Early over expression of messenger RNA for multiple genes, including insulin, in the Pancreatic Lymph Nodes of NOD mice is associated with Islet Autoimmunity.

BACKGROUND: Autoimmune diabetes (T1D) onset is preceded by a long inflammatory process directed against the insulin-secreting beta cells of the pancreas. Deciphering the early autoimmune mechanisms represents a challenge due to the absence of clinical signs at early disease stages. The aim of this study was to identify genes implicated in the early steps of the autoimmune process, prior to inflammation, in T1D. We have previously established that insulin autoantibodies (E-IAA) predict early diabetes onset delineating an early phenotypic check point (window 1) in disease pathogenesis. We used this sub-phenotype and applied differential gene expression analysis in the pancreatic lymph nodes (PLN) of 5 weeks old Non Obese Diabetic (NOD) mice differing solely upon the presence or absence of E-IAA. Analysis of gene expression profiles has the potential to provide a global understanding of the disease and to generate novel hypothesis concerning the initiation of the autoimmune process.
METHODS: Animals have been screened weekly for the presence of E-IAA between 3 and 5 weeks of age. E-IAA positive or negative NOD mice at least twice were selected and RNAs isolated from the PLN were used for microarray analysis. Comparison of transcriptional profiles between positive and negative animals and functional annotations of the resulting differentially expressed genes, using software together with manual literature data mining, have been performed.
RESULTS: The expression of 165 genes was modulated between E-IAA positive and negative PLN. In particular, genes coding for insulin and for proteins known to be implicated in tissue remodelling and Th1 immunity have been found to be highly differentially expressed. Forty one genes showed over 5 fold differences between the two sets of samples and 30 code for extracellular proteins. This class of proteins represents potential diagnostic markers and drug targets for T1D.
CONCLUSION: Our data strongly suggest that the immune related mechanisms taking place at this early age in the PLN, correlate with homeostatic changes influencing tissue integrity of the adjacent pancreatic tissue. Functional analysis of the identified genes suggested that similar mechanisms might be operating during pre-inflammatory processes deployed in tissues i) hosting parasitic microorganisms and ii) experiencing unrestricted invasion by tumour cells.

Background

Type 1 diabetes (T1D) is an autoimmune disease characterized by the absence of insulin due to the specific destruction of the insulin-producing β cells of the pancreatic islets. This is a progressive process taking over 20 weeks in the NOD mouse and several years in human patients to be completed [1]. The NOD animal model has been a valuable source of information for several aspects of disease pathogenesis [2]. Genetic studies have contributed to portray the complexity of the disease and have established that multiple loci are carrying genes implicated in T1D in human [3] and animal models [4]. In human, more than 6 genes contribute to the disease [5,6] while over 20 loci have been described in the NOD mouse [7] but only few possible candidate genes have been unequivocally defined [8] other than the H2g7 Idd1 locus [7,9].

Despite intensive research, the initial causal events remain elusive since the selection of individual mice at early stages, prior to the overt clinical signs, represents a challenge. Indeed, even though the NOD mice are inbred not all animals develop the disease, with an incidence of 40-90% in females, depending on the colony [2]. Additional hindrances for selecting individual animals that will subsequently develop the disease with certainty are the low penetrance of the implicated genes and the influence of environmental factors. For these reasons, the exact mechanisms taking place prior to the onset of the pancreatic islet-damaging sustained inflammatory processes remain largely unknown.

The aim of our investigation was to evaluate the possibility of the existence of distinct gene expression profiles in order to eventually render possible the study of the early molecular changes, taking place before the onset of inflammation in autoimmune prone mice. One hypothesis is that during the post-weaning period in genetically autoimmune prone individuals (mouse or human), homeostatic changes prompt the immune system not to conform to physiological responses, but to instead trigger pathways that lead to the final autoimmune condition in later life. Exogenous factors including the more or less sustained presence of micro-organisms in otherwise microbe-free tissues may take place during early life stages [10]. This may influence the immune regulatory processes put in place to prevent, attenuate and/or repair the acute or sustained inflammation. It has been reported that early immune stimulation prevents autoimmune diseases, while infections later in life might exacerbate their advent in genetically prone individuals [11]. A single injection of BCG (the Mycobacterium bovis vaccine strain) is highly protective against T1D in the NOD mice [12] and BB rats [13] when administered early but it has no beneficial effect given after disease onset; similar to what has been observed in human [14]. These observations lead to the indication that the time "window" of opportunity to prevent disease in human may share similarities with the NOD mice and thus BCG or other immune-system signalling, safe biological preparation, should be given as therapy before the onset of inflammation of the pancreatic islets (insulitis). This early "window" coincides in the NOD mouse model with the early appearance of insulitis, occurring between 4 and 7 weeks [2] and led to the hypothesis that early initiating mechanisms should be evaluated at this age i.e. over the weaning period. We have previously established that IAA are present early between 3 and 5 weeks of age and represent a diabetes sub-phenotype since their presence in NOD mouse sera, at this early period, correlated with the final disease phenotype [15]. Thus Early-IAA (E-IAA) have been shown to mark the first measurable phenotypic checkpoint in T1D pathogenesis. Animals positive for E-IAA at 3-5 weeks after weaning, develop autoimmune diabetes earlier (between 16 and 20 weeks) than their E-IAA negative littermates [15]. We have used this E-IAA sub-phenotype to select individual mice as diabetes prone and search for changes in gene expression patterns in the PLN, correlating with the presence of E-IAA in the NOD mouse sera.

