Potentially Pathogenic Immune Cells and Networks in Apparently Healthy Lacrimal Glands

Abstract

Lacrimal glands of people over 40 years old frequently contain lymphocytic infiltrates. Relationships between histopathological presentation and physiological dysfunction are not straightforward. Data from rabbit studies have suggested that at least two immune cell networks form in healthy lacrimal glands, one responding to environmental dryness, the other to high temperatures. New findings indicate that mRNAs for several chemokines and cytokines are expressed primarily in epithelial cells; certain others are expressed in both epithelial cells and immune cells. Transcript abundances vary substantially across glands from animals that have experienced the same conditions, allowing for correlation analyses, which detect clusters that map to various cell types and to networks of coordinately functioning cells. A core network—expressing mRNAs including IL-1α, IL-6, IL-17A, and IL-10—expands adaptively with exposure to dryness, suppressing IFN-γ, but potentially causing physiological dysfunction. High temperature elicits concurrent increases of mRNAs for prolactin (PRL), CCL21, and IL-18. PRL is associated with crosstalk to IFN-γ, BAFF, and IL-4. The core network reacts to the resulting PRL-BAFF-IL-4 network, creating a profile reminiscent of Sjögren’s disease. In a warmer, moderately dry setting, PRL-associated increases of IFN-γ are associated with suppression of IL-10 and augmentations of IL-1α and IL-17, creating a profile reminiscent of severe chronic inflammation.

I. INTRODUCTION

Dry eye disease is one of the most common morbidities eye care specialists are called upon to treat. The etiology is widely recognized as multifactorial. Certain inflammatory processes that arise in the lacrimal glands, most notably Sjögren’s disease,^1,2 graft-versus-host disease,³ sarcoidosis,^4,5 Wegener’s granulomatosis,⁶ HTLV-associated dacryoadenitis,⁷ and HIV-associated diffuse infiltrative lymphocytosis,^8,9 are associated with severe physiological dysfunction that leads to keratoconjunctival inflammation. Recent data suggest that many cases of dry eye disease result from age-related decreases in meibomian gland function, impairment of the tear film lipid layer, accelerated evaporation, and increased tear film osmolarity.^10–13 Accordingly, some authors have proposed that subsequently diagnosed lacrimal gland physiological dysfunction develops as an untoward consequence of the mechanisms that initially drive compensatory increases in lacrimal fluid production,^10,13 but possible cellular bases for such a phenomenon have not yet been proposed.

A glossary of terms used in this review is appended to the article.

A. Inflammation and Physiological Dysfunction: Pathophysiological Diversity

Classic histopathological findings show chronic inflammation to be a normal, almost ubiquitous, concomitant of aging in the lacrimal glands. Waterhouse found lymphocytic infiltrates in 65% of lacrimal glands from women aged 75 years and older.¹⁴ Damato et al reported infiltrates in 70% of glands from women and men, mean age 62 ± 17 years.¹⁵ They also suggested that cases in which atrophic and fibrotic changes are not associated with notable infiltrates may be the sequelae of previous inflammatory processes. Obata et al reported infiltrates in 69% of palpebral glands and 83% of orbital glands from women and men between 40 and 87 years of age.¹⁶ Waterhouse found infiltrates in fewer of glands from elderly men (20%), but, strikingly, he also found that infiltrates begin to appear early in adult life and that they occur at similar rates (22% and 18%, respectively) in glands from women and men younger than 40 years of age. These numbers point to the conclusion that the prevalence of lymphocytic infiltrates is substantially greater than the prevalence of clinical dry eye disease, implying that the infiltrates that commonly develop are immunopathologically diverse, with diverse impacts on lacrimal physiological function. Moreover, as dry eye disease is several-fold more prevalent in women than in men, the disparity between the processes that do and do not cause clinically significant physiological dysfunction must be greater in men.

The contrast between Sjögren’s disease and Mikulicz’s disease^17,18 illustrates the principle that some immune cell infiltrates are associated with physiological dysfunction, while others are relatively benign, at least with respect to lacrimal gland physiological function. Moreover, findings from laboratory studies suggest a general principle to account for the difference: Some mediators known to be produced by inflammatory infiltrates cause physiological dysfunction in ex vivo models, some do not, and some may augment fluid production. Nitric oxide (NO) impairs stimulus-secretion coupling in human labial salivary gland acinar cells,¹⁹ interleukin (IL)-1 suppresses protein secretion in ex vivo mouse lacrimal gland preparations,²⁰ and both IL-1β and IL-6 suppress ionic currents associated with Cl⁻ secretion and fluid production by a reconstituted rabbit lacrimal acinar epithelial model.²¹ Similarly, chronic stimulation with certain G protein-coupled receptor agonists, including cholinergic receptor agonists and biogenic amines, can impair both protein secretio ^22,23 and Cl⁻ secretion21, and other biogenic amines potentiate Cl⁻ secretion and presumably augment lacrimal fluid production, but impair protein secretion.^21–23 Interestingly, in another model, human retinal pigment epithelium, interferon (IFN)-γ enhances ion fluxes related to fluid transport.²⁴

There has been some exploration of the diversity of cellular pathophysiologies associated with lacrimal gland lymphocytic infiltrates.^2,5–27 The extent of the diversity is not yet fully established, however. The most common histopathological diagnosis, chronic dacryoadenitis, presents with diverse features, as immune cells may be concentrated in periductal/perivenular foci, or they may pervade large areas formerly occupied by acini, and they may affect a single lobule or most or all lobules; the infiltrates may or may not be accompanied by atrophic changes in the acini, in the ducts, or in both epithelial structures, and by fibrotic changes in the surrounding stromal spaces. Thus, chronic dacryoadenitis likely encompasses a spectrum of immunopathological processes.

B. Epithelial Factors and Development of Immunopathological Infiltrates

The mechanisms that cause noninfectious inflammatory infiltrates to develop in the lacrimal glands are unknown. A concept that has had considerable intellectual appeal is that the steroid reproductive hormones influence the spectrum of paracrine mediators that the gland’s epithelial cells express, and the local mediators in turn influence the balance between inflammatory and immunoregulatory activities in the gland.²⁸ The importance of ongoing, constitutive peripheral immunoregulatory mechanisms is emphasized by the identification of lacrimal epithelial cell internal vesicle trafficking pathways that may constitutively release autoantigens to the lacrimal gland’s stromal spaces.²⁹ Androgen therapy has proven highly beneficial in some murine models of inflammatory lacrimal gland diseases,³⁰ and a preliminary case review study has suggested that androgen supplementation of estrogen-progestin replacement therapy ameliorates signs and symptoms in women with severe dry eye disease.³¹ Studies have found a number of intriguing clues into how the lacrimal epithelia function as a nexus between hormonal status and local immunological activity; nonetheless, this nexus has remained largely a black box, slow to yield insights into its inner workings.

It is has been known for some time that epithelial cells in apparently healthy lacrimal glands express several important immune-response-related mediators, most notably transforming growth factor (TGF)-β and prolactin (PRL),^32,33 the latter being a cytokine as well as a hormone. Both are highly pleiotropic, and, as will be shown in this report, the influences PRL exerts in the lacrimal gland clearly are context-dependent. TGF-β mediates the immunoregulatory actions of natural (T_RN-) and adaptive (T_H3) T cells, but it also is involved in adaptive mucosal immunity, development of chronic inflammatory responses, and inflammation-associated fibrosis. Evidence from ex vivo studies indicates that TGF-β, secreted by acinar epithelial cells from rat lacrimal glands, directs monocytes to mature as dendritic cells with immunosuppressive- or immunoregulatory functions.³⁴ This finding has been interpreted as suggesting that lacrimal epithelial cell-derived TGF-β orchestrates the ongoing generation of regulatory T cells, assumed to be CD4⁺, T_H3 cells. However, it is not clear how the steroid reproductive hormones might interact with TGF-β to determine susceptibility to immunopathology. In murine lacrimal glands, steroid hormone influences on the expression TGF-β₁ and TGF-β₂, the isoforms most strongly implicated in immunoregulation, appear modest.^3,5–37 Androgens upregulate expression of TGF-β₃, but the role TGF-β₃ might play in lacrimal immunoregulation has not been investigated.

PRL actions include promoting T cell and B cell proliferation^38–40; inducing expression of IFN-γ⁴¹; inducing immature dendritic cells to differentiate as antigen presenting cells which promote T_H1 activation⁴²; and supporting expression of IL-4.^43,44 A high molecular weight form of PRL has been reported to be elevated in biopsied labial salivary glands from patients with Sjögren’s disease, and PRL has been implicated in striking cellular physiological changes in rabbit lacrimal glands⁴⁵ and ex vivo acinar cell models.⁴⁶ A study, reported in preliminary form, has linked adenovirus vectormediated, transient, overexpression of PRL to severe, acute immunopathology in rabbit lacrimal glands. However, pregnancy, a state of physiological hyperprolactinemia, has been reported to be associated with immunoarchitectural changes that appear benign.⁴⁷

Evidence confirms that the steroid reproductive hormones influence the expression of myriad additional genes, many immune response-related, in murine lacrimal glands,^36,37 and analyses indicate that the higher levels of androgens in males largely account for gender-related dimorphisms in transcript expression.³⁶ However, the impacts of steroid hormones clearly must be multifactorial, as both androgens and estrogens support expression of genes that are considered pro-inflammatory and genes that are considered anti-inflammatory.^36,37

C. Recent Insights from Rabbit Studies

RT-PCR primer and probe sequences have been designed for a number of rabbit immune response-related gene transcripts, and a study using these tools has yielded surprising findings.⁴⁸ Of 28 lacrimal glands taken from young adult virgin female rabbits that had been raised in an out-of-doors setting, only one gland presented with obvious immunopathological lesions. However, all glands contained mRNAs that typically are associated with innate inflammatory responses, e.g., CXCL8, IL-1α, IL-1β, and IL-6; mRNAs that function in the ectopic immune inductive tissues characteristic of Sjögren’s- and Mikulicz’s diseases, e.g., CCL21, CXCL13, BAFF, IL-4, and IL-18^49,50 and mRNA for IL-2, which typically is expressed by T_H1 cells. mRNA for IFN-γ, the hallmark cytokine of T_H1 responses, was detected in the glands from three of the four groups of animals that were studied.

The abundances of many transcripts varied through wide ranges across the individual glands from each group, but, despite the large intragroup variability, it was possible to discern systematic associations between the median abundances of certain transcripts and the environmental conditions the rabbits had experienced. Median abundances of one cluster of transcripts increased in association with a composite measure of the desiccating influence of the environment during the 30 days before study. In contrast, median abundances of IFN-γ mRNA appeared to decrease in association with increasing desiccating influence. Median abundances of another cluster of transcripts, which included PRL, were highest in the group that had experienced the highest temperature. These associations suggested that the immune cells present in apparently healthy lacrimal glands organize themselves into at least two different networks, one that responds to signals related to environmental dryness, and one that responds to signals related to high environmental temperature. Median abundances of additional transcripts varied across the groups, but with no simple, readily discernible association with either environmental parameter.

II. NEW FINDINGS

We undertook experimental studies and analyses on the premise that information about the composition and behavior of the immune cell networks that form in apparently healthy glands would provide clues into how the most prevalent immunopathological states arise and set the stage for understanding how reproductive steroid hormones influence susceptibility to development of one state or another. The goals were to: 1) determine how expression of selected transcripts is distributed between immune cells and the major epithelial structures; 2) assay transcript abundances in lacrimal glands from an additional group of rabbits, selected as having experienced conditions intermediate between the group that had experienced the driest conditions and the group that had experienced the hottest conditions; 3) elucidate the networks that had formed in the glands from each of the five groups; and 4) assess the separate influences that dryness and high temperature might exert on the networks’ activities. As will be seen, the findings point to a multiplicity of networks, including a core network, which responds adaptively to signals related to dryness and reactively to signals from other cells and networks, and to several cell types which respond concurrently but not necessarily coordinately to signals related to high temperature. The various cells and networks can engage in both positive- and negative crosstalk with each other, and such crosstalk can lead to the emergence of more complex networks.

