TB-IRIS: Proteomic analysis of in vitro PBMC responses to Mycobacterium tuberculosis and response modulation by dexamethasone

Paradoxical tuberculosis-associated immune reconstitution inflammatory syndrome (TB-IRIS) occurs in 8–54% of South African patients undergoing treatment for tuberculosis/human immunodeficiency virus co-infection. Improved TB-IRIS molecular pathogenesis understanding would enhance risk stratification, diagnosis, prognostication, and treatment. We assessed how TB-IRIS status and dexamethasone influence leukocyte proteomic responses to Mycobacterium tuberculosis (Mtb). Patient blood was obtained three weeks post-anti-retroviral therapy initiation. Isolated mononuclear cells were stimulated ex vivo with heat-killed Mtb in the presence/absence of dexamethasone. Mass spectrometry-based proteomic comparison of TB-IRIS and non-IRIS patient-derived cells facilitated generation of hypotheses regarding pathogenesis. Few represented TB-IRIS-group immune-related pathways achieved significant activation, with relative under-utilisation of “inter-cellular interaction” and “Fcγ receptor-mediated phagocytosis” (but a tendency towards apoptosis-related) pathways. Dexamethasone facilitated significant activation of innate-related pathways. Differentially-expressed non-IRIS-group proteins suggest focused and co-ordinated immunological pathways, regardless of dexamethasone status. Findings suggest a relative deficit in TB-IRIS-group responses to and clearance of Mtb antigens, ameliorated by dexamethasone.


Introduction
Mycobacterium tuberculosis (Mtb) is the most prevalent human immunodeficiency virus (HIV)-associated opportunistic infection (OI), with up to 70% of new TB cases in South Africa (SA) being HIV co-infected (Leone et al., 2010) and with Sub-Saharan Africa accounting for~80% of the global burden of HIV-associated TB (Lawn and Churchyard, 2009). Although antiretroviral therapy (ART) reduces the risks of active TB and HIV-related mortality (Harries et al., 2009;Meintjes and Lynen, 2008), combined treatment is associated with various complications, including tuberculosis-associated immune reconstitution inflammatory syndrome (TB-IRIS) (Leone et al., 2010;Harries et al., 2009;Schutz et al., 2010).
Immune restoration, perhaps coupled with relatively lacking immunoregulation, may lead to IRIS-characteristic tissue-damage (Lawn et al., 2005). Based on murine models, Barber et al. (2012) proposed that following lymphopaenic uncoupling of innate and adaptive responses, the timing of macrophage infection with Mtb -relative to CD4+ T-cell-mediated activation of the infected macrophage -is a critical factor in IRIS development: hyper-activation of accumulated primed macrophages may lead to inflammatory tissue destruction.
There is currently no validated immunological marker of IRIS to guide clinical management: instead, identification is largely by clinical diagnosis of exclusion (Meintjes et al., 2008a). A double-blinded, randomized, placebo-controlled study  demonstrated that prednisone reduced paradoxical IRIS morbidity, concurrent with reductions in serum IFNγ, TNFα, IL-6, IL-12, and IL-10, suggesting an effect mediated via reduced cytokine effector responses (Maertzdorf et al., 2011;Skolimowska et al., 2012). At clinical doses, glucocorticoids are known to have immunosuppressive and anti-inflammatory effects (Zen et al., 2011). However, while prednisone ameliorates IRIS symptoms, it may not resolve underlying pathogenic mechanisms. Better understanding of the molecular basis of IRIS would therefore inform improved risk-stratification, diagnosis, prognostication, and definitive prevention and/or treatment strategies. Clinical and immunological data suggest a difference in in vivo inflammatory responses in persons predisposed to IRIS. Here, we have assessed whether IRIS status influences immune cell proteomic response to Mtb, as well as how dexamethasone modulates these responses, in order to provide insight into the molecular pathogenesis.

Study design
This was a pre-clinical mass spectrometry-based differential proteomic profiling study to assess whether IRIS status influences immune cell proteomic response to Mtb, and how corticosteroids modulate this response. PBMC from IRIS (n = 30) and non-IRIS (n = 30) participants were cultured ex vivo, mimicking in vivo conditions of IRIS development and symptom alleviation (stimulation with Mtb ± dexamethasone, in the presence of first-line ART). Global protein expression differences were interrogated to generate hypotheses regarding the molecular mechanisms underlying IRIS and its symptom alleviation by corticosteroids. The sample size (N = 60) provides N95% power to detect a twofold change in expression (p b 0.01).