Overt T1D is considered to require the dominant presence of islet-destructive pro-inflammatory T lymphocytes [16]. The onset of T cell activation is known to deploy within organized secondary lymphoid tissues, draining the tissue that delivers the adequate signals for activation of T and B lymphocytes.

In the NOD mice, any early islet-disruptive remodelling process taking place, whether it implies β cell apoptosis or repair mechanisms [17], might result in delivery of immune cell-signalling to the PLN. The endpoint of such signalling might include deletion of autoreactive regulatory T cells, and/or expansion of autoreactive pro-inflammatory T lymphocytes as well as T cell-dependent autoreactive B lymphocytes. The later further differentiate as insulin auto-antibody secreting plasma cells [18]. Antigen specific T cells interact with dendritic cells (DC) within intact explanted lymph nodes (LN), and it has been observed that immunological synapse formation and prolonged interactions between these 2 cell types is followed by activation, dissociation and rapid migration of T cells away from the antigenic stimulus area [19]. Even more strikingly, excision of PLNs at 3 weeks of age protects NOD mice against IAA, insulitis and diabetes development, almost completely, but has no effect when performed at 10 weeks of age [20]. Therefore the time frame of 3 to 5 weeks seems to correspond to the initiation of the autoimmune process in the PLN and the related transcriptional profiling can be potentially sort out in this tissue.

We anticipated that at 5 weeks of age, differential gene expression correlated with the presence or absence of E-IAA, in the PLN of NOD mice, could be detected. This timing allowed for prior tests of E-IAA at 3 and 4 weeks. We used microarray analysis to determine gene expression patterns in the PLN. This approach has the potential i) to provide a global profiling of the modulated genes at this first measurable phenotypic checkpoint in T1D pathogenesis, ii) to allow novel hypothesis to be formulated concerning the initiation of the islet-destructive autoimmune process and eventually to further challenge these hypothesis. We identified ectopic high expression of insulin genes in the E-IAA positive PLN. This data confirm the possible cross talk between the pancreas and its adjacent lymph nodes via at least one pancreas specific autoantigen.

Functional annotations of up-regulated transcripts in the PLN of E-IAA positive NOD mice revealed a gene network otherwise found also to operate during inflammatory and tumour processes. Furthermore, we have identified several transcripts coding for extracellular proteins that have to be evaluated as potential disease pathogenesis associated markers and possibly therapeutic targets. This data represent a "freeze frame" of a first identifiable disease check point, associated with the presence of IAA and corresponding to the early steps of the autoimmune process.

Methods

Animals

Mice were purchased from Taconic farms (NOD/tac). Pregnant females were tested one week before delivery for the presence of IAA. To discard any possibility of trans-placental antibodies affecting our experimental design, based upon the spontaneous appearance of IAA, we have used, when possible, individual mice from progenies of negative females and from litters that they contained both positive and negative individuals.

Animals were tested as previously described [15], at 3, 4 and 5 weeks of age for the presence of IAA. PLN of IAA positive and negative animals were used for RNA preparation and pancreata have been also isolated for histology. All animals were kept under SPF conditions and all studies were performed under the recommended Laboratory Animal Care committee of the University of Colorado.

IAA assay

IAA were measured in a standard radioimmunoassay as previously described [15,21] incorporating competition with unlabeled insulin and precipitation with Protein A/G sepharose in a 96 well filtration plate.