A. Materials and Methods

1. Animals

All protocols conformed to the Association for Research in Vision and Ophthalmology (ARVO) Resolution on Use of Animals in Ophthalmic Research and were approved by the University of Southern California Institutional Animal Care and Use Committee. Groups of young adult virgin female New Zealand white rabbits, aged 16–18 weeks and weighing 3.5–4 kg, were acquired from Irish Farms (Norco, CA). Irish Farms is a non-barrier facility, and each group experienced a different combination of humidity and temperature arising from natural meterological variations.

2. Maximal Dryness and High Temperature

Meteorological data recorded at a station located near the Irish Farms facility were obtained from weatherunderground.com. The maximal daily dryness, Δ%H, was defined as 100% – daily low humidity. Mean values of maximal daily dryness and daily high temperature were calculated for the 30 days before each group of animals was moved to the University Vivaria; the mean values are plotted against Δ%H and T as orthogonal axes in Figure 1. Animals were consistently euthanized and necropsied 4 days after arrival at the University Vivaria

Figure 1.

Average daily maximal dryness and high temperature experienced by experimental groups during 30 days before arrival at USC Vivaria. Animals were consistently euthanized and necropsied 4 days after arrival.

3. Designation of Groups and Individual Glands

Lacrimal glands for this study were obtained from five groups of rabbits, designated with the letter V, for virgin, and with superscripts denoting the mean daily maximal dryness and mean daily high temperature, i.e., V^58%,17°, V^61%,27°, V^68%,37°, V^72%,32°, and V^82%,29°. Group V^58%,17° comprised two rabbits; groups V^61%,27° and V^68%,37° each comprised three rabbits; group V^72%,32°, five rabbits; and group V^82%,29°, six rabbits. A group of six term-pregnant animals, designated P^82%,29°, was obtained at the same time as group V^82%,29°. A preliminary analysis has shown that pregnancy exerts significant influences on the abundances of many immune response-related transcripts,⁵¹ and a full report of those findings is in preparation. Individual glands were designated according to the sequence in which the rabbits from each group were euthanized—V^58%,17°01 and V^58%,17°02; V^61%,27°01 – V^61%,27°03; V^68%,37°01 – V^68%,37°03; V^72%,32°01 - V^72%,32°0.5; V^82%,29°01 – V^82%,29°06; and P^82%,29°01 – P^82%,29°06—and according to whether they were from the left or right eye, i.e., OS or OD.

4. Tissue Collection and Processing

Inferior lacrimal glands from all groups except V^58%,17° were divided into parts placed in RNALater™ for RNA extraction and real time RT-PCR analyses; parts placed in 10% formalin for paraffin embedding and H&E staining; and parts placed in OCT for immunohistochemical staining. The OCT-embedded samples from V^68%,37° andV^61%,27° were lost in a freezer failure after staining for RTLA and CD18 was completed. Samples from V^82%,29° were allocated to several projects, and limited amounts were available for this study.

5. Immunohistochemical Staining and Image Analysis

Frozen sections were stained, examined, and analyzed according to methods that have been described in previous publications.⁴⁸

6. Laser Capture Microdissection

Samples of epithelial cells from acini, intralobular-, interlobular-, and intralobar ducts, and samples of immune cell accumulations were obtained by laser capture microdissection as described by Ding et al.⁵²

7. RNA Extraction, Reverse Transcription, and Real-Time RT-PCR

Methods for mRNA extraction, reverse transcription, and real-time polymerase chain reaction (PCR) are described elsewhere.⁵³ RNA extracts from the individual glands were analyzed separately. The primer- and probe sequences were based on published rabbit gene sequences. The abundance of each target mRNA in each tissue extract was calculated relative to the abundance of GAPDH mRNA in the same extract. The abbreviations for the transcripts assayed are presented in the Glossary, along with basic information about the transcripts’ products’ roles and the cells that express them.

8. Data Analysis and Presentation

Sigmaplot 12.0 (Systat Software, San Jose, CA) was used for Pearson’s correlation tests and linear- and non-linear regression analyses.

a. Glands That Appeared as Outliers

Four individual glands—V^61%,27°02.OS, V^68%,37°02.OS, V^68%,37°03.OS, and V^72%,32°01.OD— stood apart from the other glands in their respective groups by having considerably higher abundances of certain transcripts. Transcript abundances in each of these glands are designated with unique symbols in the figures, and each is discussed as an informative case study.

b. Correlation Analyses

Correlations between transcript abundances among the individual glands of each group were analyzed by Pearson’s test for multiple correlations. Transcript abundances in glands that appeared as outliers were excluded from Pearson’s analyses in order to minimize type 1 and type 2 errors. The numbers of glands for which values were analyzed were: V^58%,17°, four; V^61%,27°, five; V^68%,37°, four; V^82%,29°, twelve; V^72%,32°, nine. Rows and columns in the table of Pearson’s statistics for each group were organized to maximize the blocks of contiguous cells containing values of P ≤.05. This empirical approach, which required no assumptions about the identities of the cell types that expressed the transcripts, made it possible to discern correlation clusters, i.e., clusters of transcripts whose abundances varied coordinately.

Non-linear regression analyses were performed to identify significant exponential relationships between transcripts:

$$\begin{array}{ll}\text{exponential}\text{growth},\text{two}-\text{parameter}& [\text{mRNA}]={[\text{mRNA}]}_0·{\mathrm{e}}^{\mathrm{b}·\mathrm{x}}\\ \text{exponential}\text{growth},\text{three}-\text{parameter}& [\text{mRNA}]={[\text{mRNA}]}_{\mathrm{B}}+{[\text{mRNA}]}_0·{\mathrm{e}}^{\mathrm{b}·\mathrm{x}}\\ \text{exponential}\text{decay},\text{two}-\text{parameter}& [\text{mRNA}]={[\text{mRNA}]}_0·{\mathrm{e}}^{-\mathrm{b}·\mathrm{x}}\\ \text{exponential}\text{decay},\text{three}-\text{parameter}& [\text{mRNA}]={[\text{mRNA}]}_{\mathrm{B}}+{[\text{mRNA}]}_0·{\mathrm{e}}^{-\mathrm{b}·\mathrm{x}}\end{array}$$

where x = the abundance of a transcript selected as a reference.

c. Mapping Correlation Clusters to Cells and Networks

The correlation clusters discerned with Pearson’s analyses and clusters of non-linearly- related transcripts were mapped to cell types or to networks of cells on the basis of the data from the laser capture microdissection surveys and published information about cell types that express the various transcripts, summarized in the Glossary. The maps are presented in standard histoarchitectural schemas comprising generic epithelial cells, lymphocytes, and MMΦDC. Variations in abundance of some transcripts, e.g., transcripts expressed by epithelial cells, may reflect variations in the levels of expression in cells whose relative numbers do not vary. Variations in abundance of other transcripts may reflect both variations in the numbers of cells expressing the transcripts and also variations in levels of expression per cell.

In some cases, it was possible to infer hypotheses about the signals responsible for the associations between different transcripts’ abundances. For example, increasing expression of PRL is known to mediate increased expression of IL-4 and increased expression of IFN-γ; increasing expression of IL-4 is known to mediate decreased expression of IFN-γ; and IL-10 and IFN-γ are known to counter-regulate each other’s expression. The analyses also discerned associations that have not been reported previously. The signals that coordinated the programs of expression underlying the new associations are not known, and it is possible that those signals were mediated by products of transcripts that have not been assayed. Therefore, for the purposes of this communication, the generic term, “crosstalk,” is used as a rubric for both the hypothesized and the unknown signals.

As none of the networks has previously been described in lacrimal glands, the names assigned to them were selected to evoke salient network features, whether functions, cell types, or mediators.

d. Judging Whether Networks Might be Intact in Glands that Appear as Outliers

Both linear and nonlinear regression analyses were used to identify cases where transcript abundances in the glands that appeared as outliers were likely to have been subject to the same forms of crosstalk as the other glands of their respective groups.

e. Relationships between Transcript Abundances and Meteorological Variables

In the initial analysis of glands from V^58%,17°, V^61%,27°, V^68%,37°, and V^82%,29°, values for all transcripts in the glands that appeared as outliers were excluded from the statistical analyses, and the mean transcript abundances in the remaining glands were tested for significant relationships to a composite variable, the number of desiccating T/H days.⁴⁸ For the new analyses described in this report, the medians of the entire samples of values from each group were determined, and the 30-day mean high temperature and the 30-day mean dryness ( $\overline{\mathrm{\Delta }\%\mathrm{H}}$) were treated as independent variables. The relationships between median abundances and mean daily high temperature and mean daily dryness were tested using the same exponential growth functions listed above, with $\mathrm{x}=\overline{(\mathrm{\Delta }\%\mathrm{H})}$ or x = T̄ $\mathrm{x}=\overline{(\mathrm{\Delta }\%\mathrm{H})}$ or x= T̄. As will be shown, the exponential functions provided strong, highly significant, empirical descriptions of relationships between several transcripts’ median abundances and either dryness or high temperature.

The functions of exponential growth with dryness and exponential growth with heat also were used as heuristics for describing the behaviors of median transcript abundances over specific domains of the $\overline{\mathrm{\Delta }\%\mathrm{H}}-\overline{\mathrm{T}}$ surface when transcripts appeared to have been subject to additional influences in one group or another. As will be seen, the additional influences appeared in many cases as crosstalk arising from cells and networks that expressed other transcripts. When the median abundances and heuristics are plotted against the appropriate axis, the additional influences can be seen as acting in one group or two groups to augment abundance above-, or suppress it below, the value predicted by its heuristic.

B. Histoarchitectural Organization of Transcript Expression

As summarized in Figure 2, the laser capture microdissection survey indicated that mRNAs for lipophilin CL, CCL2, and IL-2 were most abundant in acinar cells, but they also were detectable in immune cell accumulations. mRNAs for PRL and APRIL were most abundant in acinar cells and also abundant in ductal epithelial cells, but much less abundant in immune cell accumulations. TGF-β₂ mRNA was localized to ductal epithelial cells and present at similar levels in the three duct segments. CCL4 mRNA was present at similar levels in interlobular duct cells and in immune cell accumulations. IL-1RA mRNA was present at similar levels in interlobular duct cells, intralobar duct cells, and immune cell accumulations. Decorin mRNA was most abundant in immune cell accumulations, but it also was present in acinar cells and in each of the duct segments. mRNAs for TGF-β₁, CD25, and BAFF were predominantly localized to immune cells, but they also were detectible in intralobar duct epithelial cells. mRNAs for IL-1α, IL-1β, IL-6, and IL-10 were predominantly localized to immune cell accumulations.

Figure 2.

Histoarchitectural organization of transcript expression: Relative abundances in acini, intralobular ducts, interlobular ducts, and intralobar ducts. Two V^82%,29° glands were microdissected with a laser capture system. Two additional glands from term pregnant animals (P^82%,29°) also were microdissected; those data will be reported elsewhere. The intensity of shading in each structure is proportional to the transcript’s highest mean relative abundance. Note that the highest abundances of TGF-β₁ were found in periductal/perivenular immune cell accumulations, while the highest abundances of TGF-β₂ were found in the three duct segments. The highest abundances of CCL2 were found in acinar cells, and the highest abundances of CCL4 were found in interlobular duct cells. The highest abundances of mRNAs for lipophilin CL and for the mitogenic cytokines, IL-2, APRIL, and PRL also were found in acinar cells, and mRNAs for APRIL and PRL were additionally found in cells of all three duct segments.

C. Intergroup Variations, Intragroup Variations, Correlation Clusters, Cells, and Networks

Supplemental Figure 1 presents an overview of relationships between abundances of mRNAs for IFN-γ, IL-4, and IL-1α and abundances of mRNAs for IL-10 and PRL across the glands from each of the five rabbit groups. It also presents relationships between abundances of mRNAs for IL-17A, iNOS, and CD1d and mRNAs for IL-10 and PRL across the glands from the two groups in which they could be assayed, V^72%,32° and V^82%,29°. Supplemental Figure 2 presents the abundances of transcripts that appeared to be systematically related to dryness, and Supplemental Figure 3 presents the abundances of transcripts that appeared to be systematically related to high temperature exposure. Representative relationships are described in detail in Sections II.D.1 and II.D.2. Supplemental Figure 4 presents the abundances of the transcripts for which simple relationships to dryness or temperature could not be discerned.