Ethics, patient recruitment, and sample acquisition
This study was nested within a parent study entitled "Prospective cohort study of hospitalised HIV-TB co-infected patients at Brooklyn Chest Hospital" (van der Plas et al., 2013), approved by the University of Cape Town Human Research Ethics Committee (HREC 049/2009). All participants provided written informed consent.
As previously described, hospitalised ART naïve HIV-TB patients meeting eligibility criteria were recruited and followed for development of paradoxical TB-IRIS using a consensus clinical case definition (Meintjes et al., 2008a). Venous blood samples were obtained at three weeks post-ART initiation, and subsequently classified into two groups: 'IRIS' or 'non-IRIS' (Fig. 1a). Samples were transported at room temperature to a BSL2 facility at the Institute of Infectious Diseases and Molecular Medicine within 4 h. Exhaustive cell counts were performed at every step of downstream sample processing.
PBMC culture conditions (Fig. 1b) mimicked in vivo conditions under which IRIS develops. To each well, the following were added: first-line ART cocktail to final physiologically relevant concentrations of 9 μM lamivudine (3TC; NIH, Germantown, USA), 115 nM efavirenz (EFV, NIH), and 5 μM zidovudine (AZT, NIH); heat-killed Mtb H37Rv to a final multiplicity of infection (MOI) of 5:1; and, where appropriate, dexamethasone to a final physiologically relevant concentration of 15 ng/mL. PBMC from each participant were cultured (37°C, 5% CO 2 ) independently for 20 h (the time of maximal differential gene expression, as determined by qPCR; data not shown) under the following three conditions: no stimulus (baseline), plus Mtb only, and plus Mtb and dexamethasone.
Cells from each participant were harvested, washed twice in phosphate-buffered saline (PBS, Sigma-Aldrich), and pooled only briefly (to avoid proteomic alteration by allogeneic reactions) according to IRIS status (n = 30 IRIS, n = 30 non-IRIS), before being immediately flash-frozen in liquid nitrogen prior to storage at −80°C.

Protein extraction and fractionation
3 × 10 6 frozen PBMC per group were lysed by immediate re-suspension in 3 mL phosphoprotein lysis buffer (0.25% CHAPS, 250 U benzonase, 25 mM MES pH 6.0, 1 M NaCl) at 4°C for 30 min. Insoluble debris was pelleted by centrifugation (10,000 g, 30 min, 4°C); the supernatant and cell debris were thereafter processed separately for LC-MS/MS analysis as follows.
Supernatant protein was quantitated by Bradford assay, and a phospho-purification kit (Qiagen) was employed, as per kit instructions. Briefly, supernatant was loaded onto prepared columns to capture phosphoproteins; the flow-through thus contained phospho-depleted proteins. Column-bound proteins were washed (6 mL phosphoprotein lysis buffer) and then eluted using 2 mL elution buffer (50 mM K 2 PO 4 pH 7.5, 50 mM NaCl). Phospho-enriched and -depleted proteins were separately subjected to filter-assisted sample preparation (FASP) (Wisniewski et al., 2009). Briefly, samples were reduced, alkylated, and proteolysed (protein:trypsin ratio of 100:1) on a spin-filter membrane, facilitated by exchange of detergent-containing buffer for ureacontaining buffer. Phospho-depleted peptides were further processed by high-pH reverse phase fractionation into five fractions. All peptide fractions were dried and re-suspended in 5% acetonitrile [ACN; Sigma-Aldrich], 0.1% formic acid [FA] for LC-MS/MS (Fig. 1b).
Data acquisition, controlled by the Xcalibur software package (Thermo Scientific), was performed in 'Top-20', data-dependent, positive ion mode. Precursor ion settings included resolution at 60,000, automatic gain control (AGC) target of 1 × 10 6 , and scan rage 400-2000 m/z. Selected precursor ions were fragmented in the linear ion trap via collision-induced dissociation. Dynamic exclusion was set at 60 s to minimise repeated sequencing of the same precursors. A lock mass of 445.120025 m/z was used.