Histology immunoassay

Tissues were formalin (10%) fixed and embedded in paraffin. Sections of paraffin-embedded tissues were stained with Hematoxylin & Eosin and with polyclonal guinea pig anti-insulin antibodies (Millipore) followed by incubation with a peroxidase-labeled anti-guinea pig IgG antibody (Kierkegaard & Perry Laboratories Inc.). The presence of insulitis has been evaluated by visualizing more than 10 islets in 3 different sections from the same individual. Similarly the presence of insulin expressed in the PLNs has been evaluated in more than 3 sections of each individual tissue.

Tissues and RNA preparation

PLN have been dissected from six E-IAA negative and three E-IAA positive animals at 5 weeks of age. Tissues were immediately placed in RNA later containing RNases inhibitors (Qiagen) and processed for total RNA preparation on the same day with Qiagen mini-RNA commercial kits (RNeasy) with on-column DNase treatment following the manufacturer's protocol. Quantity, purity and integrity were confirmed initially by spectrophotometry (A260/A280 ratio) and capillary electrophoresis (2100 Bio analyzer; Agilent Technologies, Palo Alto, CA). 4.5 μg of total RNA were used for the target preparations and no additional step of cDNA amplification has been used in our studies.

Microarray Analysis

Target preparations and chip hybridizations have been performed as recommended by the manufacturer MG_U74A_version 2 GeneChips (Affymetrix, Santa Clara, Ca). Each chip contained 12 486 probe sets (Supplemental Materials and Methods). All gene expression data discussed in this report have been deposited in NCBI's Gene Expression Omnibus [22] and are accessible through GEO Series accession number GSE15582 .

Chromosomal location enrichment by Map genes to chromosomes

Genes used for hierarchical clustering have been also used to create the Genome view of our transcriptome data by using the dChip software (map of genes tool), as described in the dChip site .

P-values have been calculated for all stretches containing ≤20 selected genes to assess the statistical significance of gene proximity and the significant P-values have been reported in the analysis view. The "tightness" i.e. the rank distance of the genes on the two ends of the stretch was calculated by/n genes against that of n genes randomly put on the chromosome as described .

Functional annotations

Gene Ontology analysis was carried out using DAVID NIAID/NIH online functional annotation tool and ontology terms were retained according to the significance of P-values for any particular GO term. This indicated that there were more genes associated with this GO term than would be expected by random chance [23]. The results were independently confirmed by applying the GO Browser tool in NetAffx to the same set of differentially expressed genes. The associations between altered genes or pathways were further evaluated using the Ingenuity Pathways Analysis software (IngenuityR Systems).

Reverse transcription-coupled real-time PCR

Total RNAs from PLN of E-IAA positive and negative mice were used for Real Time PCR following the ABI software and method. To confirm gene expression levels identified in the arrays by RT-PCR, RNAs were aliquoted from the same samples used for the array hybridizations. Sequences corresponding to the investigated genes have been retrieved from the public database and aligned with Affymetrix target sequences using BLAST software . Primers and TaqMan probes [see Additional File 1] have been designed using the proprietary TaqMan software. We verified that the amplified fragments corresponded to the exact Affymetrix sequences. mRNA from an insulinoma cell line has been used as positive control and 18S rRNA for data normalization.

Results

E-IAA as a marker of autoimmune-related modifications in the PLN of NOD mice

We have established the following criteria for the selection of individual mice: autoimmune positive animals have been tested at 3, 4 and 5 weeks for the presence of IAA in the serum and selected when positive at least twice [24]. Negative phenotype has been considered when animals were negative for IAA at the same ages. A total of 9 animals have been selected, 6 negative for E-IAA and 3 E-IAA positive (Table 1). Even though the presence of E-IAA have been demonstrated to be an early marker of autoimmune diabetes, animals negative for IAA throughout their life may develop T1D, yet this remains a rare event (1 animal/15) [25]. In human, some cases of IAA long term positive individuals who do not develop T1D have been observed [26]. As seen on Table 1, regardless of the sub phenotypic attribution to the individual mice used in this study, one E-IAA negative sample (A36.4) was grouped, by clustering analysis, together with the E-IAA positive group, according to gene expression patterns (see below). This observation underlines the complexity of the autoimmune condition and indicates that the balance between pathogenic and physiological immune responses controls the outcome of disease. One possible explanation for the absence of E-IAA in the A36.4 PLN sample is that the selected sub-phenotype marks the disease stage without being per se the cause for the expression patterns identified. Therefore the initiation of the autoimmune process might be dependent upon the identified gene signatures and independent of the presence of IAA in the serum. In this case the E-IAA would be the consequence of an ongoing mechanism but not the cause.

