx and Klein2009; Kenk et al., Reference Kenk, Selvanathan, Rao, Suridjan, Rusjan, Remington and Mizrahi2015). It is possible that individuals with particularly high microglial activity in adolescence in addition to another vulnerability factor, such as NMDA receptor hypofunction, could be at even higher risk of developing psychosis. NMDA receptor hypofunction, resulting in reduced transmission of excitatory glutamate into the postsynaptic neuron, has long been thought to be a key component of the pathophysiology of schizophrenia (Olney, Newcomer, & Farber, Reference Olney, Newcomer and Farber1999). NMDA receptor activity at synapses is thought to be critical to the growth and stabilization of dendritic spines (Arnsten, Reference Arnsten2011), and spines that are stable are thought to be less vulnerable to being pruned (Glausier & Lewis, Reference Glausier and Lewis2013). It is possible that individuals who have less effective NMDA receptors have greater numbers of weak spines, which could be less likely to survive normative synaptic reshaping during the first few years of life. Individuals with overactive microglia and inappropriately weak synaptic connections may be particularly likely to experience abnormally high rates of synaptic pruning in adolescence. These individuals may be at particularly high risk to cross below a threshold of synaptic density needed to support coordinated neuronal firing, resulting in poorer cognitive function (Homayoun & Moghaddam, Reference Homayoun and Moghaddam2007) and potentially the onset of psychosis (Cannon, Reference Cannon2015; Feinberg, Reference Feinberg1982).

In all, there are many possible interactions between variants affecting microglial priming and regulation, synaptic strength and functioning, and prenatal complement system activity that could be involved in the etiology of schizophrenia. Further work assessing interactions such as these would allow us to test theories linking the complement system with synaptic pruning in schizophrenia and further our understanding of the role that the complement system may play in the etiology of schizophrenia.

Summary

Initial evidence suggests that complement proteins may mediate the association between prenatal inflammation and risk for schizophrenia by affecting neurodevelopment. The complement system is most clearly implicated in synaptic pruning, but may play a role in neuronal growth and migration and glial priming as well. Interactions between complement proteins and genetic variants in these systems may contribute to specific patterns observed in schizophrenia, such as faster rates of synaptic pruning in adolescence. Further research is needed to clarify the role of complement proteins in neurodevelopmental processes, identify periods of development and specific proteins that are associated with schizophrenia risk, and assess interactions with other vulnerability markers to establish how complement proteins may play a role in the etiology of schizophrenia.

Typical immune signaling during pregnancy and risk patterns

There is some indication that the maternal immune system changes throughout pregnancy, likely to promote tolerance of fetal tissue (Girardi, Bulla, Salmon, & Tedesco, Reference Girardi, Bulla, Salmon and Tedesco2006). For example, the first trimester of pregnancy is characterized by high levels of anti-inflammatory IFNγ levels and low levels of typically pro-inflammatory IL-6 in maternal serum, whereas by the third trimester of pregnancy IL-6 levels are high and IFNγ levels are low (Aris, Lambert, Bessette, & Moutquin, Reference Aris, Lambert, Bessette and Moutquin2008). In contrast, C3, C4, and C5 levels are higher in the serum of pregnant women compared to nonpregnant women, and remain at consistent levels in maternal serum throughout pregnancy (Richani et al., Reference Richani, Soto, Romero, Espinoza, Chaiworapongsa, Nien and Mazor2005). These findings suggest that immune pathway activity may change across pregnancy, with important implications for maternal and fetal reactivity to immune challenges. However, only a few studies have assessed maternal cytokine levels longitudinally (Holtan et al., Reference Holtan, Chen, Kaimal, Creedon, Enninga, Nevala and Markovic2015; Ross et al., Reference Ross, Miller, Culhane, Grobman, Simhan, Wadhwa and Borders2016), and typical patterns of marker levels across pregnancy have yet to be established. Characterizing the typical patterns of cytokines and complement in maternal and fetal serum will help clarify which alterations in cytokine and complement levels are atypical for that period in development, and could yield insight into why similar prenatal exposures can have different effects at different times. For instance, preterm-equivalent rodents exposed to bacteria and hypoxic-ischemic injury at birth exhibited upregulation of only anti-inflammatory cytokines, whereas term-equivalent rodents exposed to these two insults exhibited upregulation of both pro-inflammatory and anti-inflammatory cytokines (Brochu, Girard, Lavoie, & Sébire, Reference Brochu, Girard, Lavoie and Sébire2011). This knowledge may also nominate periods of development and cellular systems that may be particularly associated with risk for schizophrenia. Thus far, the studies that have examined whether elevations in maternal cytokines during pregnancy are associated with risk for schizophrenia in human samples have been cross-sectional (Allswede et al., Reference Allswede, Buka, Yolken, Torrey and Cannon2016; Brown, Hooton, et al., Reference Brown, Hooton, Schaefer, Zhang, Petkova, Babulas and Susser2004; Buka, Tsuang, Torrey, Klebanoff, Bernstein, et al., Reference Buka, Tsuang, Torrey, Klebanoff, Bernstein and Yolken2001; Ellman et al., Reference Ellman, Deicken, Vinogradov, Kremen, Poole, Kern and Brown2010; Goldstein et al., Reference Goldstein, Cherkerzian, Seidman, Donatelli, Remington, Tsuang and Buka2014). Elucidating different prenatal inflammatory trajectories and examining their association with postnatal development may further specify different etiological pathways that can increase risk for schizophrenia.