Bioinformatic analysis
The .RAW data files were converted to mzXML format using the Trans-Proteomic Pipeline (Deutsch et al., 2010). Crux-Barista (version 1.37 (Park et al., 2008;Spivak et al., 2009)) software mapped MS/MS fragmentation spectra to a UniProtKB Human Reference Proteome FASTA file (Consortium, 2014) to identify peptide-spectrum matches (PSMs). Search parameters included one missed cleavage and carbamidomethylation of cysteine as a fixed modification. Decoy searches performed in Crux-Barista facilitated calculation of q-values for identified peptides; quantitation was by spectral counting using the normalised spectral abundance factor (NSAF) method (Fig. 1e). In MS and MS/MS modes respectively, mass ranges were set to 400-3500 Da with a mass tolerance 10 ppm, and 200-3500 Da, tolerance 0.5 Da. The q-value threshold was set at 0.05 and protein identification required at least one high-confidence unique peptide. Where different proteins could not be distinguished based on peptide evidence (e.g. splice variants), these were gathered into a 'protein group' for further analysis. Computation was performed using facilities provided by the University of Cape Town ICTS High Performance Computing team (http://hpc.uct.ac.za).
NSAF values of non-redundant protein groups were used to normalise each test condition to baseline within a group. The resulting values were compared both between conditions and between groups by generating per-protein expression ratios and thus fold-changes. Withingroup normalisation of NSAF values facilitates analysis of test condition influence on cell responses relative to baseline. Significant differential expression was defined as fold-changes greater than two standard deviations from the mean fold change for all protein groups. Fold-changes were log 2 transformed for subsequent analysis using the Ingenuity Pathway Analysis (IPA) suite (Ingenuity® Systems) to identify biological canonical pathways with significantly over-represented members in the datasets.
To determine the differential proteomic response to Mtb antigens, Mtb-stimulated IRIS and non-IRIS PBMC datasets were compared, at the levels of fold-change from baseline, pathway enrichment (-log(pvalues)), and pathway activation (z-score).
To determine the manner in which dexamethasone modulates this response, datasets for Mtb-stimulated IRIS PBMC in the absence and presence of dexamethasone were compared, at the levels of fold-change from baseline and pathway activation. To improve confidence in in vivo applicability of findings, ex vivo differences between unstimulated IRIS and non-IRIS PBMC were also independently compared (using alternate software approaches) at the level of pathway activation.

Results
Differential proteomic profiling revealed significant differences in proteomic responses to Mtb antigens in two different experimental comparisons: (1) IRIS vs. non-IRIS PBMC, and (2) IRIS PBMC in the presence/absence of dexamethasone. This facilitated generation of hypotheses regarding the mechanisms underlying IRIS pathogenesis and corticosteroid-based symptom alleviation.

Protein identification, quantitation, and comparisons
A total of 701 and 911 non-redundant PBMC protein groups were identified from IRIS and non-IRIS samples, respectively (Table 2). Proteins that could be identified and quantified in test conditions and at baseline (Table 3a) were used to generate expression ratio values for input into IPA. Significantly differentially expressed proteins are presented as ranked lists with fold-changes from baseline (Table 3b and c).
Key functional pathways were significantly differentially enriched across datasets (colour-coded in Fig. 2a). Activation scores of more focused biological functions (Fig. 2b-d) indicate modulation of the type of response that PBMC are capable of.

Characteristics of non-IRIS PBMC proteomic responses to Mtb antigens
At the level of fold-change from baseline, members of the top 10 differentially regulated proteins in the non-IRIS group play prominent roles in key canonical pathways, such as Rac, RhoGDI, and integrin signalling, as well as F c γ receptor (F c γR)-mediated phagocytosis.
At the level of pathway enrichment, non-IRIS PBMC responses encompass more pathways that are significantly enriched: when -log(pvalues) are plotted to identify significantly differentially enriched pathways between the IRIS and non-IRIS datasets, consistently higher levels of significant enrichment are seen in the non-IRIS group (Fig. 3, left).
At the level of pathway activation, fewer pathways are active during non-IRIS PBMC responses to Mtb antigens, but many achieve significant activation scores (Fig. 2d). Many of the significantly-activated pathways are associated with both innate and adaptive immune functions.
Non-IRIS group proteins contributing to enrichment, activation, and efficacy of the F c γR-mediated phagocytic pathway are also relevant to other actin cytoskeletal dynamics-related pathways, and some even act as regulators in several different pathways (Fig. 4): enriched pathways are tightly functionally integrated (linked via overlapping regulatory mechanisms or roles in actin cytoskeletal dynamics).

