Pavlovian or classical conditioning is a phenomenon of associative learning, in which a behavioral response is induced by establishing a temporal pairing between a conditioned stimulus (CS) and unconditioned stimulus (US). ¹ In addition to the well-known phenomenon of classic conditioning of physiological reflexes, immune responses can also be memorized and recalled using Pavlovian conditioning, based on a reciprocal communication between the central nervous system (CNS) and the peripheral immune system. ²

This concept of behaviorally conditioned immune modulation was rediscovered by Ader and Cohen ³ in the 1970s, who used cyclophosphamide as an immunosuppressive agent (US) and illness-induced taste aversion by lithium chloride (LiCl) solution as CS and then measured the changes in immune responsiveness by a specific antibody reaction, using sheep erythrocytes as antigen. Since the study was published, this phenomenon of conditioned immunosuppression has been replicated in a multitude of studies ^{4 – 7} by different researchers all over the world and current knowledge about this phenomenon including potential pathways at work and experimental paradigms employed have been comprehensively summarized in a recent review by Hadamitzky et al. ⁸

While most of our knowledge on conditioning of immunological functions is derived from immunosuppression paradigms, relatively few studies have been focusing on immune enhancement. A series of these studies targeted the conditioned enhancement of both CTLs and natural killer (NK) cells in rodents, identifying multiple possible signaling molecules and pathways driving these effects. ^{9 – 12}

Historically, behaviorally conditioned immune functions are suggested to be regulated by bidirectional interaction between CNS and peripheral immune system via the hypothalamic-pituitary-adrenal (HPA) axis, sympathetic nervous system (SNS), and the parasympathetic nervous system (PNS). ¹³ In addition to the HPA axis, the brain limbic system (cortex, hippocampus, amygdala) has also been associated with behaviorally conditioned immune responses. ¹³ In a previous study, ¹¹ both adrenocorticotropic hormone (ACTH) and interferon-alpha (IFN-α) were shown to be involved in the efferent signaling pathways during recall of the conditioned enhancement of NK cell activity. ¹¹ However, the exact biological basis (biochemical, neuroanatomical, genetic) and the underlying molecular mechanisms driving these processes are still unclear.

To shed more light on these pathways, we conducted a pilot study, using the formerly published 3-day conditioning paradigm originally developed by Hsueh et al.. ¹¹ Following that paradigm, animals were first exposed to camphor smell and then injected with polyinosinic:polycytidylic acid (poly I:C), simulating a viral infection and also inducing so-called sickness behavior, an animal model for depression-like behavior in rodents. ^{14 , 15} 48h later, animals in the test group were re-exposed to camphor smell, only, to recall a behaviorally conditioned immune response in rats while animals in a positive control group receive a re-injection of poly I:C and animals in a negative control group receive smell conditioning, only, but no poly I:C re-injection.

We then used a gene-agnostic expression analysis approach to identify first gene candidates along the 4 major tissues – hypothalamus, pituitary gland, adrenal glands, and spleen - of the HPA axis while in parallel looking at cytokine/chemokine protein levels in the plasma of the conditioned animals. Results from both approaches were combined and aligned with the existing literature to provide a list of candidate genes and proteins for future studies.

In the present pilot study, we tried to shed preliminary light on the underlying molecular mechanisms, driving behaviorally conditioned immune enhancement. For that purpose, we employed a conditioning paradigm using camphor smell as conditioned stimulus and an i.p. injection of poly I:C as unconditioned stimulus and hypothesized that the efferent signals for the recall of immune response originate from the hypothalamus. To test our hypothesis, we analyzed differential gene expression in the organs along the HPA axis in test (smell re-exposure and sham injection on D2), positive (poly I:C injection on D0 and D2, only) and negative control (smell re-exposure however sham injection both on day 0 and day +2, respectively) animals along with plasma levels of cytokines/chemokines.