The Supplemental Figures demonstrate that many transcripts’ abundances varied considerably across the individual glands of each group. As illustrated in Figure 3, much of the intragroup/intergland variability stemmed from phenomena that were localized within individual glands, as there was little correlation between many transcripts’ abundances in the OS gland and the OD gland from the same animal. Nevertheless, such intragroup/intergland variability was highly systematic in the sense that the abundances of clusters of transcripts varied coordinately. Statistics from Pearson’s multiple correlation tests are presented in Supplemental Tables 1–5; values of P ≤.05 are shaded to highlight the empirical correlation clusters. Nonlinear regression analyses identified additional significant relationships. Figure 4, Figure 5, Figure 6, Figure 7, Figure 8 present the schemas with the correlation clusters and nonlinear relationships mapped to various cells and networks.

Figure 3.

Transcript abundances in OS and OD lacrimal glands from V^82%,29°. The transcripts that showed the largest OS-OD variations were positively associated with a core adaptive-reactive network. The variability between companion glands emphasizes the importance of stochastic events and positive crosstalk at the level of the individual gland.

Figure 4.

Schema depicting the correlation clusters detected and the cell types and networks inferred in glands from V^58%,17°. Pearson’s correlation statistics and empirical correlation clusters are presented in Supplemental Table 1. Relationships between the abundances of selected transcripts and the abundances of mRNAs for IL-10 and PRL are presented in Figure 9. Brackets { } indicate transcripts that sorted to correlation clusters and parenthesis ( ) indicate transcripts that were assayed but not found to be significantly associated with other transcripts. Green dots (●) indicate positive associations; red dots (●) indicate negative associations. The large cells at the left represent epithelial cells without distinguishing between the acini and the three duct segments analyzed. CCL4 mRNA was found both in interlobular duct epithelial cells and in immune cell accumulations; whether variations at one site are associated with variations at the other has not been determined. Decorin mRNA was detected in acinar cells and in all ductal segments, but it was most abundant in immune cell accumulations, and, for simplicity, it is shown as being localized to immune cell accumulations. In groups that had been exposed to more dryness and higher temperatures (Figure 5 – Figure 8), mRNAs for IL-1β and IL-2; IL-6; CD8; and IL-1α, CTLA-4, and IL-10; and, in most cases, CCR5 sorted to the same correlation cluster, creating the signature of the core, adaptive-reactive network. Theses transcripts are denoted in bold type.

Figure 5.

Schema depicting the correlation clusters detected and the cell types and networks inferred in glands from V^61%,27°. Pearson’s correlation statistics are presented in Supplemental Table 2. Relationships between abundances of selected transcript and abundances of mRNAs for CD4 and IL-10 are presented in Figure 10. Details are as described in the legend to Figure 4. The core, adaptive-reactive network developed independently of the T_H3-remniscent network, and it generated negative crosstalk to cells—presumably T_H1 cells or NKT—that expressed IFN-γ mRNA. T_H3-reminiscent network engaged in positive crosstalk with cells—possibly, T_H2 cells—that expressed IL-4 mRNA and with cells that expressed mRNAs for CXCL13, Casp1, MMP-9, CCL21, CD3ε, and CD3ζ. The phenomena depicted as emergent were associated with appearance of the frank immunopathological state in gland V^61%,27°02.OS. These phenomena might have been triggered either by some adventitious event, such as trauma or infection, or by a stochastic event, such as the coincidence of large numbers of T_H2-reminiscent cells and large numbers of cells of the core, adaptive-reactive network giving rise to a new network.

Figure 6.

Schema depicting the correlation clusters detected and the cell types and networks inferred in glands from V^68%,37°. Pearson’s correlation statistics are presented in Supplemental Table 3. Details are as described in the legend to Figure 4. Associations between the abundances of mRNAs for PRL, IFN-γ, IL-4, and BAFF are shown in Figure 12.A. Epithelial cells expressing PRL mRNA engaged in positive crosstalk with TH1- reminiscent cells expressing IFN-γ and with T_H2-reminicent cells expressing IL-4. They also engaged in negative crosstalk with T_H3-reminiscent cells. Negative crosstalk between IL-4-expressing cells and BAFF-expressing cells was abrogated at higher levels of PRL expression, and negative crosstalk between IL-4 expressing cells and IFN-γ-expressing cells emerged at the highest level of PRL expression. The core, adaptive-reactive network initially developed independently of the interplay between cells expressing mRNAs for PRL, IL-4, BAFF, and IFN-γ. The emergent phenomena depicted represent the state in gland V^68%,37°02.OS, associated with positive crosstalk between the core, adaptive-reactive network; cells expressing PRL, BAFF, and IL-4; and cells expressing CCL2 and CCL4. These interactions played out in a local setting of concurrently high levels of IL-18 mRNA expression and high levels of CCL21 mRNA expression. With the exception of negative crosstalk between cells expressing CCL21 and cells expressing CXCL13, the resulting transcript expression profile is reminiscent of Sjögren’s ectopic immune inductive tissue. The state that emerged in gland V^68%,37°03.OS (not shown) may have been associated with the appearance of and additional source of IL-10 mRNA expression (Figure 12.B).

Figure 7.

Schema depicting the cell types and networks inferred in glands from V^72%,32°, based on Pearson’s correlation statistics presented in Supplemental Tables 4 - 6; certain associations that were significant across the entire sample of glands (Supplemental Table 6) are omitted to simplify depiction of the emergent phenomena. Associations between the abundance of IFN-γ mRNA and the abundances of mRNAs for IL-10 and PRL are shown in Figure 13.A. Relationships between the abundances of additional transcripts and the abundances of mRNAs for IL-10 and IFN γ are shown in Figure 14. Details are as described in the legend to Figure 4. The findings presented in Figure 13.A suggest that a critical feature of the initial state was positive crosstalk between epithelial cells expressing PRL mRNA and cells—presumably T_H1 cells or NKT cells—expressing IFN-γ mRNA. Rather than abrogating this crosstalk, T_H3-reminiscent cells expressing TGF-β₂ accumulated coordinately. Emergence of the second state may have depended on abrogation of the PRL mRNA - IFN-γ crosstalk by increasing levels of IL-10 expression, associated with ongoing development of the core, adaptive-reactive network. Emergence of a third state, in gland V^72%,32°01.OD, was characterized by disproportionately large increases in the expression of IL-17A, IL-10, CD8, and CD28. As can be seen in Supplemental Figure 1, all three states are characterized by relatively low levels of IL-10 expression and relatively high levels of IL-1α and IL-17A expression.

Figure 8.

Schema depicting the correlation clusters detected and the cell types and networks inferred in glands from V^82%,29°. Pearson’s correlation statistics are presented in Supplemental Table 7. The abundance of IFN-γ mRNA ( ) was below the limit of detection. Details are as described in the legend to Figure 4. Development of the core, reactive-adaptive network was associated with positive crosstalk to T_H2-reminiscent cells, which expressed IL-4; with positive crosstalk to cells that expressed CCL21; and, indirectly, with positive crosstalk to cells that expressed CXCL13. As shown in Figure 11, none of the glands presented with frank immunopathology, but as discussed in Section III, increasing levels of iNOS, IL-1, and IL-6 may come to be associated with physiological dysfunction.

1. A Core, Adaptive-Reactive Network

The abundances of mRNAs for IL-1α, IL-10, and CTLA-4 correlated significantly in all five groups (Supplemental Figure 1, Supplemental Tables 1–5). In V^61%,27°, V^68%,37°, V^72%,32°, and V^82%,29°, these transcripts sorted to correlation clusters that also included mRNAs for CD8, CTLA-4, IL-1β, and IL-6. In V^61%,27°, V^72%,32°, and V^82%,29°, the correlation cluster also included mRNAs for IL-2 and CCR5. The abundances of mRNAs for IL-17A and iNOS could be assayed only in V^72%,32° and V^82%,29°; both sorted to the cluster in both groups.

The transcripts in the cluster map to a novel network of diverse cells. IL-10 is classically expressed by T_H2 cells, but it also is expressed by T_R1 cells, NKT cells, and MMΦDC lineage cells. IL-1α, IL-1β, and IL-6 are classically expressed by MMΦDC lineage cells involved in innate responses; they were expressed primarily in immune cell accumulations (Figure 2), but the identities of the cell types that express them remain to be determined. IL-2 is classically a T_H1 cytokine, but IL-2 mRNA was expressed primarily in acinar cells (Figure 2). CTLA-4 is typically expressed by T cells. CD8 is expressed by T cells with conventional—i.e., hypervariable—TCR α-subunits with both effector and regulatory functions; it also may be expressed by NKT cells and MMΦDC lineage cells. CCR5 is the primary receptor for CCL4. IL-17A is typically expressed by eponymous T_H17 cells, which are positive for CD4,⁵⁴ but it also may be expressed by NKT cells and γδ T cells. iNOS can be expressed by various cell types, including, notably, activated MMΦDC lineage cells. As will be shown in Section II.D.1 and II.D.2, the abundance of each of the cluster transcripts can be described as having increased in association with increasing exposure to dryness, and the increases appear adaptive in that they are associated with decreased expression of IFN-γ mRNA. The network reacts to crosstalk from other cells and networks in certain settings. In other settings, it appears to engage with other cells and networks to form more complex networks. Therefore, for the purposes of this report, it will be referred to as the core, adaptive-reactive network.

2. Core, Adaptive-Reactive Network Transcripts in V^58%,17°

It is not clear that the core, adaptive-reactive network had formed in V^58%,17°. Figure 9 presents abundances of selected transcripts in V^58%,17°. The statistical power for V^58%,17° is quite low, but, as shown in Figure 4, the core, adaptive-reactive network transcripts appeared to sort to three smaller clusters: 1) mRNAs for IL-1β and IL-2 associated with mRNAs for CD80 and decorin (ρ_mean=.936, P_mean=.064); 2) mRNAs for IL-1α, IL-10, and CTLA-4 associated with mRNAs for CD28 and CD86 (ρ_mean=.933, P_mean=.067); and 3) IL-6 mRNA associated with IL-4 mRNA (ρ=.971, P=.029). Apparent cross-correlations between cluster 1 and cluster 2 (ρ_mean=.850, P_mean=.150) were not statistically significant; possible correlations between CD8 mRNA and the two clusters were, at best, weak (.591 ≥ρ≥.309); and the strongest of the possible associations for CD8 mRNA was with CD25 mRNA (ρ=.922, P=.0779).

Figure 9.

Relationships between abundances of selected transcripts and abundances of mRNAs for IL-10 and PRL in glands from V. G4. Significant associations with IL-10 mRNA are projected onto the back walls. Significant associations with PRL mRNA are projected onto the left side walls.

3. Additional Transcripts Associated with the Core, Adaptive-Reactive Network in V^82%,29°

Statistical powers for V^61%,27° and V^68%,37° are lower than for V^82%,29°. As will be discussed in Sections II.C.7 and II.C.8, analyses of transcript abundances in V^72%,32° glands, with gland V^72%,32°01.OD set aside as an apparent outlier, reveal that the glands sorted to two clusters, comprising, respectively, four glands and five glands, representing two phases of network development. The mean values of Pearson’s ρ and P statistics summarized in Table 1 identify several transcripts that were not likely to have been associated with the core, adaptive-reactive network in V^61%,27° or V^68%,37° but were significantly associated with it in V^82%,29°. Most notably, CD4 mRNA was associated with the network in V^82%,29°, but clearly not associated with it in V^61%,27° or V^68%,37°. mRNAs for CD25, MHC II, BAFF, CD28, PRL, IL-4, CCL2, CCL4, CCL21, and CD4 were significantly associated with the core, adaptive-reactive network in V^82%,29°. The data in Table 1 indicate that factors other than differences in the state of the core, adaptive-reactive network’s development accounted for most of the variations in the abundances of mRNAs for CD25 and CD28 in V^61%,27°, and for most of the variations in the abundances of mRNAs for MHC II, BAFF, PRL, IL-4, CCL2, and CD4 in V^68%,37.

Table 1.

