CD5L Constitutes a Novel Biomarker for Integrated Hepatic Thyroid Hormone Action

Animal husbandry and treatment

Animals were housed in groups under constant 12 hours/12 hours light/dark cycle at ambient temperature (22°C ± 1°C) and had ad libitum access to food and water. TRβ^−/− and TRα1^+/m mutant mice (38,39) were bred in the animal core facility Lübeck and compared with TRβ^+/− or TRα1^+/+ littermates. Wild-type C57BL/6NCrl animals were purchased from Charles River Laboratories GmbH (Erkrath, Germany) and rendered hypothyroid with methimazole (0.1% m/v) (M8506; Sigma-Aldrich Chemie GmbH, Taufkirchen, Germany) and sodium perchlorate (0.2% m/v) (106564; Merck KGaA, Darmstadt, Germany) for three weeks in the drinking water. Mild or severe thyrotoxicosis was induced by adding either T4 (T2376; Sigma-Aldrich Chemie GmbH) at 1 mg/L or T3 (T6397; Sigma-Aldrich Chemie GmbH) at 0.5 mg/L to the drinking water, respectively. To maintain regular drinking behavior, 0.1% m/v bovine serum albumin (01400.1; BIOMOL GmbH, Hamburg, Germany) was added into the drinking water of the thyrotoxic and the untreated control groups, and 40 pills/L of the nondigestive sweetener, consisting of 33.8% sodium cyclamate (E952) and 3.8% sodium saccharin (E954) (Suketter; Cederroth International, Väsby, Sweden), was added into the drinking water of the hypothyroid group. All animal procedures were approved by the Schleswig-Holstein Ministerium für Energiewende, Landwirtschaft, Umwelt, Natur und Digitalisierung, Kiel.

Analysis of omics datasets

Plasma proteome data of a previously conducted human longitudinal thyrotoxicosis study were further analyzed (40,41). Briefly, 16 healthy male volunteers were experimentally rendered hyperthyroid by a daily oral administration of 250 μg LT4 for eight weeks. Plasma samples were drawn before, during, at the end of an eight-week T4 administration, and after a subsequent eight-week recovery period. Plasma proteins were analyzed by using an untargeted LC-MS/MS approach after depletion of six abundant plasma proteins with an MARS-6 column.

The plasma proteomes of animals from a previously performed mouse study (42), in which one group of C57BL/6 wild-type mice was rendered hyperthyroid for two weeks (thyrotoxicosis group), one group was rendered hyperthyroid for two weeks followed by two weeks of recovery (recovery group), and one control group remained untreated for four weeks (control group), were analyzed by applying a similar untargeted proteomics workflow as for the human samples but substituting the MARS-6 by an MARS-3 column.

For target gene identification, four filter steps were used: (1) exclusion of proteins with FDR-adjusted p-values >0.05, (2) comparison of recovery group or control group with the thyrotoxicosis group and exclusion of proteins with a fold change < −/+1.5, (3) exclusion of all proteins that are not present in both comparisons, and (4) exclusion of all discordantly altered proteins between mice and humans. The final list after all filtering steps can be found in Table 1. Microarray-based analysis of the hepatic transcriptome was performed as previously described (43). For identifying target gene expressing tissues, the databases UniProt (44) and Expression Atlas (EMBL-EBI, Hinxton, United Kingdom) were used. For creating the heatmap and performing a hierarchical clustering, the R package “complexHeatmap” was used.

ELISA

Serum concentrations of tT4 (EIA-1781; DRG Diagnostics GmbH, Marburg, Germany) and tT3 (DNOV053; NovaTec Immundiagnostica GmbH, Dietzenbach, Germany) were measured with commercially available immunoassays in accordance to the manufacturer's instructions.

Western Blot analysis

A set of serum samples with relatively elevated or suppressed TSH values, respectively, were selected from a cross-sectional study of thyroid patients (45). Serum aliquots (0.4 μL per sample) were separated by SDS-PAGE in a 10% SDS gel at 80 V. After blotting onto a nitrocellulose membrane (Protan™; Merck, Darmstadt, Germany), total protein was quantified by Ponceau S staining. CD5L was detected by rabbit antibodies against the N-terminus of human CD5L (AHP1169; Bio-Rad Laboratories GmbH, Feldkirchen, Germany) in combination with goat anti-rabbit-immunoglobulin-HRP antibody (P0448; Agilent Dako, Santa Clara, CA). Relative CD5L expression was quantified by using ImageJ 1.48v (National Institutes of Health).

Separation of macrophages and parenchymal cells from mouse livers

Mouse livers were perfused and digested as previously described to allow for the isolation and separation of parenchymal and nonparenchymal cells (46). Parenchymal cells (mainly hepatocytes) were pelleted by centrifugation at 35 g for 3 minutes. To obtain liver macrophages, supernatants were centrifuged at 650 g for 7 minutes; pellets were subsequently resuspended in Dulbecco's modified Eagle's medium (DMEM) and subjected to selective plating by using uncoated tissue culture flasks. After two to three hours, flasks were vigorously washed to remove nonadherent cells. Adherent cells were collected as the liver macrophage fraction.

RNA isolation, cDNA generation, and gene expression analysis

Primary cell and cell line culture

All cells and cell lines were incubated at 37°C in a humidified 5% CO₂-containing atmosphere.

Statistical analysis

For all statistical analyses, Graphpad Prism 6 (Graph Pad Software, San Diego, CA) was used. Statistical significance of differences between variables was determined by an unpaired two-tailed Student's t-test and Welch's correction and was considered significant at p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), and p < 0.0001 (****). Values are illustrated as mean ± SEM. Pearson correlation coefficient was determined for calculating variable dependence, and a two-tailed p-value was used to ascertain statistical significance.