As opposed to our initial hypotheses, it seems there could not only be one but multiple signaling pathways post recall along the HPA axis. Analyzing gene expression patterns in hypothalamus, pituitary, adrenal glands, and spleen, we found evidence for the involvement of WNT/β-catenin pathways, including genes as Otx2, Fzd6, Zic1, and Sox7/8/9/10/11 and 13 in the hypothalamus during the early phases of the recall, corroborating earlier studies and recent studies. ^{24 , 25} These findings are complemented by the elevated expression of secreted immunomodulators (osteopontin, OPN) in the hypothalamus and dopaminergic as well as opioid signaling. In the pituitary of test animals, we observed a strong upregulation of steroid synthesis related genes, while gene expression of interferon stimulated genes (ISGs) as Cxcl11, Cxcl10, Cxcl9, Ccl2, Ccl7, Ccl12, Gbp5, Rsad2, Oasb1a/b, and Ch25h seemed mostly confined to positive control. We also identified a strong and immediate upregulation of thermogenesis and prostaglandin synthase associated genes in the adrenal glands and strong upregulation of immunoregulatory molecules in the spleen. Overall, the molecular signaling post recall seems to travel from hypothalamus to pituitary and then onwards to adrenal and spleen, both via neuronal pathways as blood borne messengers and then reach back to the hypothalamus via feedback loops. The same expression pattern can also be observed in positive control; however, here the reaction seems to occur much faster, especially regarding downregulated genes, which show up 3h earlier with poly I:C re-injection compared to re-exposure to the conditioned stimulus.

Looking at the cascade in more detail, one of the key players in the hypothalamus seems to be Otx2, with its immediate and strong upregulation in the test arm, detected by two independent array probes. The upregulation of frizzled receptor Fzd6 and transcription factor Zic1 in test further point towards an involvement of the WNT/β-catenin pathway. While Wnt1 upregulation in the test arm did not reach the >2fold threshold, we observed an immediate upregulation of Wnt5a and Wnt7b in the hypothalamus of test animals. This observation of Wnt regulation in association with genes involved in dopaminergic signal transmission (Slc18a2) links to a recent study by Zhang & Yang, 2021, ²⁶ which found two other Wnt genes, Wnt1 and Wnt3a play a critical role in the differentiation of neural stem cells into dopaminergic neurons using tetrahydroxy stilbene glycoside and Wnt-signaling specific inhibitor (IWR1).

Downstream of Otx2, we saw an upregulation in neural transcription factors Sox8, 9, 10, 11 and 13, with Sox11 being upregulated immediately post recall in test and Sox10 showing the strongest upregulation at +3h post recall. Of note, Sox9, 10, and 11 have previously been shown to interact with Otx2 during the neural development ²⁷ and regulation of visual cycle gene expression in the retinal pigment epithelium. ²² In addition, we also observed a single point upregulation of the homeobox transcription factor Engrailed 1 (En1) +3h post recall in test and immediately in positive. Engrailed 1 has been shown to interact with Otx2, repressing canonical Wnt-signaling during murine embryonic development of the mid and hindbrain and its expression seems to be an important survival factor both for serotonergic and dopaminergic neurons. ²⁸

Beside Otx2, two further genes were found strongly upregulated in hypothalamus post recall in test: Spp1, coding for immune regulator osteopontin (OPN) and Slc18a2, coding for vesicle transporter VMAT2. OPN, beside its historical role in biomineralization, has been shown to hold a central role in immune activation and regulation e.g. by interacting with leukocyte integrin receptors (α4β1, α9β1, and α9β4) ^{29 , 30} and cell-surface marker (CD44), ³¹ involved in lymphocyte activation, recirculation, and homing. OPN is also a chemoattractant for neutrophils, expressed by wide range of immune cells and has been shown to play a role in immunomodulation by blocking IL-10 in Th2 type cytokine responses and promoting inflammatory IL-12 and Th1 type expansion. ^{32 , 33} Considering its pronounced upregulation during the first 6h post recall in test and its humoral immunomodulatory functions, we postulate a central role for OPN in the recall of conditioned immune enhancement, either directly or working in concert with release of CRH and/or other soluble factors into the median eminence, triggering the release of ACTH and further immunomodulatory molecules from the pituitary gland. While this hypothesis is in line with earlier findings by Hsueh et al., ¹¹ which demonstrated the involvement of plasma ACTH in the recall of the behaviorally conditioned immune enhancement, except a short peak +3h in positive, we did not detect any upregulation of Pomc, precursor of ACTH and multiple other peptide hormones, on mRNA level in any of the tissues. We did, however, detect an immediate, single point upregulation of IFN-γ gene expression in the hypothalamus of test and a similar upregulation of IFN-α in the spleen +6h post recall of positive control animals.