Pearson’s Statistics for Associations with the Core Adaptive-Reactive Network Transcripts: ρ and (P)

Values presented are mean Pearson’s ρ (upper cell) and P (lower cell) statistics for associations between the indicated transcripts’ abundances and the abundances of the core adaptive-reactive network transcripts (mRNAs for IL-10, CD8, CTLA-4, IL-1α, IL-1β, IL-2, IL-6, and CCR5; values for CCR5 in V^68%,37° were excluded due to likely associations with additional networks in that setting).

mRNA	V61%,27° (n=5)	V68%,37° (n=4)	V82%,29° (n=12)
CD25	0.024 (0.901)	0.623 (0.377)	0.881 (0.002)
MHC II	0.668 (0.227)	0.202 (0.767)	0.790 (0.006)
BAFF	0.338 (0.582)	0.194 (0.720)	0.879 (0.001)
CD28	−0.065 (0.849)	0.478 (0.522)	0.745 (0.007)
PRL	0.476 (0.426)	0.072 (0.866)	0.716 (0.010)
IL-4	−0.602 (0.286)	0.070 (0.834)	0.745 (0.007)
CCL2	0.715 (0.179)	−0.155 (0.800)	0.788 (0.005)
CCL4	0.644 (0.246)	0.611 (0.389)	0.854 (0.002)
CCL21	−0.555 (0.337)	−0.576 (0.424)	0.665 (0.027)
CD4	−0.341 (0.578)	−0.033 (0.848)	0.829 (0.002)
Lipoph.CL	—	−0.289 (0.647)	−0.671 (0.023)
IL17A	—	—	0.914 (0.001)
iNOS	—	—	0.849 (0.002)

Association Significant

Bold Association Appears Unlikely

4. A T_H3-Reminiscent Cluster and Network Associations in V^61%,27°

Across the first five V^61%,27° glands, the numbers of cells positive for the T cell marker, RTLA, and the abundances of mRNAs for TGF-β₁, CD4, CD25, CD28, TNF-α, CD80, CD86, CCL21, CD3ε, CD3ζ, CXCL13, and decorin sorted to the same correlation cluster (Figure 10.A). Associations between RTLA⁺ cells and mRNAs for TGF-β₁, CD4, CD25, and CD28 are reminiscent of T_H3 cells. Associations between mRNAs for CD86, CD80, and TNF-α are reminiscent of MMΦDC lineage cells. Associations between mRNAs for CCL21, CD3ε, and CD3ζ are reminiscent of T cell inductive tissue, as CCL21 recruits naïve T cells,⁵⁵ and the CD3ε and CD3ζ subunits of the TCR signaling complex are highly expressed by naïve T cells.^56,57 CXCL13, typically expressed by MMΦDC lineage cells, recruits B cells^58,59; therefore, the association of CXCL13 mRNA with the T_H3-reminiscent cluster transcripts also is reminiscent of immune inductive tissue.

Figure 10.

Relationships between abundances of selected transcripts and abundances of mRNAs for CD4 and IL-10 in glands from V^61%,27°. Values for gland V^61%,27°02.OS ( ) were omitted from the initial regression analyses to minimize both type 1 and type 2 errors. Solid lines indicate range of X-axis values over which regressions were calculated, and dashed lines indicate projections. A. Associations with CD4 mRNA. The number of T cells (RTLA⁺ cells) and the abundance of CCL21 mRNA increased in approximately constant proportions across the entire sample of glands. Exponential regressions of IL-4 and MMP-9 mRNA abundances across the first five glands fell short of the P <.05 criterion for statistical significance. However, projections of both predicted values in gland V^61%,27°01.OS that were near the observed values, and regressions calculated across all six glands were highly significant (dotted lines). B. Lack of a significant association between the abundances of mRNAs for CD4 and IL-10 across the first five V^61%,27° glands. C. Associations with IL-10 mRNA. With the exception of CCR5 mRNA, the abundances of core, adaptive-reactive network transcripts in gland V^61%,27°02.OS fell within two standard errors of the proportions in the other V^61%,27° glands. This finding suggests that the network remained intact as it reacted to positive crosstalk associated with the immunopathological process. Abundances of mRNAs for CCL4, CCL2, CD25, CD28, BAFF, and numbers of bone marrow-derived cells—marked by CD18—were not significantly associated with the core, adaptive-reactive network transcripts across the first five glands but were significantly increased in gland V^61%,27°02.OS.

5. A T_H2-Reminiscent Profile in Gland V^61%,27°02.OS

Gland V^61%,27°02.OS stood apart from the other V^61%,27° glands in having large, pathological accumulations containing RTLA⁺ cells and CD18⁺ cells.⁴⁸ It also had higher abundances of mRNAs for CD4 and CCL21, but not of the other T_H3-reminiscent cluster transcripts; higher abundances of mRNAs for IL-4 and MMP-9 (Figure 10.A); and higher abundances of other transcripts, as well (Figure 10.B, discussed below). Although elevated, the number of RTLA⁺ cells and the abundances of mRNAs for CD4 and CCL21 were in the same proportions as in the other V^61%,27° glands. The abundances of mRNAs for IL-4 and MMP-9 could be described as having increased exponentially as the abundance of CD4 mRNA increased across the V^61%,27° glands. The association between mRNAs for IL-4 and CD4 is reminiscent of T_H2 cells. These findings imply that T_H2-reminiscent cells accumulated, and MMP-9 expression increased, concurrently with accumulation of T_H3-reminiscent cells, but with different kinetics.

Images of sections of V^58%,17°, V^61%,27°, V^68%,37°, and V^82%,29° glands stained for RTLA and CD18 have been presented by Mircheff et al. Representative images of V^72%,32° glands are presented in Figure 11. The transcript expression profile of gland V^61%,27°02.OS implicates cells, possibly of the MMΦDC lineage, expressing high levels of CCR5 mRNA, and T cells expressing high levels of mRNAs for CD4, CD25, and CD28 in the immunopathological process. Therefore, it is notable that more T cells, more CD18⁺ cells, and higher abundances of transcripts such as IL-4 mRNA, MMP-9 mRNA, and CCL21 mRNA were found in glands from other groups which were free of evident immunopathology (Supplemental Figures 2, 3, and 4).

Figure 11.

Representative sections of V^72%,32° glands stained for RTLA and CD18. Images of similarly stained sections of V^58%,17°, V^61%,27°, V^68%,37°, and V^82%,29° glands have been presented elsewhere. Immunopathological lesions are not evident in any gland, and gland V^72%,32°01.OD, which had higher abundances of numerous transcripts, had fewer than the median number of CD18⁺ cells and the fewest RTLA⁺ cells.

6. The Core, Adaptive-Reactive Network and the Inflammatory Process in Gland V^61%,27°02.OS

The plot of CD4 mRNA and IL-10 mRNA abundances in Figure 10.B indicates that the core, adaptive-reactive network developed independently of the accumulating T_H3-reminiscent cells and T_H2-reminiscent cells across the first five V^61%,27° glands. Like mRNAs for IL-4 and MMP-9, the core, adaptive-reactive network transcripts were considerably more abundant in gland V^61%,27°02.OS than the other V^61%,27°glands (Figure 10.B). With the exception of CCR5 mRNA, the core, adaptive-reactive network transcripts’ abundances remained in the same proportions in gland V^61%,27°02.OS as in the other V^61%,27° glands. Likewise, the abundances of most of the additional transcripts that associated with the core, adaptive-reactive network in V^82%,29° (Table 1) also were considerably higher in gland V^61%,27°02.OS than their V^61%,27° median values (Figure 10.B). These findings imply that the core, adaptive-reactive network remained intact in gland V^61%,27°02.OS, even though its activity had increased in reaction to crosstalk from the immunopathological process.

7. Crosstalk between Cells Expressing mRNAs for PRL, IL-4, IFN-γ, and IL-10

The abundance of IFN-γ mRNA increased exponentially as the abundance of PRL mRNA increased across the first five V^68%,37° glands (R²=0.993, P =.0003, Figure 12.A). This relationship accords with the pro-T_H1 influences that PRL exerts in other systems. However, in gland V^68%,37°02.OS, where PRL mRNA was most abundant, the abundance of IFN-γ mRNA was 100-fold less than predicted by the exponential growth relationship across the first five glands. Notably, the abundance of IL-4 mRNA increased exponentially as the abundance of PRL mRNA increased across the entire sample of glands (R²=0.996, P <.0001), such that it was highest in gland V^68%,37°02.OS. This relationship accords with reports that PRL supports expression of IL-4,^43,44 and it therefore suggests that epithelial cells which expressed PRL mRNA formed a network with T_H2-reminiscent cells which expressed IL-4 mRNA. The relationships also suggest that T_H1-reminiscent cells or NKT cells expressing IFN-γ mRNA (referred to as T_H1/NKT-reminiscent cells for purposes of this communication) received contradictory crosstalk from the PRL-mediated epithelial-T_H2-reminiscent network, i.e., positive crosstalk associated with expression of PRL mRNA and negative crosstalk associated with expression of IL-4 mRNA, and that the highest level of IL-4 mRNA expression was associated with abrogation of PRL-associated support for IFN-γ mRNA expression.

Figure 12.

Relationships between abundances of selected transcripts and abundances of mRNAs for PRL and IL-10 in glands from V^68%,37°. A. PRL mRNA. The regression of IFN-γ mRNA abundances calculated across the first four glands fell short of the P < .05 criterion, but the projection for gland V^68%,37°02.OS (⊗) was nearly identical to the observed value. The regression calculated through gland V^68%,37°02.OS was highly significant (dotted line). The regression of BAFF mRNA abundances calculated across all six glands was significant, but it obscures a significant negative association (ρ = −0.961, P = .0386) between the abundance of BAFF mRNA and the abundance of IL-4 mRNA across the first four glands. B. IL-10 mRNA. Abundances in glands V^68%,37°02.OS and V^68%,37°03.OS (⊕) were omitted from the regression analyses to minimize type 1 and type 2 errors. Values for gland V^68%,37°02.OS projected by regressions calculated over the first four glands were similar to the observed values for mRNAs for IL-1α, CTLA-4, CD8, and IL-6.

Positive and negative associations between the abundance of IFN-γ mRNA and the abundances of other transcripts also played out in V^72%,32°, where the abundance of IFN-γ mRNA reached higher levels than in any other group. The multiphasic relationship between the abundance of IFN-γ mRNA and the abundances of mRNAs for IL-10 and PRL in V^72%,32° (shown in Supplemental Figure 1) is illustrated on an expanded scale in Figure 13.A. As across the first five V^68%,37° glands, the abundances of IFN-γ mRNA and IL-4 mRNA increased with increasing abundance of PRL mRNA across the first four V^72%,32° glands; these increases may reflect development of the PRL-mediated epithelial-T_H2 network in association with accumulation of T_H1/NKT-reminiscent cells. The abundance of IFN-γ mRNA decreased as the abundance of IL-10 mRNA increased across the next five V^72%,32° glands (R²=0.944, P=.0012). The phase of decreasing IFN-γ mRNA abundances accords with the T_H1/T_H2 paradigm.⁶⁰ It also accords with the T_H1/IL-17 paradigm,^61,62 as IL-17A mRNA was associated with the core, adaptive-reactive network in V^72%,32°; but the association of IFN-γ mRNA with IL-10 mRNA was stronger (ρ=−0.929) than the association with IL-17A mRNA (ρ = −0.740). The abundance of IFN-γ mRNA in the tenth gland, V^72%,32°01.OD, was considerably higher than predicted by the inverse relationship; the unique transcript abundance profile of this gland is described in Section II.C.9.

Figure 13.

Relationships between the abundances of IFN-γ mRNA and the abundances of mRNAs for IL-10 and PRL in glands from groups V^61%,27° and V^72%,32°. Dark areas in projections onto rear and side walls indicate ranges of IL-10 mRNA abundances and IFN-g mRNA abundances over which regressions were calculated. A. Group V^72%,32°. The abundance of PRL mRNA was significantly related to the abundance of IL-10 mRNA across the first nine glands. The abundance of IFN-γ mRNA increased with increasing abundance of PRL mRNA across the first four glands and decreased with increasing abundance of IL-10 mRNA across the next five glands. B. Group V^61%,27°. The abundance of IFN-γ mRNA decreased with increasing abundance of IL-10 mRNA across the first five glands. The value the regression projected for gland V^61%,27°02.OS was similar to the observed value.