Table 1

Subphenotypes of PLN used for transcriptome

Sample Name	E-IAA Phenotype	Gene expression patterns
A8.1	IAA neg	NEG
A12.1	IAA neg	NEG
A12.2	IAA neg	NEG
A36.1	IAA neg	NEG
A37.1	IAA neg	NEG
A36.4	IAA neg	POS
A9.6	IAA pos	POS
A15.4	IAA pos	POS
A15.6	IAA pos	POS

Correlation of E-IAA subphenotypes with gene expression patterns as shown by unsupervised hierarchical clustering on Fig. 2.

E-IAA related gene expression patterns in PLN

Our interest in this study was to identify gene expression variations solely due to the presence or absence of IAA, in a set of identical animals of the same age and same genetic background, which however are not synchronized for the evolution of the disease. Albeit the NOD strain is inbred, homeostatic differences inherent to each animal might influence the physiopathological status and in consequence the disease stage. Therefore, RNAs have been prepared from PLN, isolated from individual mice and used without prior amplification. This allows identifying and controlling for inherent biological variability of gene expression, between individual animals that might not be related to the E-IAA sub-phenotype and allowed us avoiding the possibility for any preferential transcript amplification.

The quality of whole gene expression data generated from each of the 9 individual animals has been checked using box plots metrics representation [see Additional File 2]. The mean fluorescence intensity values and the range of signal amplitudes were similar across all samples and consistent with a good hybridization quality of the arrays. A scatter plot of the raw data for all transcripts of the arrays from either E-IAA negative or positive samples clearly showed several probe sets to be differentially expressed between the 2 conditions (Figure 1). Expression levels for the up-regulated genes were higher in E-IAA positive PLN, while up-regulated variations in the E-IAA negative samples have been observed at a lower level (Figure 1).

Figure 1

Correlation of gene expression between E-IAA positive and negative PLN. Scatter plot representation of all MU74Av2 array elements (GS6.0 software), after hybridization with the RNA probes. A subset of elements that are distinct between the two arrays and which deviate the most in signal intensity is depicted by the colour codes: blue for the highly expressed in the E-IAA positive PLNs, red for the highly expressed in the E-IAA negative samples and yellow for the probe sets that do not show statistical significant changes between the two sets of samples in both X and Y "fluorescence intensities".

Statistically significant differences in gene expression levels between the E-IAA positive and E-IAA negative lymph nodes were determined (adjusted P-values < 0.05 and log2 fold changes >0.42). A total of 177 probe sets were deemed statistically significant by these criteria and corresponded to 165 unique genes, whilst 12 transcripts were identified repeatedly because they were represented on the chip by more than one probe set [see Additional File 3]. Unsupervised hierarchical clustering of the modulated probe sets confirmed clear sample segregation in the two phenotypic groups (Figure 2). Gene expression patterns correlated well with phenotypic evaluation for all individual mice except one: sample A36.4 is classified with the positive group (Figure 2, noted by *) whilst its phenotypic evaluation, as described above (Table 1), has identified it as a negative sample. This discrepancy, as discussed above, correlates with previous findings in human and the NOD mouse, whereas autoantibody negative individuals have still a probability to develop the disease [25,27].

Figure 2

Hierarchical clustering representation of differentially expressed probe sets between E-IAA positive and negative PLN samples. Log2 transformed data from 177 probe sets are represented in a matrix format wherein each row displays expression results for a single gene across the arrays and each column shows the relative expression levels for all the genes in each sample. Red represents relative expression greater than the median expression level across all samples, and green represents an expression level lower than the median expression level. The colour intensity represents the magnitude of the deviation from the median. The dendrogram at the left lists the genes and provides a measure of the relatedness of their expression profile in each sample. *Sample A36.4 corresponds to E-IAA negative phenotype despite its clustering according gene expression with the E-IAA positive samples (see also Table 1).

Interestingly, 77% (119 genes and 8 ESTs) of the differentially expressed transcripts were up regulated in the E-IAA samples while only 22% (34 genes and 3 ESTs) were down regulated. The majority of the up-regulated transcripts (69%, ie. 82 transcripts and 5 ESTs) have expression patterns varying over 2 fold while only 5 out of the 34 down-regulated genes and one out of the 3 ESTs (16%) showed over 2 fold differences between the two sets of samples [see Additional File 3].