Another - not necessarily mutually exclusive - option of how OPN might be involved in the signaling cascade during recall of the conditioned immune reaction, could be the HPA feedback loop. As shown by Wang et al., ³⁴ OPN -/- mice show reduced levels of ACTH and elevated levels of cortisol during chronic restraint stress and as speculated, same could be due to a negative feedback loop involving OPN. Interestingly, we also detected decreased levels of Cxcl10 and Ccl2, both in hypothalamus of test +6h post recall and immediately post recall in the pituitary of positive control animals. Further to this hypothesis, Trinh et al. in 2020 ³⁵ further demonstrated that stress, induced by systemic poly I:C injection, can increase plasma OPN, triggering the release of ACTH. This is in line with the Wang’s observation that OPN injection results in partial restoration of the ACTH response to stress in OPN -/- mice, while an anti-OPN antibody was shown to partially inhibit the stress response in OPN WT mice. ³⁴

The third strongly upregulated gene in hypothalamus in test post recall was Slc18a2, encoding vesicular monoamine transporter (VMAT2). VMAT2 is responsible for the transport of cholinergic neurotransmitters norepinephrine and dopamine into synaptic vesicles. Interestingly, both Reserpine, which irreversibly inhibits VMAT2, and 6-Hydroxydopamine (6-OHDA), a neurotoxin inhibiting dopamine signaling, have been shown to downregulate Slc18a2 transcription, blocking both the conditioned recall of an enhanced NK reaction ³⁶ and the conditioned recall of the enhanced neutrophil reaction. ¹² As further shown by Hsueh et al., ^{11 , 12 , 37} both cholinergic and serotonergic pathways in the CNS seem to be involved during acquisition as well as recall of the conditioned NK cell response and same has been shown to involve both muscarinic and nicotinic receptors. In line with these data, we saw a downregulation of both dopamine receptors (Drd1, Drd2) and serotonergic receptors (Htr1f, Htr2a) in the hypothalamus of test animals +3h post recall, further pointing towards an involvement of dopaminergic and/or noradrenergic signaling pathways during the recall of the conditioned response. Figure 7 shows the putative interaction between the upregulated genes and potential interactors.

Additional genes strongly downregulated in hypothalamus included pro-melanin-concentrating hormone gene (Pmch), precursor of the orexigenic peptide MCH, vasoactive intestinal peptide (Vip) as well as neuropeptide Y (Npy), together with an upregulation of neuropeptide S (Nps). MCH, in addition to its role in appetite stimulation, has been shown to play a critical role in regulating dopamine signaling by suppressing DA release in the nucleus accumbens while NPY seems to play a similar role by directly inhibiting dopaminergic neurons in the ventral tegmental area (VTA). ³⁸ Neuropeptide S has the exact opposite effect as neuropeptide Y, acting anxiolytic and anorectic, which adds to the pattern of appetite-reduction and energy conversation by immediate and strong downregulation of Pmch and Vip, further augmented by downregulation of Npy and upregulation of Nps, observed immediately in positive and +3h post recall in test.

We also observed a strong downregulation of opioid inhibitory G-protein coupled receptor 88 (Gpr88) and regulator of G-protein signaling 9 (Rsg9) in concert with an upregulation of prepronociceptin (Pnoc) and proprotein convertase subtilisin/kexin type 1 inhibitor (Pck1n), detected by two independent probes; like with Npy and Nps, regulation was observed +3h in test and immediate in positive. Again, these observations are in line with former studies by Hsueh et al., ³⁹ which have demonstrated the involvement of opioid pathways during the recall of behavioral conditioned immune enhancement.

Finally, we also identified RT-1A upregulated in both hypothalamus and adrenal of test animals. RT-1A is a classical rat MHC class Ia gene, involved in the presentation of peptides to cytotoxic T lymphocytes and has been shown to interact with the Ly-49 family of receptor on NK cells as “self-identifier”, thereby providing protection against NK cell lysis. ⁴⁰

Most of the other genes we observed regulated in hypothalamus shortly post recall such as cilia and flagella associated protein Cfap43, involved in olfactory detection, Npas4, key regulator of GABAergic synapse development, stabilizing the activity of neurons during glutamatergic input, or dynein subunits Dnah12 and Dnah1, point towards a heavy involvement of microtubular transport processes during the recall of the conditioned immune reaction.

Pituitary onwards, we observed an upregulation of steroidogenic, nuclear, and metabolism-related genes like Fam111, Star or Rgs1 in test animals post recall as opposed to a strong upregulation of ISGs in positive. Same view, however, might be distorted as we were not able to analyze time points +0h and +3h in pituitary, and hence might have missed upregulation during first 3h.