Also in accord with the T_H1/T_H2 paradigm, the abundance of IFN-γ mRNA decreased exponentially (R² = 0.994, P=.0061) as the abundance of IL-10 mRNA increased across the V^61%,27° glands (Figure 13.B). Moreover, IFN-γ mRNA was not detectable in glands from V^82%,29°, the group with the highest abundances of mRNAs for IL-10 and IL-17A.

8. Positive- and Negative Crosstalk to a T_H1/NKT-Reminiscent-TNF-α Network in V^72%,32°

The numbers of RTLA⁺ cells and the abundances of mRNAs for TNF-α, CCL2, CCL4, CD3ε, CD3ζ, IL-1RA, and IL-18R increased exponentially with increasing abundance of IFN-γ mRNA across first four V^72%,32° glands (Figure 14). Similar to IFN-γ, mRNA, the abundance of IL-18R mRNA decreased exponentially as the abundance of IL-10 mRNA increased across the next five glands. While the abundances of mRNAs for IFN-γ and IL-18R decreased, the abundances of mRNAs for CCL4, CD3ε, CD3ζ and IL-1RA remained associated with each other. This sequence of associations suggests that: 1) the T_H1/NKT reminiscent cells engaged in crosstalk with a network comprising epithelial cells expressing mRNAs for CCL2 and CCL4, MMΦDC lineage cells, and, perhaps, additional cell types, expressing mRNAs for TNF-α, CD80, CD86, IL-1RA and MMP-9; and 2) elements of this network persisted as negative crosstalk associated with increasing levels of IL-10 mRNA expression suppressed expression of mRNAs for IFN-γ and IL-18R

Figure 14.

Relationships between the abundances of selected transcripts and the abundances of mRNAs for IL-10 and IFN-γ in glands from group V^72%,32°. Increasing abundances of IFN-γ mRNA across the first four glands were associated with increases in the abundances of mRNAs for IL-18R, TNF-α, IL-1RA, CD3ε, CD3ζ, CCL4, and CCL2 and with increases in the number of T cells, marked by RTLA. Increasing abundances of IL-10 mRNA across the next five glands were consistent with increasing abundances of core, adaptive-reactive network transcripts, e.g., CD8 mRNA and associations with the abundances of mRNAs for iNOS, CD1d, and IL-17A, as well as with mRNAs for CD4,CD25, MHC II, IL-4, and BAFF. Abundances of transcripts in gland V^72%,32°01.OD were omitted from the regression analyses to reduce the possibility of type 1 or type 2 errors. Like the abundance of IFN-γ mRNA (Figure 13.A), the abundance of IL-17A mRNA in gland V^72%,32°01.OD, was considerably higher than projected by the linear regressions; this conclusion was confirmed by projections of multiple linear regressions. It may be noted that IL-17A mRNA abundances with gland V^72%,32°01.OD included are strongly and significantly described by three-parameter exponential growth regression. According to such a model, the high abundance of IL-17A mRNA in gland V^72%,32°01.OD would be a predictable development, rather than an emergent phenomenon.

Additional quantitative relationships indicate that the core, adaptive-reactive network received both positive crosstalk and negative crosstalk from the T_H1/NKT-reminiscent-TNF-α network during both phases of its development in V^72%,32°. While the abundances of mRNAs for IL-1α and IL-10 remained in similar proportions across V^58%,17°, V^61%,27°, V^68%,37°, and V^82%,29° (Supplemental Figure 1), the proportions were different in V^72%,32°, where IL-1α mRNA was considerably more abundant and IL-10 mRNA considerably less abundant. These findings imply that crosstalk from the T_H1/NKT-reminiscent-TNF-α network suppressed the core, adaptive-reactive network’s expression of IL-10 mRNA and augmented its expression of IL-1α mRNA.

9. Emergence of Novel Profiles in Glands V^68%,37°03.OS and V^72%,32°01.OD

Gland V^68%,37°03.OS, the V^68%,37° gland in which IFN-γ mRNA was most abundant, had a disproportionately high abundance of IL-10 mRNA (Figure 12.B). As noted above, a high abundance of IL-10 mRNA also coincided with high abundances of mRNAs for IFN-γ, PRL, and IL-4 mRNA in gland V^72%,32°01.OD (Figure 14). The abundances of most of the other core, adaptive-reactive network transcripts in both glands remained in similar proportions as in the other glands of their respective groups. Moreover, glands V^68%,37°03.OS and V^72%,32°01.OD each contained notably fewer T cells than the other glands of their respective groups.

Despite their qualitative similarities, the profiles differed quantitatively. Most notably, IL-10 mRNA was 5.4-fold more abundant in gland V^68%,37°03.OS, while IFN-γ mRNA was 13.5-fold more abundant in gland V^72%,32°01.OD. The quantitative differences suggest that some additional source of IL-10 mRNA expression had emerged in gland V^68%,37°03.OS and that some additional source of IFN-γ mRNA expression had emerged in gland V^72%,32°01.OD. While IL-17A mRNA was associated with the core, adaptive-reactive network across the first nine V^72%,32° glands, it was 1.9-fold more abundant than this association predicted in gland V^72%,32°01.OD; this discrepancy suggests that an additional source of IL-17A mRNA expression had emerged coincidently with the additional source of IFN-γ mRNA expression. Thus, the profile in gland V^72%,32°01.OD contradicts the T_H1/T_H17 paradigm as well as the T_H1/T_H2 paradigm; qualitatively; it appears reminiscent of IL-17⁺, IFN-γ⁺ γδ T cells found in mice.⁶³

10. BAFF mRNA-Expressing Cells, the PRL-Mediated Epithelial-T_H2 Network, and the PRL-Mediated Epithelial-T_H1 Network

BAFF is typically expressed by MMΦDC lineage cells in immune inductive tissues, where it promotes B cell activation and proliferation. The abundance of BAFF mRNA could, like the abundance of IL-4 mRNA, be described as having increased exponentially as the abundance of PRL mRNA increased across the V^68%,37° glands (Figure 12A). However, this description may obscure a more complex sequence of interactions. Across the first four glands, where the abundances of mRNAs for BAFF and IL-4 were near or below their median values, the abundance of BAFF mRNA was negatively associated with the abundance of IL-4 mRNA (ρ=−.961, P =.0386). These relationships imply that the PRL-mediated epithelial - T_H2 network engaged in crosstalk with BAFF-expressing MMΦDC lineage cells; negative crosstalk between the program for IL-4 mRNA expression and the program for BAFF mRNA expression played out across the lower range of PRL mRNA abundances, but the negative crosstalk was abrogated across the higher range of PRL mRNA abundances, and positive crosstalk emerged. As positive crosstalk influencing BAFF mRNA expression may have operated both in gland V^68%,37°03.OS and in gland V^68%,37°02.OS, it would not have depended on the negative crosstalk to T_H1/NKT-reminiscent cells that uniquely emerged in gland V^68%,37°02.OS.

11. Crosstalk between the Core, Adaptive-Reactive Network and Other Cells and Networks in Glands V^68%,37°02.OS and V^68%,37°03.OS

The core, adaptive-reactive network developed independently as the PRL-mediated epithelial-T_H2 network developed and T_H1/NKT-reminiscent cells accumulated across the first four V^68%,37° glands. The abundances of most of the core, adaptive-reactive network transcripts in glands V^68%,37°02.OS and V^68%,37°03.OS (except IL-10 mRNA, which, as noted above, was associated dually with the network and with an additional source in gland V^68%,37°03.OS) remained in proportions similar to the proportions in the other V^68%,37° glands (Figure 12.B). These associations imply that, like BAFF-expressing cells, the core, adaptive-reactive network engaged in positive crosstalk with the PRL-mediated epithelial - T_H2 network in glands V^68%,37°02.OS and V^68%,37°03.OS.

12. Correlation Clusters Reminiscent of Immune Inductive Tissues

The role BAFF plays in B cell inductive tissue was noted above. Likewise, associations of CCL21 mRNA with mRNAs for CD4, CXCL13, CD3ε, and CD3ζ in V^61%,27° were noted above to be reminiscent of immune inductive tissue. As also already noted, CCL21 mRNA was associated with mRNAs for CD3ζ, CCL28, and PRL in V^58%,17°. In V^68%,37°, CCL21 mRNA was associated with mRNAs for CD3ε and CD3ζ. In V^82%,29°, CCL21 mRNA was associated with mRNAs for CD3ε and CCL28, as well as with the core, adaptive-reactive network transcripts. In V^72%,32° (Supplemental Table 4, Figure 7), CCL21 mRNA sorted to a cluster that included mRNAs for IRF-1, IL-15, IL-18, LTβ, and E selectin, and this cluster engaged in crosstalk with the core, adaptive-reactive network, the T_H1/NKT-reminiscent-TNF-α network, and cells expressing VCAM-1 mRNA. These associations indicate that apparently healthy lacrimal glands may contain cells which are capable of recapitulating characteristic functions of organized immune inductive tissues.

D. Associations between Environmental Variables and Network Functions

1. Empirical Descriptions

The median abundance of IL-6 mRNA (associated with the core, adaptive-reactive network) can be empirically described as having increased exponentially with exposure to increasing dryness (R² = 0.997, P ≤.0001). This relationship is projected across the range of temperature exposures in Figure 15.A. With the caveat that data are not available for IL-18 mRNA in V^58%,17° nor for lipophilin CL mRNA in V^61%,27°, the median abundances of mRNAs for CCL21, IL-18, MMP-9, CCL28, and lipophilin CL can be empirically described as having increased exponentially with increasing exposure to increasing temperature (0.822 ≤ R² ≤ 1.000, 0.0337 ≥ P ≥.0005). These relationships are projected across the range of dryness exposures in Figure 15.B.

Figure 15.

Relationships between abundances of selected transcripts and environmental dryness and high temperature. Median abundances and abundances in individual glands that appeared as outliers are presented. A. Abundances of transcript that could be described as changing with increasing exposure to dryness. B. Abundances of transcripts that could be described as changing with exposure to increasing temperature. Continuous surfaces represent the domains over which median abundances could be described as conforming to empirical exponential growth curves or heuristics. Discontinuities represent domains in which surfaces could not be projected because median abundances in one or more groups were augmented above or displaced below the values predicted by their respective heuristics. Smaller font indicates transcripts that behaved similarly to the plotted transcripts.

2. Exponential Growth Heuristics and Crosstalk

The exponential growth functions also are useful heuristics for describing relationships between transcripts’ abundances and the environmental variables that were not monotonic. Thus, abundances can be described as having been displaced from their heuristics’ predictions in one group or another. Most such cases coincided with the emergence of crosstalk between cells or networks. For example, the median abundance of PRL mRNA increased according to an exponential temperature heuristic, but it was augmented above the heuristic’s prediction in V^58%,17° (Figure 15.B, Supplemental Figure 3). The median abundance of CD8 mRNA increased according to an exponential dryness heuristic, but, like the abundance of PRL mRNA, it was augmented above its heuristic’s prediction in V^58%,17° (Figure 15.A, Supplemental Figure 2). The augmentation of PRL mRNA expression in V^58%,17° coincided with the association of PRL mRNA in the correlation cluster that also comprised mRNAs for CD3ζ, CCL21, and CCL28 (Figure 4). The augmentation of CD8 mRNA expression coincided with the possible association of CD8 mRNA with CD25 mRNA in a network independent of the core, adaptive-reactive network (Section II.C.2).

Although median abundances of CCL21 mRNA could be empirically described as having increased exponentially with exposure to increasing temperature (Figure 15.B, Supplemental Figure 3), the value in V^61%,27° fell perceptibly above the regression, and the value in V^82%,29° fell perceptibly below the regression. Augmentation of CCL21 mRNA expression in V^61%,27° coincided with an association with the T_H3-reminiscent correlation cluster (Sections II.C.4 – II.C.5, Figure 5). Suppression of CCL21 mRNA in V^82%,29° coincided with an interplay between positive crosstalk associated with the core, adaptive-reactive network and negative crosstalk associated with cells expressing mRNAs for IL-15 and IL-18.