The biological relevance of these trends in gene expression changes will obviously need to be tested on an individual gene basis.

Validation of microarray data by Real time PCR and immunohistochemistry

In order to validate the expression levels observed in the arrays, we selected 4 genes for quantitative real time PCR analysis (Q-RT-PCR): 2 regenerating islet-derived coding genes, Reg2 and Reg3a (Figure 3A &3B) and the 2 insulin genes, Ins1 and Ins2 (Figure 3C &3D) that have shown high gene expression differences in the arrays [see Additional File 3]. Similar trends of gene expression differences between the two sets of samples, in the arrays and in the Real time PCR were observed (Figure 3A-D).

Figure 3

Validation of expression patterns of selected genes differentially expressed in the PLN of NOD mice. A and B: Reg2 and Reg3a genes (A: Real Time PCR and B: arrays row data). C and D: Ins1 and Ins2 genes (C: Real Time PCR; D: arrays row data). Samples are PLN RNA from E-IAApos mice (A9.6 & A15.4) and from E-IAA neg mice (A12.2 & A8.1). E, F &G. Staining of histological sections with anti-Insulin antibodies, from pancreas (E), PLN (F) and Inguinal lymph nodes (G) from NOD mouse at 5 weeks.

Expression of insulin genes in the PLN was not expected since insulin is synthesized, stored and secreted by the β cells of the pancreatic islets in a highly regulated manner in human and mouse. However ectopic expression or so-called illegitimate transcription of self-antigens has been previously reported, mainly in the thymus [28] and spleen [29]. Cells expressing pancreatic autoantigens are also found in peripheral lymphoid tissues and the same cell can express more than one self antigen [15,30]. The functional significance of these observations has been questioned and evidence from studies in the mouse strongly suggested that self-antigen-expressing cells, in the thymus, may participate in negative selection and mediate tolerogenic signals, some of which may culminate in the apoptotic death of autoreactive lymphocytes [31].

Considering this data, we have stained NOD PLN with anti-insulin antibodies by immunohistochemistry (Figure 3F) and used pancreas as positive control (Figure 3E). Sparse distribution of insulin expressing cells was noted in the PLN (Figure 3F), while no staining has been obtained in inguinal lymph nodes used as negative controls (Figure 3G), thus confirming the ectopic insulin gene expression observed in our data. It is interesting to note that insulin genes have been also found expressed in the PLN of NOD mice at 4 weeks in a recent publication [32] although at lower levels.

Noteworthy that except insulin no other endocrine-specific transcripts (somatostatin, glucagon, somatostatin receptor or Islet amyloidal polypeptide) have been identified in our data set [see Additional File 4]. As expected and similarly to the endocrine specific transcripts, no other transcripts known to be specific either only to the exocrine pancreas or to the developing pancreas were expressed in the PLN at 5 weeks [see Additional File 4]. While expression of insulin genes in the PLN is not a novel finding, the high levels of expression found in the PLN of E-IAA positive NOD mice was not expected and is bewildering. One explanation might be that in the E-IAA positive PLN, APCs expressing insulin might act, locally in the PLN, as insulin-presenting cells. This would contribute to lack of tolerance to this molecule and consequently to the instigation of insulin autoreactive lymphocyte trafficking, leading to infiltration of the islets resulting to T1D. However a role of insulin in local tissue architecture, independent of its autoantigen function, cannot be excluded.

Transcriptional signatures specific to E-IAA sub-phenotype

An important aspect of our experimental design concerns the early selection of NOD mice by the presence of E-IAA, used as a marker of early autoimmune activity, at an age whereas homeostatic changes in the pancreas might take place. It has been demonstrated that increased neonatal beta cell apoptosis is taking place with a peak at 2-3 weeks after birth, in T1D animal models including the BB rat and the NOD mouse [33]. The cross-talk between the pancreas and the adjacent LN during or immediately after this phase might be reflected by discrete gene expression changes in this later tissue.