Characteristics of IRIS PBMC proteomic responses to Mtb antigens
At the level of fold-change from baseline, TXNDC5 is down-regulated in IRIS relative to non-IRIS PBMC. At the level of pathway enrichment, only four pathways are identified as significantly enriched in the IRIS group (Fig. 3, right), and the canonical pathway for antigen presentation is significantly enriched in the non-IRIS group. Although the eIF2 canonical pathway is relatively more enriched in the non-IRIS group, at the level of pathway activation, it is relatively more up-regulated in the IRIS group. Activation of one critical pathway (Interaction of Blood Cells) is even relatively inhibited in the IRIS group (Fig. 2b). Many more immune and inflammation-related pathways are activated in the IRIS group than the non-IRIS group. Additionally, higher activation scores are observed for the FcγR-mediated phagocytosis pathway in non-IRIS (relative to IRIS) PBMC.

Modulation of these responses by dexamethasone
At the level of fold-change from baseline during IRIS PBMC responses to Mtb antigens, dexamethasone differentially regulates  proteins associated with pathways more closely resembling those identified as enriched during non-IRIS PBMC responses to Mtb antigens. At the level of pathway activation, dexamethasone also increases the number of activated pathways in the IRIS group that achieve significance, and many of these pathways are associated with innate immune responses (Fig. 2c). Even when dexamethasone is present during Mtb stimulation of IRIS PBMC, Mtb-stimulated non-IRIS PBMC still display relatively higher activation scores for the FcγR-mediated phagocytosis pathway (Fig. 5). The pattern of relatively increased FcγR-mediated phagocytosis pathway activation in the non-IRIS (relative to the IRIS) group persists when comparing ex vivo proteomic differences at baseline (z = 3.6).