The median abundances of mRNAs for IL-4, BAFF, and APRIL increased according to dryness heuristics but can be described as having been augmented above their heuristics’ predictions in V^68%,37°; IL-4 mRNA can be described as also having been augmented above its heuristic’s prediction in V^72%,32° (Figure 15.A, Supplemental Figure 2). The augmentation of the median abundances of mRNAs for IL-4 and BAFF in V^68%,37° coincided with the high abundance of PRL mRNA (associated with the high temperature V^68%,37° had experienced) and development of the PRL-mediated epithelial - T_H2 network (Sections II.C.7 and II.C.10; Figure 6).

The median abundance of IL-10 mRNA increased according to an exponential dryness heuristic, but it was suppressed below its heuristic’s prediction in V^72%,32. In contrast, the median abundances of mRNAs for CTLA-4, IL-1α, IL-1β, IL-2, CCL2, CCL4, MHC II, and TGF-β², and the median number of RTLA⁺ cells also increased according to exponential dryness heuristics, but, like IL-4 mRNA, all were augmented above their heuristics’ predictions in V^72%,32°° (Figure 15.A, Supplemental Figure 2). The median abundances of mRNAs for CD3ε, CD3ζ, and CXCL8 increased according to exponential temperature heuristics, but they, also, were augmented above their heuristics’ predictions in V^72%,32° (Figure 15.B, Supplemental Figure 3). The median abundance of IFN-γ mRNA decreased with increasing dryness according to an exponential decay heuristic, but it was augmented markedly above its heuristic’s prediction in V^72%,32° (Figure 15.A, Supplemental Figure 2). The displacements of the core, adaptive-reactive network transcripts and associated transcripts, IFN-γ mRNA, and mRNAs for CD3ε, CD3ζ, and CXCL8 coincided with crosstalk between the core, adaptive-reactive network, the PRL-mediated epithelial - T_H2 network, and the T_H1/NKT-reminiscent - TNF-α network (Section II.C.7 – II.C.8, Figure 7).

III. HOW DETERMINED, ADAPTIVE NETWORK RESPONSES MAY LEAD TO DIVERSE IMMUNOPATHOLOGICAL STATES

The new findings and analyses reported above introduce the concept that, at least by early adult life, each lacrimal gland has become a dynamic system comprising immune cells that engage in crosstalk with each other and with epithelial cells under the influences of physiological signals related to environmental dryness, high temperature, and perhaps also low temperature. As will be discussed in this section, several of the networks may have adaptive value but also may be antecedents of different types of chronic immunopathological infiltrate, with diverse impacts on the glands’ physiological function.

A. Possible Adaptive Value of the Core Network and of T_H2-Reminiscent Cells

Both the core, adaptive-reactive network and at least one of the cell types (T_H2-reminiscent cells) with which it engages in positive crosstalk may have adaptive value in opposing T_H1 cell-mediated autoimmune responses. Certain regulatory MMΦDC lineage cells exert their influences through IL-4, which suppresses IFN-γ expression. IL-10 and IL-17A also suppress IFN-γ expression. IL-10 is, like IL-4, expressed by T_H2 cells, and it also is expressed by T_R1 cells, regulatory B cells, and regulatory MMΦDC cells. IL-17A is commonly associated with inflammatory pathogenesis, as it increases expression of IL-1 and of various chemokines, but it also is well known to suppress IFN-γ expression. Foxp3⁺ cells, presumed to be regulatory T cells, are present in murine lacrimal glands,⁶⁴ but their phenotypes have not yet been characterized. In gland V^68%,37°02.OS, a high level of IL-4 mRNA expression was associated with abrogation of PRL-associated support for IFN-γ mRNA expression. The core, adaptive-reactive network accounted for most of the IL-10 mRNA expression in V^61%,27°, V^68%,37°, V^72%,32°, and V^82%,29° and also for most of the IL-17A mRNA expression in V^72%,32° and V^82%,29°. Increasing abundances of IL-10 mRNA were significantly associated with decreasing abundances of IFN-γ mRNA in V^61%,27° and in a subset of the V^72%,32° glands. Moreover, IFN-γ mRNA was not detectable in V^82%,29°, where median abundances of mRNAs for IL-10 and IL-17A were highest.

1. Counter-regulatory, Anticipatory, or Corollary?

It would be plausible to hypothesize that the core, adaptive-reactive network developed as a counter-regulatory response to desiccation-induced inflammation of the cornea and conjunctiva. It appears that T_H1 cells effect the transformation from an innate inflammatory ocular surface response to an autoimmune process, and autoreactive, pathogenic T_H1 cells can be detected in the draining lymph nodes. If such cells traffic to the lacrimal glands, they would have the potential to propagate the inflammatory process.⁶⁵ The theoretical prediction that they provoke a counterregulatory activation of the core, adaptive-reactive network is consistent with the finding that the network engages in crosstalk with other cells and networks in other settings. However, if the activity of the core, adaptive-reactive network in V^82%,29° had increased reactively, in response to the arrival of increasing numbers of T_H1 cells, it appears to have maintained itself and developed further even after it had effectively suppressed T_H1 activity.

3. Physiological Signals Related to Dryness

It also would be plausible to hypothesize that the core, adaptive-reactive network is activated by autonomic secretomotor neurotransmission triggered by activation of cold receptors or nocioceptors in the cornea and conjunctiva. Findings from preliminary experiments with an ex vivo acinar cell model are consistent with a neurally-mediated mechanism, as the β-adrenergic receptor agonist, isoproterenol, increases expression of mRNAs for CCL2 and CCL4,⁶⁶ both of which recruit both T cells and MMΦDC lineage cells. The findings in Figure 2 both confirm that epithelial cells express mRNAs for CCL2 and CCL4 in vivo and also indicate that acinar cells are the predominant cell type expressing CCL2 mRNA. Like the core, adaptive-reactive network transcripts, the abundances of mRNAs for CCL2 and CCL4 could be described as having increased according to exponential dryness heuristics. They were significantly associated with the core, adaptive-reactive network transcripts in V^82%,29°, and hypothetical associations with the network in other settings cannot be rejected (Table 1). Moreover, mRNA for the primary CCL4 receptor, CCR5, is one of the core, adaptive-reactive network transcripts.

In view of findings suggesting that meibomian gland dysfunction leading to increased evaporation and tear fluid hyperosmolarity frequently precedes signs of aqueous insufficiency,^10–13 it should be noted that cold receptors would be activated by the heat loss of evaporation,⁶⁷ thus eliciting a response that is anticipatory rather than counter-regulatory. However, a neurally-mediated mechanism can be predicted to also involve negative crosstalk, as the muscarinic cholinergic receptor agonist, carbachol, attenuates, but does not entirely abrogate, the ability of isoproterenol to increase expression of mRNAs for CCL2 and CCL4.⁶⁶ The median abundances of mRNAs for CCL2 and CCL4 were elevated above their heuristics’ predictions in V^72%,32°, and the abundances of both also were elevated in glands V^61%,27°02.OS, V^68%,37°02.OS, and V^72%,32°01.OD—three of the four glands that appeared as outliers. Therefore, it would be plausible to predict that epithelial cells expressing CCL2 and CCL4 contributed to the crosstalk which promoted development of the core adaptive-reactive network in each setting.

3. Physiological Signals Related to High Temperature

It is not a straightforward task to formulate hypotheses for how physiological signals related to high temperature might elicit increases in the expression of such transcripts as PRL mRNA, CCL21 mRNA, IL-18 mRNA, and lipophilin CL mRNA. The abundances of the various high temperature-responsive transcripts appeared to increase concurrently with each other, but not coordinately. The signals that elicit these increases might include hormones associated the physiological heat stress response, most notably PRL.⁶⁸ Of the transcripts that became more abundant in association with exposure to increasingly high temperature, only mRNAs for PRL and lipophilin CL have so far been clearly localized to epithelial cells (Figure 2). However, our preliminary studies with ex vivo models have not shown that ambient PRL levels consistently influence expression of either transcript (unpublished data). The failure to demonstrate a consistent influence in ex vivo models does not necessarily preclude a role for hormonal prolactin, however, as responses may be mediated by complex local signaling interactions not yet reconstituted in the models. A plausible-seeming hypothesis, yet to be tested, is that hormonal PRL influences resident immune cells that engage in crosstalk with the epithelial cells and lead the epithelial cell to up-regulate expression of PRL and lipophilin CL.

B. T_H3-Reminscent Cells and Constitutive, Adaptive Immunoregulation

In addition to T_H2-reminiscent cells and IL-10-expressing cells of the core adaptive-reactive network, the T_H3-reminiscent cells (whose signature correlation cluster was detected in V^61%,27° and V^68%,37°) also can be predicted to exert regulatory influences in the lacrimal gland. The abundance of TGF-β₁ mRNA did not vary in a simple relationship with dryness or temperature across the five groups of rabbits. As variations in the numbers of other cells (T_H1/NKT-reminiscent cells in V^72%,32° and T_H2-reminiscent cells in V^82%,29°) accounted for most of the variation in CD4 mRNA abundance, the failure to detect the T_H3-reminiscent signature in other groups cannot be taken to imply that T_H3-reminiscent cells were absent from group V^72%,32° and group V^82%,29°. Rather, it is more likely that T_H3-reminiscent cells were present constitutively.

C. Crosstalk and Inflammatory Pathogenesis

Crosstalk between the core, adaptive-reactive network and other cells and networks was associated with obvious immunopathological changes in gland V^61%,27°02.OS. Interestingly, the pathological process appeared not to involve IFN-γ, as IFN-γ mRNA expression was suppressed in association with the high level of IL-10 mRNA expression in that gland (Figure 13.B). The relationships between the numbers of RTLA⁺ cells and abundances of mRNAs for TGF-β₁, IL-4, and CD4 across the V^61%,27° glands (Figure 10) support estimates that T_H3-reminiscent cells and T_H2-reminiscent cells accounted for roughly similar numbers of T cells in gland V^61%,27°02.OS. In contrast, T cells associated with the core, adaptive-reactive network can have accounted for only a small proportion of the total number of T cells in gland V^61%,27°02.OS. It is plausible to hypothesize that ongoing, exponential accumulation of T_H2-reminiscent cells drove development of the pathological process. The competing hypothesis, that some adventitious event, such as infection or trauma triggered the process, also deserves consideration. In either case, it also is plausible that the incipient process elicited a reactive increase in the activity of the core, adaptive-reactive network, and that the network, in turn, contributed mitogenic cytokines (IL-1α, IL-1β, IL-2, and IL-6) that accelerated the process.

D. Crosstalk Promoting Elements of Ectopic Immune Inductive Tissue

Gland V^68%,37°02.OS and gland V^68%,37°03.OS (the two V^68%,37° glands in which the core, adaptive-reactive network engaged in crosstalk with the PRL-mediated epithelial - T_H2- reminiscent network) were the same glands in which negative crosstalk associated with expression of mRNAs for IL-4 and BAFF was abrogated and replaced by positive crosstalk (Figure 12.A). Neither gland presented with evident pathological changes, but features of their transcript expression profiles suggest that they had accumulated several elements of ectopic immune inductive tissues such as characterize Sjögren’s disease. As noted, the products of several of the core, adaptive-reactive network transcripts are mitogenic; importantly, IL-2, IL-6, and IL-10, like IL-4 and BAFF, are associated with B cell activation. Like PRL mRNA, the abundances of two additional transcripts whose products have been implicated in Sjögren pathophysiology (CCL21 and IL-18) increased with increasing exposure to high temperatures and were highest in V^68%,37°. The role CCL21 plays in recruiting T cells to immune inductive tissues has already been noted. IL-18 may abrogate the concerted influences of IL-6 and TGF-β in promotion of B cell IgM-to-IgA class switching, a normal phenomenon in mucosal immune effector tissues,⁶⁹ and favor IgM-to-IgG class switching, a phenomenon associated with Sjögren’s- and Mikulicz’s diseases.