Inflammatory patterns associated with specific exposures

Prenatal immune activation increases fetal inflammation broadly, but certain exposures may be more strongly associated with specific markers. Generally, injections of bacterial and viral components are known to induce elevations in IL-1β, IL-6, and TNFα in maternal serum. Viral components may also particularly activate IFNγ (Bilbo et al., Reference Bilbo, Biedenkapp, Der-Avakian, Watkins, Rudy and Maier2005; Boksa, Reference Boksa2010; Cai et al., Reference Cai, Pan, Pang, Evans and Rhodes2000; Smith et al., Reference Smith, Li, Garbett, Mirnics and Patterson2007). Elevations in IL-1β and IL-2 have been observed following hypoxic-ischemic injury (Girard, Kadhim, et al., Reference Girard, Kadhim, Larouche, Roy, Gobeil and Sébire2008; Girard, Larouche, et al., Reference Girard, Larouche, Kadhim, Rola-Pleszczynski, Gobeil and Sébire2008). Nevertheless, there is some evidence suggesting that distinct cytokines have been implicated in mediating the effects of maternal inflammation on neurodevelopment. The effects of viruses on neurodevelopment may be mediated by changes in IL-6 specifically, and not IL-1β or IFNγ (Smith et al., Reference Smith, Li, Garbett, Mirnics and Patterson2007). In contrast, IL-1β may be particularly involved in neurodevelopmental changes resulting from bacteria (Girard, Tremblay, Lepage, & Sébire, Reference Girard, Tremblay, Lepage and Sébire2010). Assessing cognitive changes in response to viral and bacterial infection similarly suggests many common deficits and the potential for distinct deficits. Exposure to each pathogen has been shown to lead to poorer prepulse inhibition, spatial learning, and novel object recognition. Latent inhibition may be particularly disrupted following viral infection (Boksa, Reference Boksa2010). Further work in both rodent models and longitudinal human samples with birth record data is needed to assess whether certain prenatal exposures are consistently associated with specific cytokines and neurodevelopmental deficits or if the effects are consistent across exposures. There may be less specificity to exposure type in the complement system, as C3 seems to be activated by bacteria, viruses, and hypoxic-ischemic injury (Cowell et al., Reference Cowell, Plane and Silverstein2003; Gewurz et al., Reference Gewurz, Shin and Mergenhagen1968; Sahu et al., Reference Sahu, Isaacs, Soulika and Lambris1998). However, more work is needed to assess whether certain exposures are associated with distinct patterns of complement system activity and neurodevelopmental deficits as well. If there is specificity associated with particular exposures, this information could eventually be used to develop treatments for neonates designed to counteract the specific associated inflammatory patterns. If exposures exhibit the same general pattern, a common treatment could be used to support optimal neurodevelopment for all neonates.