Independently of this, we did observe an upregulation of orphan nuclear receptors Nr4a2 and Nr4a3 in test - while the exact function of both receptors has not been fully elucidated, recent studies point towards a role in antigen presentation and viral response. ⁴¹ Downregulation did not yield a consistent pattern and beside few, single point downregulations as observed with Tetraspanin (Tspan8) and Claudin (Cldn1), only serotonin-related tryptophan hydroxylase Tph1 showed a clear pattern. This rather mixed expression pattern in test is in strong contrast to the clear and strongly ISG-dominated expression pattern in positive control, showing more than 50x upregulation for genes like Cxcl9, 10, 11 or Gbp5, Rsad2, and Ccl2, which points to a different mechanism when comparing poly I:C re-injection to the recall of the conditioned immune reaction.

Further along the HPA axis, the strong immediate upregulation of Ucp1 (>300fold) in the adrenal glands of test animals point towards a dramatic shift in metabolism towards heat generation as e.g. required for fever induction. Likewise, strongly upregulated Thrsp has also been shown to play a role in thermogenesis of brown adipocytes, ⁴² which was recently confirmed in fed and refed hatchling chicks. This metabolic shift towards heat generation is also in line with the upregulation of fatty acid binding protein Fabp3, involved in the regulation of mitochondrial thermogenesis and temperature homeostasis. At the same time, we saw the upregulation of both adrenergic receptors Adr1a and Adr3b. Both receptors have been shown to be involved in metabolic abnormalities and associated diseases and mutations in these receptors have been speculated to be associated with weight gain in association with anti-psychotic drugs, further corroborating a role of gene expression towards metabolic adaptation to sickness-related conditions.

In the spleen, we saw a strong upregulation of known and partially unknown immunomodulatory genes. The highest upregulation was observed for a 354bp uncharacterized transcript, coding for a 118aa Ig-like domain containing protein (M0R5S4; “Hiramoto factor”). Reactome analysis suggests the involvement of this protein in immunoregulatory interaction between lymphoid and non-lymphoid cells involving lymphoid expressed Fc-gamma receptors and/or in signaling of the B cell receptor. The second upregulated gene, Ly6aI, also known as stem cell antigen 1 (SCA-1), has recently attracted attention after it was identified as key gene in the transport of neurotropic vector adeno-associated virus serotype 9 across the blood-brain barrier in C57BL/6J mice. ⁴³ Its classic function, however, is more known in hematopoiesis; in fact, Sca1 is one of the most common cell surface marker used to enrich adult hematopoietic stem cells. Furthermore, a regulatory role of Ly6 proteins in nicotinic acetylcholine receptors (nAChRs) has been suggested. ⁴⁴

In addition to M0R5S4 and Ly6aI, many other genes upregulated in spleen fall into the category of immunomodulation and -regulation. RT1-Bb is a member of the MHC class II family interacting with CD4 receptors on helper T cells. SIRPδ (Sirpd) is a member of the signal regulatory protein family, involved in signal transduction and cell adhesion and more researched member SIRPα acts as inhibitory receptor by interacting with transmembrane protein CD47, controlling effector functions of the innate immune response. ⁴⁵ Loc690948, predicted to code for Lilrb3a, is an inhibitory member of the leukocyte immunoglobulin-like receptor family and LILRBs have been documented on a broad range of immune cells including NK cells, where they can modulate immune cell functions as cytokine release, antibody production and -presentation ⁴⁶ as well as Toll-like receptor signaling. ⁴⁷ Apol11a, coding for a Apolipoprotein L, a Bcl-2 like protein, has been proposed to have cell death related function due to their pore-forming domain, inflicted by dendritic cells after viral stimulation and also gets strongly expressed splenic DCs post stimulation with poly I:C, dependent on TLR3/TRIF, a reaction mimicked by IFN-ß. ⁴⁸ CD72 is an inhibitory co-receptor on B cells, that recognizes the RNA-containing Sm/RNP (Smith/Ribonucleoproteins as resulting from cell death) by an extracellular C-type lectin domain (CTLD) while inhibiting the corresponding B cell response through it intracellular immunoreceptor tyrosine-based inhibition motif (ITIM), thereby blocking the TLR7-mediated activation of antibody production. ⁴⁹ Finally, RT1-N2 belongs to the family of rat MHC class Ib molecules. This class has recently been shown to interact with both inhibitory and activating Ly49 receptors, resulting in the specific expansion of allospecific NK cells. ⁵⁰ All these findings point towards a central role for spleen in the modulation of the behavioral conditioned immune enhancement.