An additional mediator known to be implicated in Sjögren’s immune inductive tissues is CXCL13, which recruits B cells. CXCL13 mRNA was detected in all glands in which it was assayed. However, it was no more abundant in V^68%,37° glands than in glands from V^72%,32° (Supplemental Figure 3) and, moreover, across the V^68%,37° glands, it appeared subject to negative crosstalk associated with increasing CCL21 expression (Figure 6). This finding suggests that if potential antecedents of Sjögren’s ectopic immune inductive tissues were present gland V^68%,37°02.OS and gland V^68%,37°03.OS, they were rather primitive.

E. Crosstalk Promoting IL-17-Mediated Inflammation

In V^72%,32°, the core, adaptive-reactive network appeared as the recipient of positive and negative crosstalk from the T_H1/NKT-reminiscent-TNF-α network, and both forms of crosstalk may have been modulated by crosstalk from other elements of the complex network that emerged in this setting. The negative crosstalk was evident in the suppression of IL-10 mRNA expression, and the positive crosstalk was evident in the augmentation of expression of IL-1α mRNA and other core and extended network transcripts. Perhaps surprisingly, given the relatively high abundances of mRNAs for CCL2, CCL4, IFN-γ, IL-1α, IL-1β, IL-6, IL-17A, and iNOS in V^72%,32° glands (Figure 15.A, Supplemental Figure 2), none presented with obvious pathological changes (Figure 11). However, it seems straightforward to predict that inflammatory lesions would arise during extended exposure to an environmental setting similar to that experienced by V^72%,32°, either as the network that formed continues to develop, or as a population of cells expressing high levels of IFN-γ in a setting of high IL-17A expression (features of the profile of gland V^72%,32°01.OD) emerges.

F. Physiological Dysfunction without Frank Inflammation: Stochastic Processes and Determined Responses

It also is plausible to predict that physiological dysfunction would arise in the absence of frank pathological lesions as the core, adaptive-reactive network continues to develop during extended exposure to an environmental setting such as group V^82%,29° experienced, i.e., a setting where expression of mRNAs for IFN-γ and CCL21 is suppressed. This prediction stems from published findings that the products of several core, adaptive-reactive network transcripts (NO,¹⁹ IL-1α, IL-1β, and IL-6^20,21) cause physiological dysfunction in ex vivo models.

A simple model for projecting how the core, adaptive-reactive network would develop over time (absent crosstalk from other cells or networks) arises from the exponential growth functions and heuristics that describe relationships between network transcript abundances and environmental dryness. The 30d mean dryness, $\overline{\mathrm{\Delta }\%\mathrm{H}}$, can be written as

$$\sum _{\mathrm{d}=1-30}\mathrm{\Delta }\%{\mathrm{H}}_{\mathrm{d}}÷30\mathrm{d}=\frac{\overline{\mathrm{\Delta }\%\mathrm{H}}}{30\mathrm{d}}·30\mathrm{d}$$

Therefore, the two-parameter exponential growth equation,

$$[\text{mRNA}]={[\text{mRNA}]}_0·{\mathrm{e}}^{\mathrm{b}·\overline{\mathrm{\Delta }\%\mathrm{H}}}$$

can be rewritten,

$$[\text{mRNA}]={[\text{mRNA}]}_0·{\mathrm{e}}^{\frac{\mathrm{b}}{30\mathrm{d}}·\overline{\mathrm{\Delta }\%\mathrm{H}}·30\mathrm{d}}$$

The term b is determined empirically, by the nonlinear regression algorithm. Therefore, using the term,

$$\frac{\mathrm{b}}{30\mathrm{d}}$$

as an empirically determined constant allows one to project values of [mRNA] after varying numbers of days, nd, at varying at values of $\overline{\mathrm{\Delta }\%\mathrm{H}}$, i.e.,

$$[\text{mRNA}]={[\text{mRNA}]}_0·{\mathrm{e}}^{\frac{\mathrm{b}}{30\mathrm{d}}·\overline{\mathrm{\Delta }\%\mathrm{H}}·{\mathrm{n}}_{\mathrm{d}}}$$

Figure 16.A projects the median abundance of IL-6 mRNA over time for each of the values of $\overline{\mathrm{\Delta }\%\mathrm{H}}$the groups in this study had experienced. According to these projections, the median abundance after 50 days at the dryness experienced by V^61%,27° would be greater than after 30 days at the dryness experienced by V^82%,29°. This simple model implies that the core, adaptive-reactive network develops inexorably in a wide range of environmental settings.

Figure 16.

Theoretical projections of IL-6 mRNA abundance changes over time. (A) Projections of median abundances at the values of $\overline{\mathrm{\Delta }\%\mathrm{H}}$ experienced by each of the five groups. (B) Projections of the abundances in the individual V^82%,29°.OS glands at the value of $\overline{\mathrm{\Delta }\%\mathrm{H}}$ that the V^82%,29° animals had experienced. The projections assumed that values of [mRNA]₀ had varied stochastically among the individual glands and the value of $\frac{\mathrm{b}}{30\mathrm{d}}$ remained constant.

If the core, adaptive-reactive network develops inexorably, the wide ranges of variability across the individual glands from each of the groups studied indicate that it must develop according to different kinetics in different subjects and in the OS- and OD glands of each subject. The simple model points to two theoretical explanations for stochastic variations in the core, adaptive-reactive network’s kinetics: the term b may be a variable, or the term [mRNA]₀ may be a variable. Figure 16.B projects the abundance in each of the V^82%,29°OS glands based on values of [mRNA]₀ calculated by solving for each of the observed values of [mRNA]. This theoretical explanation implies that stochastic variations in the numbers of core, adaptive-reactive network cells present in a gland early in life would determine the network’s kinetics throughout life.

IV. CONCLUSIONS

The present findings expand the repertoire of mediator transcripts found to be expressed by epithelial cells in rabbit lacrimal glands to include mRNAs for IL-2, IL-1RA, APRIL, CCL2, and CCL4, as well as mRNAs for TGF-β₂ and PRL; they confirm that mRNA for CCL28 (typically expressed by epithelial cells in mucosal immune effector tissues) also is expressed in the lacrimal gland.

The findings also substantiate the hypothesis that immune cells in rabbit lacrimal glands form networks with each other and with epithelial cells. A core, adaptive-reactive network developed with increasing exposure to dryness; it was evident in glands from all settings except the coolest and least dry (V^58%,17°). Numbers or activities of cell types expressing mRNAs for PRL, CCL21, IL-18, and several additional transcripts increased with exposure to increasingly high temperatures; the increases appeared to be concurrent, rather than coordinate, and variations in the abundances of these transcripts were associated with different networks in different settings.

Expression of each of the epithelial transcripts was associated with crosstalk to and from immune cells and immune cell networks in one setting or another. IL-2 was consistently associated with the core, reactive-adaptive network, and CCL2, CCL4, and PRL clearly were associated with it in a setting where it developed to high levels of activity (V^82%,29°, Figure 8, Figure 15.A). The core, adaptive-reactive network expresses IL-10, and, as discussed in Section III, in several settings it was associated with negative crosstalk to T_H1-reminiscent cells expressing IFN-γ. In other settings, PRL was implicated in positive crosstalk to T_H1-reminiscent cells; to T_H2-reminiscent cells expressing IL-4; and to MMΦDC lineage cells expressing BAFF. In a dry setting (V^82%,29°), development of the core, adaptive-reactive network and negative crosstalk suppressing IFN-γ expression may lead to a state of lacrimal physiological dysfunction independent of frank immunopathology. In other settings, the interplay between positive crosstalk associated with PRL-; positive crosstalk associated with other activities of the core, adaptive-reactive network-; and positive- and negative crosstalk associated with T_H2 cells and BAFF-expressing cells can promote the accumulation of several of the elements of Sjögren’s ectopic immune inductive tissues (V^68%,37°, Figure 6, Figure 12) or to a state of chronic, inflammatory pathology associated with severe physiological dysfunction (V^72%,32°, Figure 7, Figure 14). The concept that different immunopathological processes arise over time, as stochastic events interact with determined system responses to physiological signals, leads one to predict that pathways to additional, immunopathologically distinct states will emerge in other natural settings.

The findings presented herein support the notion that T_H3 cells, expressing TGF-β₁, are present in the rabbit lacrimal gland; that the T_H3 cells form networks with antigen presenting cells, expressing CD80 and CD86; and that they engage in different forms of crosstalk with other cells and networks in different settings, e.g., with T_H2 cells in V^61%,27° and V^82%,29° and with T_H1 cells in V^72%,32°.

If each lacrimal gland is a system that comprises immune cells and networks that are potential antecedents of immunopathological infiltrates, the distinction between immunoregulation and inflammation must be much more richly nuanced than previously contemplated. The predicted pathways to immunopathological states involve changes in the behaviors of cells and networks vis-à-vis autoantigens that may have been constantly surveilled, and the changes are driven by determined network responses to normal, physiological signals. None of the predicted pathways invokes events, such as infection or ocular surface trauma, that often are thought to trigger losses of tolerance to autoantigens. Such triggers include activation of toll-like receptors.⁷⁰, development of adaptive immune responses to pathogen antigens that mimic autoantigens,⁷¹ generation of previously subdominant autoantigen epitopes not subject to central or peripheral tolerance, ²⁹ or direct, MHC II-mediated presentation of autoantigen epitopes by epithelial cells acting as surrogate antigen presenting cells.⁷² The findings do not exclude such events. To the contrary, conceiving of the lacrimal gland as a dynamic system may provide a framework for understanding how such events impact the gland’s immunophysiological state.

None of the predicted pathways invokes changes in steroid reproductive hormone levels (associated with the female life cycle, the female reproductive cycle, or the male life cycle) to tip the balance from regulatory activity to inflammatory activity. As noted, the hormonal milieu of pregnancy is associated with changes in the expression of numerous transcripts.⁵¹ Analyses that will be described in a report being prepared for publication indicate that many of those changes reflect complex changes in interactions within the core, adaptive-reactive network and between the core network and other cells and networks. In some cases, the crosstalk leads to transcript expression profiles reminiscent of those that emerged in glands V^68%,37°02.OS and gland V^68%,37°03.OS, described in Sections III.D as having features reminiscent of Sjögren’s immune inductive tissues. Moreover, many hormones in addition to prolactin and the steroid reproductive hormones, including cortisol, which mediates the physiological cold stress response,⁷³ may prove to impact immune cell- and network activities in the lacrimal gland. Thus, one can now reasonably expect that the immunophysiological state of a lacrimal gland at any moment will depend on the subject’s cumulative history of hormonal states interacting with physiological signals related to environmental exposure.

The pathways that have been predicted in this report are biologically plausible. The immunopathology of Sjögren’s disease has been described in numerous studies of labial salivary glands of patients with Sjögren’s syndrome. In contrast, the other chronic immunopathological states predicted to evolve in lacrimal glands are described with more molecular and mechanistic specificity than has yet been documented by clinical or experimental studies. It would be worthwhile to begin work to validate, or to reject and revise, all the predictions, as the findings can be expected to lead to improvements in clinical practices. Different immunopathological processes will generate different biomarker signatures, and an understanding of how those processes develop should make it possible to diagnose them differentially, in their incipient stages. Likewise, different processes would present different targets of therapeutic opportunity, and an understanding of those processes may lead to identification of the most opportune targets.

Experiments to test the predictions can be done with animal models, but it is not yet clear which model will prove to be most informative of how immune cells and networks function in human lacrimal glands. Differences between the histoarchitectures of rodent lacrimal glands and rabbit- and human lacrimal glands have been described, as have histoarchitectural similarities between rabbit and human lacrimal glands.⁷⁴ It would be straightforward to design studies comparing network functions in different animal models. Discerning how immune cells and networks function in human lacrimal glands may prove to be a formidable challenge.

Supplementary Material

Fig S1. Supplemental Figure 1.

Overview of relationships between abundances of mRNAs for IFN-γ, IL-4, and IL-1α and abundances of mRNAs for IL-10 and PRL across the glands from each group. mRNAs for IL-17A, iNOS, and CD1d were assayed in glands from V^72%,32° and V^82%,29°; their abundances also are presented.

Fig S2. Supplemental Figure 2.