Potential for synergism with multiple inflammatory insults

As similar cytokines and complement proteins tend to be secreted in response to multiple types of prenatal insults, exposure to multiple prenatal inflammatory events may lead to a potentiated inflammatory effect that puts individuals at greater risk for altered neurodevelopment. For example, rats exposed to either bacterial infection in the last few days of gestation or hypoxic-ischemic insult at birth exhibited an increase in proinflammatory IL-1β and IL-2 levels and a decrease in anti-inflammatory IL-1RA levels in the brain in the first 2 postnatal days, and rats exposed to both exhibited the greatest changes in these levels (Girard, Kadhim, et al., Reference Girard, Kadhim, Larouche, Roy, Gobeil and Sébire2008; Girard, Larouche, et al., Reference Girard, Larouche, Kadhim, Rola-Pleszczynski, Gobeil and Sébire2008). In adulthood, rodents exposed to both prenatal insults exhibited the greatest changes in behavior relative to controls, such as reduced exploratory behavior (Girard et al., Reference Girard, Sébire, Brochu, Briota, Sarret and Sébire2012). Similarly, rats exposed to viral infection before hypoxic-ischemic insult at birth exhibited greater reductions in myelin basic protein in the neonatal period compared to rodents exposed only to hypoxia-ischemia (Stridh et al., Reference Stridh, Mottahedin, Johansson, Valdez, Northington, Wang and Mallard2013). Higher levels of pro-inflammatory cytokines and lower levels of anti-inflammatory microglia were observed following viral infection and prior to hypoxic-ischemic insult, and could explain the stronger effect of exposure to viral infection on neural damage (Stridh et al., Reference Stridh, Mottahedin, Johansson, Valdez, Northington, Wang and Mallard2013). However, similar studies have identified that infection prior to hypoxia-ischemia can have a preconditioning effect that mitigates damage to the developing brain (Mallard & Hagberg, Reference Mallard and Hagberg2007). This contradiction could be explained by differences in length of time between infection and hypoxic-ischemic injury. As typical inflammatory responses first exhibit an increase in pro-inflammatory factors and then an increase in anti-inflammatory factors, different lengths of time between the first and second insult could result in different environmental contexts when the second insult hits, yielding different effects on overall inflammation (Mallard & Hagberg, Reference Mallard and Hagberg2007). A more detailed understanding of these factors could help us identify infants who are at particular risk for cytokine-mediated neurodevelopmental damage and potentially inform the strength of anti-inflammatory treatment needed.

Correspondence between maternal and fetal immune signaling

Examining the association between maternal cytokine and complement levels and risk for schizophrenia among offspring is a critical step toward translation of rodent literature to humans. However, these studies are vulnerable to the assumption that elevations in maternal serum reflect changes in the fetal brain. Rodents studies suggest that changes in maternal cytokine levels in serum are sometimes, but not always, consistent with cytokine changes in the placenta, fetal serum, and fetal brain (Ashdown et al., Reference Ashdown, Dumont, Ng, Poole, Boksa and Luheshi2006; Garay, Hsiao, Patterson, & McAllister, Reference Garay, Hsiao, Patterson and McAllister2013; Gilmore, Jarskog, & Vadlamudi, Reference Gilmore, Jarskog and Vadlamudi2005; Urakubo et al., Reference Urakubo, Jarskog, Lieberman and Gilmore2001). Apparent inconsistencies may be influenced by the timing of measurement of cytokine levels. For example, injection of bacterial components has been shown to cause a rapid increase in IL-1β (~1 hr) and a more gradual and protracted increase in TNFα (~4–24 hr) in the fetal brain (Cai et al., Reference Cai, Pan, Pang, Evans and Rhodes2000). This suggests that some elevations in cytokines in the fetal brain may be missed if only one measurement is taken. True differences between maternal and fetal cytokine levels may be influenced by the transfer of the inflammatory response at the placenta. Cytokines are thought to not be able to cross from maternal to fetal tissue, with the possible exception of IL-6 (Meyer et al., Reference Meyer, Feldon and Yee2009). The maternal inflammatory response at the boundary of the placenta is thought to induce an inflammatory response from the fetal side of the placenta, much like the transduction of cytokines at the blood–brain barrier (Meyer et al., Reference Meyer, Feldon and Yee2009). Thus, it is possible that changes in cytokine signaling are made at these interfaces. Less is known about the consistency of complement signaling in maternal and fetal serum and tissues. Studies of the maternal–fetal interface suggest that complement inhibitors are highly expressed to prevent potential damage of placental tissue (Girardi et al., Reference Girardi, Bulla, Salmon and Tedesco2006). Additional rodent studies that examine the consistency of cytokine and complement expression in maternal serum and the fetal brain across multiple time points could help clarify which cytokines are transduced from maternal to fetal tissue and which are not. Further investigation of roles of specific immune proteins in neurodevelopment, and comparison of these findings to changes observed (e.g., neural structure) in offspring whose mothers were exposed to that protein during neurodevelopment, could also help assess whether the changes observed in offspring match those that would be expected. Once sufficient rodent literature has been established to generate predicted effects of changes to specific maternal immune proteins on fetal development, these predictions could be assessed in human samples utilizing maternal serum collected during pregnancy and neonatal imaging measures.