Discussion
Here, a label-free proteomic discovery strategy identified and quantified 701 IRIS and 911 non-IRIS PBMC unique protein groups. In principle, a discovery proteomic study of paradoxical TB-IRIS has the capacity to reveal pathogenic mechanisms not accessible to immunological studies that assess a limited number of known secreted or surface proteins. Biological pathway analyses based on our unbiased data revealed clear differences in the response to Mtb antigens between IRIS and non-IRIS groups, as well as significant modulation of IRIS PBMC responses by dexamethasone. These differences may be at least partly responsible for the IRIS phenotype and corticosteroid-mediated symptom alleviation, leading to hypotheses regarding biological mechanisms underlying these phenomena.
Patient-derived PBMC were used as a patient-specific model of IRIS pathogenesis, representing the immune cell types likely involved in this process, while dexamethasone was used as a stable analogue of prednisolone -itself the pro-drug of prednisone.
An important caveat to our approach is that, since blood used in this study was taken three weeks post-ART initiation, conclusions can only be drawn regarding IRIS pathogenesis, not predisposition. A further caveat is that limiting clinical sample quantities necessitated use of a sample pooling strategy. Pooling is controversial because it precludes identification of outliers contributed by any single sample; however, levels of an individual protein in a pool by definition represent the average expression of those proteins in the samples making up the pool, and proteins detectable in individual samples are usually also detectable in the pool (Diz et al., 2009). Pooling additionally reduces the major contributor to variance (biological variance), thereby increasing power: the averaging effect of pooling makes the most evident differences and similarities between groups easier to detect (Diz et al., 2009;Oberg and Vitek, 2009). Exhaustive cell counts at every step of processing aimed to minimise technical bias in contribution from any one sample to the pool.
An additional caveat in interpretation of our results is that -for technical reasons -for the IRIS sample insoluble protein fractions, the pool consisted of six samples rather than 30, which may have led to relatively fewer proteins being identified in the IRIS group. However, despite this, the total number of non-redundant proteins identified here represents 25% of the currently observable PBMC proteome (Končarević et al., 2014) and likely represents the most abundant cellular proteins (most of the dry protein mass). This extent of proteome coverage proved sufficient to observe multiple proteins from key differentially-regulated biological pathways, enabling canonical pathway enrichment analysis as a means to generate biologically sound mechanistic hypotheses for validation.
Our findings regarding non-IRIS PBMC proteomic responses to Mtb antigens suggest improved consensus and co-ordination of response pathways, executed by well-organised cells. The majority of enriched a Relatively low statistical significance, but high fold-change. a Relatively low statistical significance, but high fold-change.  pathways are tightly functionally integrated, being linked via overlapping regulatory mechanisms or roles in actin cytoskeletal dynamics. Additionally, most are pre-requisites for effective protective immunity: since actin cytoskeletal dynamics are a key determinant of PBMC function (including signal transduction, motility, phagocytosis and antigen clearance, antigen presentation, cellular interaction, cellular homeostasis, and apoptosis, among others), appropriate functioning of this system is by extension crucial to combating infection and resolving inflammation. Additionally, many significantly-activated pathways are associated with both innate and adaptive immune functions.
Motility is primarily regulated and integrated by the Rho family of small GTPases (e.g. RhoA, Rac, Cdc42), involved in key canonical pathways significantly over-represented in the non-IRIS group. This suggests non-IRIS PBMC have a preserved capacity to co-ordinate and regulate motility for access to and egress from sites of infection, as well as phagocytosis. Indeed, TNFαup-regulated in IRIS patient bloodis known to induce monocyte/macrophage hypercytokinesis (increased but undirected motility) (Jones, 2000), which may impair antigen clearance.
Actin reorganisation is also required for phagocytosis: reorganisation is triggered by receptor ligation and G protein (e.g. Rac, Cdc42) signalling, including activation of the ARP 2/3 complex (Fig. 4), while internalization is controlled by a molecular complex containing WASP. It is notable that the canonical pathway for antigen presentation -required for adaptive response activation -is significantly enriched in the non-IRIS group, while no adaptive-associated pathway activation is observed in the IRIS group.
Taken together, our data suggest that non-IRIS PBMC display preservation of functionalities required for protective immunity, including those related to appropriate cytoskeletal reorganisation, and both arms of the immune response.
By contrast, our findings regarding IRIS PBMC proteomic responses to Mtb antigens suggest the novel idea of poor consensus and co-ordination of response pathways (i.e. few activated pathways achieve significance), possibly due to relatively deficient intercellular interaction (relative inhibition of the Interaction of Blood Cells pathway), or preexisting cellular fatigue (due to e.g. over-activation or a relative deficiency of trophic factors experienced in vivo).
In support of the notion of fatigue, TXNDC5 (down-regulated in the IRIS group), an endoplasmic reticulum lumen protein, is associated with suppression of apoptosis in endothelia (Bruneel et al., 2005) and other human cells. Additionally, the eIF2 canonical pathway -involved in protein synthesis and cellular response to stressis relatively more activated in the IRIS group, favouring apoptosis. These findings suggest that IRIS PBMC (e.g. monocytes) may be more prone to in situ apoptosis, instead of efferocytosis and antigen clearance.
While pathways enriched and activated in the IRIS group are apparently not tightly integrated, they are at least functionally linked, as expected in a setting of inflammation. Concordant with literature, the overall inflammatory response magnitude appears higher in the IRIS dataset, with many more immune and inflammation-related pathways activated.