Transcript abundances that could be described as having changed with increasing exposure to environmental dryness or as changed according to dryness heuristics. Numbers of T cells, marked by RTLA, also could be described as having increased according to an exponential dryness heuristic. Circles, diamonds, and squares represent values in individual glands; horizontal bars represent group median values. Arrows indicate median values that can be described as having been augmented above or suppressed below the values predicted by their heuristics. mRNAs for IL-17A, iNOS, IL-15, IL-1R, and CD1d could be assayed only in V^72%,32° and V^82%,29°, as samples from other groups were exhausted before gene sequences became available, and relationships more complex than depicted by the heuristics cannot be excluded.

Fig S3. Supplemental Figure 3.

Transcript abundances that could be described as having increased with exposure to increasing temperature or as having increased according to temperature heuristics. Details are as in the legend to Supplemental Figure 2. In the case of CXCL13 mRNA, the heuristic assuming suppression in V^68%,37° was selected as consistent with negative crosstalk associated with high levels of CCL21 mRNA expression (Figure 6).

Fig S4. Supplemental Figure 4.

Transcript abundances that could not be described as having changed according to simple dryness- or temperature heuristics. Total numbers of bone marrow-derived cells, marked by CD18, also could not be described as having changed according to a simple heuristic, but, like the numbers of T cells and abundances of many transcripts, were highest in V^72%,32° glands. mRNAs for CD36, IL-18R, E-selectin, VCAM-1, MHC I, IRF-1, and LTβ could be assayed only in V^72%,32° and V^82%,29°, as samples from other groups were exhausted before gene sequences became available. The relationships with T would be consistent either with temperature heuristics or with dryness heuristics positing augmentation in V^72%,32°; more complex relationships are not excluded.

Tab S1. Supplemental Table 1.

Pearson’s statistics ρ (upper cell) and P (middle cell) and values n (bottom cell) for correlations between transcript abundances across the V^58%,17° glands. Cells containing values of P ≤.0500 are shaded to highlight significant correlations. Where possible, rows and columns were arranged to emphasize correlation clusters. The blank diagonal depicts identities such that values below the diagonal mirror values above the diagonal.

Tab S2. Supplemental Table 2.

Pearson’s statistics for correlations between transcript abundances across the V^61%,27° glands. Details are as in the legend to Supplemental Table 1. Calculated with gland V^61%,27°02.OS were omitted to avoid type 1 or type 2 errors.

Tab S3. Supplemental Table 3.

Pearson’s statistics for correlations between transcript abundances across the V^68%,37° glands. Details are as in the legend to Supplemental Table 1. Glands V^68%,37°02.OS and V^68%,37°03.OS were omitted to avoid type 1 or type 2 errors. Non-linear relationships and emergent phenomena contributing to the transcript abundance profiles in those glands are presented in Figure 12.

Tab S4a. Supplemental Table 4.

Pearson’s statistics for correlations between transcript abundances across the V^72%,32° glands. Details are as in the legend to Supplemental Table 1. A. Calculated across the first four glands, in order of increasing IL-10 mRNA abundance, to capture correlations in the initial state (see also Figures 13.A and 14). B. Calculated across glands five through nine, in order of increasing IL-10 mRNA abundance, to capture correlations in the state which emerged when increasing abundance of IL-10 mRNA came to abrogate PRL-associated support for IFN-γ expression. C. Calculated with gland V^72%,32°01.OD omitted to avoid type 1 or type 2 errors.

Tab S5. Supplemental Table 5.

Pearson’s statistics for correlations between transcript abundances across the V^82%,29° glands. Details are as in the legend to Supplemental Table 1.

GLOSSARY

APRIL

(A Proliferation-Inducing Ligand, TNFSF-13). Expressed by lacrimal gland acinar- and ductal epithelial cells (this study). Mitogenic cytokine supporting T cell- and B cell survival.⁷⁵

BAFF

(B Cell Activating Factor, TNFSF-13C). Classically expressed by follicular dendritic cells. Expressed by lacrimal gland immune cells (this study). Mitogenic cytokine promoting B cell activation and proliferation

Caspase-1

(Interleukin-Converting Enzyme). Cleaves the immature forms of IL-1β and IL-18 to generate the mature, catalytically active forms

CCL2

(Monocyte Chemotactic Protein 1, MCP-1). Recruits memory T cells and MMΦDC lineage cells

CCL4

(Macrophage Inflammatory Protein 1beta, MIP-1b). Recruits T cells, including regulatory T cells.⁷⁶

CCL21

Classically expressed by stromal reticular fibroblasts. Also expressed by renal mesangial cells in humans,⁷⁷ and by vascular endothelial cells in mice.⁷⁸ Recruits naïve T cells to lymphoid follicles. ^55,77

CCL28

(Mucosa-associated Epithelial Chemokine, MEC). Classically expressed by mucosal epithelial cells. Recruits IgA⁺ plasmablasts to mucosal immune effector sites. ^79–82

CCR5

Classically expressed by MMΦDC lineage cells. Receptor for CCL4

CD1d

Major histocompatibility complex molecule. Presents glycolipid, galactosylceramide, to invariant T cell receptor a-subunit expressed by NKT cells

CD3ε

Epsilon-subunit of the signal-transducing complex that associates with activated T cell antigen receptors

CD3ζ

Zeta-subunit of the signal-transducing complex that associates with activated T cell antigen receptors. CD3ζ, in particular, is highly expressed in naïve T cells and down-regulated upon activation

CD4

Classically expressed by T cells; also may be expressed NKT cells. Complexes with T cell antigen receptors to recognize MHC II-epitope complex.

CD8

Classically expressed by cytotoxic effector T cells; also may be expressed NKT cells and regulatory T cells, as well as by MMΦDC lineage cells. Complexes with T cell antigen receptors to recognize MHC I-epitope complex

CD18

Classically expressed by bone marrow-derived cells

CD25

(IL-2 receptor α-subunit). Classically expressed by activated T cells; expressed at high levels by regulatory T cells; mediates mitogenic signal that abrogates proapoptotic actions of TGF-β.

CD28

Classically expressed by T cells. Co-receptor for CD86

CD36

Expressed by various cells, including epithelial cells and other parenchymal cells. Functions as a scavenger receptor include uptake of oxidized lipids and apoptotic cells leading to cross-presentation of epitopes to CD8⁺ cells.⁸³

CD80

Classically expressed by MMΦDC lineage cells. Co-receptor for CTLA-4. Mediates co-stimulatory signal required for activation of T cells in conjunction with MHC-molecule-mediated epitope presentation

CD86

Classically expressed by MMΦDC lineage cells. Co-receptor for CD28. Mediates co-stimulatory signal required for activation of T cells in conjunction with MHC molecule-mediated epitope presentation

CTLA-4

Classically expressed by T cells. Co-receptor for CD80

CXCL8

(IL-8). Recruits neutrophils in innate inflammatory responses

CXCL13

Classically expressed by dendritic cells. Recruits B cells to germinal centers.^58,59

Crosstalk

Signals, whether direct or indirect, that cause changes in the activity of one gene expression program, cell, or network to be associated with changes in the activity of another gene expression program, cell, or network

Decorin

Extracellular matrix proteoglycan

E-Selectin

(Endothelial Leukocyte Adhesion Molecule 1, ELAM-1; leukocyte-endothelial cell adhesion molecule 2, LECAM-2). Classically expressed by endothelial cells. Promotes extravasation of bone marrow derived cells and T cells

IFN-γ

Classically expressed by T_H1 cells; also may be expressed by NKT cells. Promotes MMΦDC cell activation in adaptive, cell-mediated responses; directs B cell class switching toward complement-binding inflammatory IgG isoforms.

IL-1α

Expressed by MMΦDC lineage cells and parenchymal cells. In rabbit lacrimal gland expressed by immune cells (this study). Secreted in mature, active form. Pleiotropic cytokine; mitogenic for bone marrow-derived cells; induces expression of chemokines and iNOS.

IL-1β

Expressed similarly to IL-1α. Secreted in precursor from, converted to mature form by caspase-1

IL-1RA

IL-1 receptor antagonist

IL-2

Classically expressed by T_H1 cells.⁶⁰ Expressed by acinar epithelial cells in rabbit lacrimal gland (this study). Mitogenic autocrine factor and cytokine; promotes T cell and B cell proliferation.

IL-4

Classically expressed by T_H2 cells; also expressed by MMΦDC lineage cells. Directs B cell activation, differentiation, and isotype class switching; suppresses IFN-γ expression.

IL-6

Expressed similarly to IL-1α and IL-1β. Mitogenic cytokine supporting bone marrow-derived cell survival in innate immune responses; interacts with TGF-β to direct plasmacytes to undergo terminal differentiation to plasmablast; also interacts with TGF-β to induce IL-17A expression.⁶⁹

IL-10

Classically expressed by T_H2 cells and MMΦDC lineage cells; also expressed by T_R1 regulatory CD4⁺ cells, regulatory CD8⁺ cells,⁸⁴ and regulatory B cells. Expressed by immune cells in rabbit lacrimal gland (this study). Pleiotropic cytokine, suppresses IFN-γ expression, promotes B cell differentiation

IL-15

IL-15 is typically a mitogen for CD8⁺ T cells and NKT cells

IL-17A

Classically expressed by CD4⁺, T_H17 cells; also expressed by NKT cells, γδ T cells,⁸⁵ and by parenchymal cells of various tissues. Promotes chemokine expression but generally suppresses IFN-γ expression.⁶² Implicated in many autoaggressive responses, including rheumatoid arthritis and Sjögren’s disease, and chronic inflammatory processes, including osteoarthritis, atheroma formation, and dry eye disease

IL-18

Classically expressed by MMΦDC lineage cells; IL-1 family cytokine secreted in immature form; promotes IFN-γ expression and IL-17A expression; promotes B cell isotype class switch to IgG.

IL-18R

IL-18 receptor

iNOS

(Inducible Nitric Oxide Synthase). Expressed by various cells, including parenchymal cells, typically induced by IL-1α and IL-1β

IRF-1

(Interferon Response Factor 1). Activates transcription of Type 1 interferons

Lipophilin CL

Secretoglobin - uteroglobin family lipid-binding protein. Highly expressed in lacrimal gland and secreted into tears in rabbits; may substitute for lipocalin

LTβ

(Lymphotoxin beta). Typically expressed by lymphoid tissue inducer cells

MHC I

(Major Histocompatibility Complex Antigen Type I; in humans, HLA Class I). Classically expressed by MMΦDC lineage cells and virus-infected cells. Presents peptide epitopes to T cell receptor - CD8 complex

MHC II

(Major Histocompatibility Complex Antigen Type II; in humans, HLA Class II). Classically expressed by MMΦDC lineage cells. Presents peptide epitopes to T cell receptor - CD4 complex

MMΦDC lineage cells

Monocytes as well as macrophages and dendritic cells, which develop from monocytes. MMΦDC express a number of the assayed transcripts in common, including mRNAs for CD80, CD86, MHC I, and MHC II

MMP-9

Degrades collagen inflammation and tissue remodeling

PRL

(Prolactin). Anterior pituitary hormone. During pregnancy, interacts with high levels of estradiol and progesterone to promote lactogenesis; following declines of estradiol and progesterone at parturition, promotes lactation; promotes physiologic heat stress response. Also a pleiotropic cytokine, expressed by T cells, B cells, and epithelial cells, including acinar- and ductal epithelial cells. Cytokine actions include up-regulating IFN-γ expression⁴¹; up-regulating expression of IL-4, ^43,44 supporting proliferation and survival of T cells, B cells, and MMΦDC lineage cells^38–40; promoting MMΦDC expression of pro-T_H1 phenotype.⁴²

RTLA

(Rabbit Thymic Lymphocyte Antigen). Used as a marker for rabbit T cells

TGF-β₁

Expressed by lacrimal gland immune cells (this study). Pleiotropic cytokine, supports or suppresses T cell proliferation; suppresses IFN-g expression; promotes plasmablast differentiation; promotes T_H17 cell differentiation.

TGF-β₂

Expressed by lacrimal gland ductal epithelial cells (this study). See TGF-b₁

TNF-α

(Tumor Necrosis Factor alpha). Classically expressed by MMΦDC lineage cells

VCAM-1

(Vascular Cell Adhesion Molecule 1). Classically expressed by endothelial cells. Mediates T cell extravasation