We undertook a series of exploratory functional data analysis aiming to comprehend the roles of the identified gene expression variations in E-IAA-dependant islet autoimmunity. Forty-one out of the 119 up-regulated genes showed differential expressions over 5 fold [Additional File 3]. This indicates that the autoimmune process as reflected by the presence of IAA in the positive NOD mice is marked by local gene expression changes in the PLN. This might include changes in the immune process, occurring in this lymphoid tissue and/or changes in local tissue architecture and remodelling, not only of the lymph node but also of the pancreas, as mentioned above, assisted or mediated by the genes expressed in the PLN.

i. Cellular compartment localization

An over representation of genes coding for extracellular proteins has been identified as 38% (62/165) modulated genes belonged to this category (P-value = 8.49E-16) (Table 2). This does not seem to be a random distribution since the first array of the MG-UG74Av2 series contains 2,057 transcripts coding for secreted proteins in a total of 12,488, representing 16% of the total number of genes on the arrays. Moreover this class of proteins is overrepresented in the E-IAA positive list of genes in the PLN (Figure 4a). The majority (30/41) of the over 5 fold expressed genes in the up-regulated samples, code for secreted proteins (GO P-value = 3.46E-22). This might reflect the cross talk taking place between the PLN and the pancreas, during the pre-inflammatory stages. Moreover secreted and membrane-associated proteins are an important class of proteins, as they have the potential to be detected in biological fluids and thus may be potential diagnostic markers or therapeutic targets.

Table 2

Annotations chart report for over-representation of functional categories.

	UP & DOWN REGULATED
GO CATEGORY/Term	Gene Count (Ease score P-value)
Cellular compartment
Extracellular region	60 (9E-22) ↑, 2 ↓
Endoplasmic reticulum	11 (0.02) ↑
Nuclear	8↑ 6 (0.02) ↓
Organelle	35↑, 18 (0.03) ↓
Mitochondrion	10 (0.04) ↑
Molecular function
Catalytic activity	65 (1.4E-5) ↑, 9 ↓
Hydrolase	41 (8.9E-8) ↑, 3 ↓
Peptidase activity	27 (2.9E-10) ↑ 1 ↓
Serine-type peptidase activity	25 (8.3E-19) ↑
Actin binding	15 (8.2E-4) ↑ 7 (0.02) ↓
Calcium binding **	10 (NS) ↑
Vitamin binding	6 (0.0008) ↑
Transferase NO-	3 (0.05) ↑, 3 ↓
Antigen binding**	5 (3.6E-4) ↓
Tissue kallikrein activity	8 (8.5E-12) ↑
Elastase activity	2 (2.5E-2) ↑
Biological process
Metabolic process	74 (6.4E-4) ↑, 17 (0.02)
Immune-related process**	68a ↑ 14 (3.6E-6)
Proteolysis	27 (1.5E-11) ↑
Cell structure**	26 (0.02) ↑
Lipid, Fatty acid & steroid metabolism**	15 (1.31E-4) ↑
mRNA transcription regulation**	22(0.09) ↓
Cell communication**	9 (0.05) ↑
Digestion	8 (3.73E-8) ↑
Cellular homeostasis	7 (0.01) ↑
IMMUNE-RELATED PROCESS**
B-cell & Antibody-mediated immunity**	9 (1.2E-4) ↓
Response to stress	18 (8.3E-5) ↑ 2.3E-2↓
MHC-I-mediated Immunity**	16 (0.09) ↑
Immunity & Defense**	16 (0.03) ↑ 4 (0.2) ↓
MHC-II-mediated Immunity**	15 (0.002) ↑
Response to external stimulus	14 (4.9E-5) ↑, 6 (0.5) ↓
Response to wounding	10 (5.09E-4) ↑
Inflammatory response	8 (0.001) ↑
Cytokine & chemokine mediated signaling
pathway**	7 (0.06) ↑
Other Immune defense**	7 (0.05) ↑
Humoral Immune response	4 (2.2E-3) ↑
Antigen presentation	3 (1.1E-2) ↓
Lymphocyte-mediated immunity	3 (0.008) ↓
Adaptive Immunityb	3 (0.008) ↓

The most relevant GO [23] and PANTHER [38] terms associated with the list of 165 E-IAA modulated genes are shown. Gene counts are the number of genes in the list at the corresponding term and arrows indicate up and down-regulated gene expression. P-value is a modified Fisher exact P-value, Ease score. Graphical representations of functional categories are shown on Figure 4. The list of all genes is available in Additional File 5.

** PANTHER annotations

a Significance for combined annotation from GO (EASE): 8.35E-5

b Adaptive immune response based on somatic recombination of immune receptors built from immunoglobulin super family domains

Figure 4

Functional annotation for up- and down-regulated genes. Functional categories distribution are according to GO and PANTHER annotations. a. Cellular component, b. Molecular function and c. Biological process. Significance for functional categories retained was 8.3E-19

a Significance for combined annotation from GO (EASE): 8.35E-5