Interestingly, the apparent lack of significant involvement of adaptive response-associated pathways in the IRIS group may suggest deficient adaptive sustenance of the innate activation required for successful destruction of Mtb by phagocytes. While HIV infection leads to T-cell depletion and functional exhaustion, in vivo development of TB-IRIS is associated with antigen-specific T-cell expansion (Bourgarit et al., 2009;Meintjes et al., 2008b), which is seemingly at odds with our proteomic data. We cannot yet explain this apparent contradiction, but note that inappropriate T-cell activity or reduced innate cell responsiveness to adaptive signals remain possibilities. Taken together, however, our data suggest that IRIS-group PBMC responses, though vigorous, may be too uncoordinated to effectively clear Mtb antigens, leading to inappropriate inflammation.
Our data also show that dexamethasone markedly alters IRIS-group PBMC responses to Mtb antigens in vitro: instead of multiple IRIS-group activated pathways with few achieving statistical significance, in the presence of dexamethasone the number of significantly activated pathways associated with innate immune responses is increased. This suggests that dexamethasone partially ameliorates IRIS-group PBMC dysfunction by restoring a more unified response (although centred on innate rather than both innate and adaptive response-associated pathways). This is consistent with the idea of deficient innate activation by adaptive signals and is also consistent with the observation that 20% of corticosteroid-treated TB-IRIS patients relapse after prednisone withdrawal (Meintjes and Sonderup, 2011). Molecular mechanisms by which dexamethasone achieves these effects are not immediately apparent from our data, but steroids generally suppress pro-inflammatory T-cell activity, while inducing regulatory T-cell and tolerogenic dendritic cell activity (Zen et al., 2011). At low, physiological-range levels, steroids can also enhance macrophage phagocytic and efferocytotic capability (McColl et al., 2007), providing a further potential route for dexamethasone to modulate the IRIS-group PBMC response to Mtb antigens. Taken together, our data therefore suggest that dexamethasone partially rescues the uncoordinated IRIS-group phenotype, either by dampening adaptive pathways to unmask focused innate pathways, or by directly boosting innate pathways, perhaps promoting local antigen clearance.
Notably, our proteomic data point to key differences in utilisation of the F c γR-mediated phagocytosis pathway between IRIS-and non-IRISgroup PBMC. F c γR are expressed on phagocytes, and ligation by IgGopsonised particles or immune complexes (IC) leads to signalling via immunoreceptor-based tyrosine activation motifs (ITAMs). Wholeblood gene expression studies have identified F c γRIA (a high-affinity IgG receptor mainly expressed on monocytes/macrophages) and F c γRI signalling as classifiers of active TB, indicating their likely importance in normal TB pathogenesis (Sutherland et al., 2014;Maertzdorf et al., 2011;Berry et al., 2010). This agrees with findings of significantly reduced F c γRIA expression following successful TB treatment (Sutherland et al., 2014;Cliff et al., 2013). Non-opsonic or surfactantmediated phagocytosis is thought to be important early in Mtb infection (Schafer et al., 2009), and complement-mediated opsonisation is critical (when C3 is absent, phagocytosis of Mtb by monocytes/macrophages declines by 70-80% ( van Crevel et al., 2002)). However, complementmediated uptake induces a relatively non-robust Th response, while F c γR-mediated uptake of Mtb strongly "activates the antimicrobial, degradative and … antigen-presentation activities of macrophages and dendritic cells to enhance the Th1 response" (Igietseme et al., 2004). Maertzdorf et al. (2011) also found exacerbated expression of complement in tuberculosis, which is consistent with the findings of Tran et al. (2013) that inappropriate control of complement activation may contribute to TB-IRIS.
The higher F c γR-mediated phagocytic pathway activation observed here in the non-IRIS group -including in our 'baseline-only' comparison of IRIS-and non-IRIS-group PBMC -therefore suggests a novel, testable hypothesis that a relative deficit in F c γR-mediated phagocytosis occurs in TB-IRIS patient PBMC in vivo, and that this directly affects innate cell-mediated antigen clearance capacity, thereby contributing to higher Mtb antigen loads and aberrant immune responses during TB-IRIS. Whether the deficient F c γR-mediated phagocytosis response observed here in IRIS-group PBMC is a manifestation of simple fatigue, B-cell dysfunction, or an as yet unrecognised factor remains to be determined.
In summary, our data suggest that TB-IRIS patients may take up, present, and clear Mtb antigens less effectively, resulting in an increased antigen load that contributes to the TB-IRIS phenotype through inappropriate or prolonged stimulation of inflammatory cascades after ART initiation. In particular, our data suggest that IRIS-group PBMC respond to Mtb antigens in a dysregulated manner in vitro, with possible underlying mechanisms of dysfunction including: cellular fatigue (including in situ apoptosis), unfocused/uncoordinated cellular responses (possibly related to poor cellular interaction), dysregulated motility, deficient FcγR-mediated phagocytosis, and lack of activating adaptive signals (or lack of innate responses to these signals). Dexamethasone ameliorates this IRIS-group PBMC dysregulation in vitro by modulating functions to resemble those of non-IRIS-group PBMC (e.g. by unmasking or boosting innate response pathways), suggesting that alternative therapeutic strategies for TB-IRIS may yet prove more effective than the currently used prednisone. By contrast, our data suggest that non-IRIS-group PBMC display proteomic characteristics consistent with known immune responses to TB infection, including preservation of functionality required for protective immunity, including those related to appropriate cytoskeletal reorganisation and both innate and adaptive immune responses. Overall, our proteomic findings are thus generally consistent with the literature, including the model of Barber et al. (2012), but provide new insight into plausible underlying molecular mechanisms. Experiments to validate these hypotheses are currently underway.

Conflict of interest statement
The authors declare that there are no conflicts of interest. The funders had no role in the study design, data collection, data analysis, data interpretation, or writing of this report. The opinions, findings, and conclusions expressed in this manuscript reflect those of the authors alone.