With reference to Hsueh et al. ¹¹ original study in mice, we also looked at IFN-related gene expression in spleen and while we couldn’t find an upregulation of Ifn-α or -β in spleen of test, we did observe an upregulation of Ifn-α2 gene expression in positive +6h post recall. More importantly, we also saw an immediate peak of Ifn-γ expression in hypothalamus of test, also reflected by elevated IFN-gamma levels in plasma. Likewise in line with Hsueh et al., we did observe elevated ACTH plasma levels in test and positive during the first 6h post recall but not in negative control animals ( Figure 6).

For other chemokines, we saw a rather mixed pattern when comparing plasma levels and gene expression: while we did see a peak of GRO/KC/Cinc-1, MIP-1α, and Rantes +6h in plasma of both test and positive control, a corresponding upregulation of Cxcl1, Ccl3 and Ccl5 on gene level could only be observed in pituitary and adrenal of positive. Likewise, a plasma peak of TNF-α did show a corresponding upregulation of Tnf gene expression. For plasma MCP-1, we observed immediate upregulation of Ccl2 in the spleen of positive and 3h later in spleen of test. This pattern resembles very much the expression of Cxcl10 and corresponding plasma levels of IP-10, where we first saw an upregulation of Cxcl10 in spleen of positive and 3h later in test plus a similar, 3h delayed pattern in hypothalamus. This pattern, an upregulation in spleen, first, and 3h later in hypothalamus, starting in positive, immediately and 3h delayed in test, was also observed for other ISGs as Cxcl11, Gbp5, and Rsad2, and might suggest that the upregulation of ISGs in hypothalamus is rather a secondary effect, triggered by an upregulation of ISGs in spleen.

In this context, it might be interesting to mention that we also observed a short but clear upregulation of acetylcholine receptor Chrna3 in the spleen of test animals immediately post recall. Most of the available literature suggest Chrna7 as the main receptor for acetylcholine ⁵¹ in the spleen, involved in the relay of vagus nerve signals to cytokine-producing macrophages via acetylcholine-synthesizing T cells in the spleen ⁵² and as recently discovered, this Chrna7-based activation of macrophages seems to be mediated by a downregulation of α1,3-Fucosyltransferase Fut7. ⁵³ We could not see any upregulation of Chrna7 in spleen, however we did see a downregulation in Fut7 +3h post recall in both test and positive, hinting towards a potential involvement of the vagal system in modulating the recall of the conditioned immune reaction. And while a recent paper by Verlinden et al. 2019 ⁵⁴ demonstrated catecholaminergic but no direct cholinergic innervation of the spleen, this finding does not rule out the possibility of an indirect vagal innervation via postganglionic non-cholinergic fibers, further supported by a recent study by Zhang et al. ⁵⁵ that has shown that CRH neurons originating from the CeA of the amygdala and PVN of the hypothalamus can stimulate splenic plasma cells formation via direct splenic innervation, potentially via intermediate T cells, translating the noradrenalinergic signal into an acetylcholinergic message.

In summary, our data indicate that in addition to the classic HPA axis involving CRH, ACTH and cortisol release, there could be additional pathways connecting brain and immune system during the recall of behavioral conditioned immune enhancement e.g. by direct splenic innervation, modulating and finetuning the message via cholinergic signaling from the PVN. We further postulate that recall of the conditioned immune response starts in the hypothalamus via a Wnt/β-catenin related pathway, involving Otx2, Sox7/8/9/10/11, Zic1, and Fzd6 as well as monoaminergic (Slc18a2) and opioid messaging (Gpr88). While still to be analyzed in more detail, both osteopontin as strongly upregulated, potentially in concert with Pmch and Vip as strongly downregulated immediately post recall, seem to play an important role during the early phases - either in the communication between hypothalamus and pituitary and/or as blood borne immune modulators themselves. While pituitary onwards, it is hard to identify one specific pathway, we observe the upregulation of steroidogenic and thermogenic genes in pituitary and adrenal, respectively, pointing towards an early hormonal activation in parallel to a strong immunomodulation observed in spleen, from where the feedback loop involving strongly upregulated ISGs as Cxcl9/10/11 reaches back to the brain.

Underlying data deposited in Zenodo.org: https://doi.org/10.5281/zenodo.7086375. ⁵⁶ •

20220902 Affymetrix all tissues all timepoints red green incl p v2.xlsb (Affymetrix raw data and analysis across all tissues and time points)