Assessing cytokine pathways

Though previous investigations of maternal cytokine levels in association with risk for psychosis have predominantly focused on single immune markers, assessing clusters of cytokines may more comprehensively reflect the dynamics of cytokine signaling pathways (Saito et al., Reference Saito, Nakashima, Shima and Ito2010). One such pathway could group cytokines based on their association with pathways of immune cell activation in peripheral blood mononuclear cells: pro-inflammatory M1 macrophage polarization, which in turn activates pro-inflammatory helper T (Th) cell types Th1 and Th17 subtypes (cytokines: TNFα, IFNγ, GM-CSF, IL-1β, IL-2, IL-6, IL-12, and IL-17), and anti-inflammatory M2 macrophage polarization, which in turn activates anti-inflammatory Th2 and Treg cell types (CSF1, IL-4, IL-5, IL-10, and IL-13; Italiani & Boraschi, Reference Italiani and Boraschi2014; Ponomarev et al., Reference Ponomarev, Veremeyko and Weiner2013). Grouping cytokines by their common pathways may improve signal-to-noise ratio as well as facilitate interpretation of findings within the biological context. However, it is also possible that examining trajectories of single cytokines accounts for more signal in association with neurodevelopmental outcomes and may be most informative. Comparing groupings of cytokines versus individual assessment would help clarify which approach may be most beneficial in investigations of prenatal immune activation and neurodevelopmental outcomes, such as risk for schizophrenia.

Interactions between immune proteins and other biological mechanisms

Neurodevelopmental changes observed in response to maternal immune activation are not the result of cytokines and complement proteins alone. Cytokines are known to interact with a variety of other factors that influence neurodevelopment. For example, IL-1β has been shown to interfere with the neuroprotective effects of brain-derived neurotrophic factor in cultured neurons (Tong, Balazs, Soiampornkul, Thangnipon, & Cotman, Reference Tong, Balazs, Soiampornkul, Thangnipon and Cotman2008). Cytokines are known to interact with the hypothalamic–pituitary–adrenal axis as well, which regulates stress reactivity. For example, IL-1β and IL-6 can induce upregulation of corticotropin-releasing hormone and vice versa, leading to increased activation of both systems (Theoharides, Weinkauf, & Conti, Reference Theoharides, Weinkauf and Conti2004). Cytokines also increase oxidative stress through activation of the tryptophan catabolite pathway (Anderson et al., Reference Anderson, Berk, Dodd, Bechter, Altamura, Dell'Osso and Maes2013). Similarly, complement proteins may interact with other neurodevelopmental factors. Elevations in C3a in pituitary cultures has been shown to increase release of growth hormone, prolactin, and adrenocorticotropic hormone (Francis et al., Reference Francis, Lewis, Akatsu, Monk, Cain, Scanlon and Gasque2003). Oxidative stress is known to activate the complement system as well (Collard et al., Reference Collard, Vä, Morrissey, Agah, Rollins, Reenstra and Stahl2000). This suggests that neurodevelopmental changes observed in response to prenatal immune challenge are the result of interactions between these pathways. More work is needed to elucidate potential interactions between these biological systems.