Target gene identification by an untargeted omics approach

For the identification of new TH-dependent serum biomarkers, a comparative plasma proteome analysis was performed by using data from a previously conducted experimental human and mouse thyrotoxicosis study (Fig. 1A) (40,41). We decided to use this two-species approach for several reasons. First, we aimed at eliminating any genetic effects in the human sample by confirming the effects in genetically homogenous wild-type animals. Second, we reasoned that any regulation that could be detected consistently in both mice and humans would be more robust, and third, this approach would allow us to use organs or cells from animal models lacking TR for in-depth studies.

OPEN IN VIEWER

FIG. 1.

Identification of biomarker candidates in previously conducted thyrotoxicosis studies in human and mice. (A) Schematic overview of the human and mouse study for Omics comparison of baseline, hyperthyroid, and recovery state. (B) Plasma proteome data sets were determined as described elsewhere (41), and human and mouse data sets were analyzed comparatively. Proteins with a p-value >0.05 and a fold change of <1.5 in both directions between control group vs. LT4-treated group and LT4-treated vs. recovery group were excluded, resulting in 16 concordantly differentially abundant targets. (C) Heat map of concordantly differentially abundant targets proteins in the human and mouse plasma proteome. LT4, levothyroxine. Color images are available online.

After filtering for statistically significant differences in protein levels between the euthyroid (control or recovery) groups and the thyrotoxic T4-treated groups, we identified 32 common plasma proteins, with 16 of these concordantly affected across species (Fig. 1B+C and Table 1). A subsequent comparison with the respective liver transcriptomes revealed that three of these candidates, CD5L, IL1AP, and TTR, could be of hepatic origin.

Induction of acute thyrotoxicosis in wild-type mice

To confirm whether the genes encoding these 16 proteins are, indeed, differentially expressed in hypothyroidism and thyrotoxicosis, we conducted an additional independent animal study, in which wild-type mice were treated with either a mixture of methimazole and perchlorate to induce hypothyroidism and T4 or T3 to induce states of mild or severe thyrotoxicosis in the drinking water, respectively (Fig. 2A). The efficiency of this treatment was confirmed with serum tT4 and tT3 ELISA (Fig. 2B, C), confirming the expected hypo- or hyperthyroidism. We then determined gene expression of all 16 candidate genes in liver, spleen, and bone, as these tissues were identified by database searches as being potentially relevant sources for the circulating proteins. Altogether 14 genes were found to be expressed in the liver; the mRNA levels of Angt, Apob, Camp, Fbln1, Il1ap, and Ttr were not affected by TH status. Statistically significant changes in MMI/ClO₄-treated mice were found for the transcript levels of Apoa1, Apoc3, Fetua, Lum, and Pgrp2; while in the moderately thyrotoxic T4-treated group those of Apoc3, Apoc4, and Cd5l were significantly altered (Fig. 2D). In addition, expression of Apoc4, Cd5l, Lum, and Vcam1 was significantly increased in highly thyrotoxic T3-treated animals (Fig. 2D). The known hepatic TH target genes Spot14 (48) and Dio1 (49) were included as positive controls and showed the expected positive regulation in the hyperthyroid groups. In spleen, the expression of 6 of our 16 candidate genes was changed with Cd5l and Vcam1, showing decreased transcript levels in the MMI/ClO₄ group, but no changes in hyperthyroid groups (Fig. 2E). In bone, transcripts of six of the candidate genes were detected, but only Cd5l revealed altered gene expression in T3-treated animals compared with controls (Fig. 2F). To test which of the hepatic expressed genes display a good correlation with serum tT3 or tT4 levels, we plotted the mRNA levels and serum tT3 or tT4 levels of the individual animals. The transcript levels of Apoc3, Apoc4, Cd5l, Fetua, Lum, and Vcam1 correlated significantly with serum tT3 with the exception of Vcam1, and also with serum tT4 levels (Supplementary Fig. S1A–L), suggesting that these genes constitute valuable novel readouts for hepatic TH action in animal models, in addition to well-known targets such as Dio1 and Spot14.

OPEN IN VIEWER

FIG. 2.

Validation of biomarker candidates under thyrostatic or thyrotoxic conditions. (A) Mouse study design to render mice hypo- or hyperthyroid with methimazole+perchlorate or T4 or T3 for three or two weeks, respectively. (B) Average serum tT4 concentrations of the respective groups. (C) Average serum tT3 concentrations of the respective groups. (D) Determination of gene expression of biomarker candidate genes as well as known TH responsive genes by RT-qPCR in liver. (E) Determination of biomarker candidate gene expression by RT-qPCR in spleen. (F) Determination of biomarker candidate gene expression by RT-qPCR in bone. Relative expression values are represented as mean ± SEM of n = 6–8. Statistically significant differences between groups were determined with an unpaired two-tailed t-test and Welch's correction and are illustrated as *for p < 0.05, **for p < 0.01, ***for p < 0.001, and ****for p < 0.0001. RT-qPCR, real-time quantitative PCR; T3, 3,3′,5-triiodothyronine; TH, thyroid hormone; tT3, total T3; tT4, total T4.

Short-term effects of THs on potential target genes

To evaluate whether the candidate genes rapidly respond to TH, which would point to a direct rather than indirect target of TH action, we conducted a short-term thyrotoxicosis study. To this end, mice were rendered hyperthyroid with T4 (1 mg/mL) for 4 days or with T3 (0.5 mg/mL) for 24 hours in drinking water (Fig. 3A). The resulting serum tT4 and tT3 concentrations of the T4-treated group (Fig. 3B+C) already reflected a similar degree of hyperthyroidism and hyper-T3-emia as after the two-week treatment (Fig. 2B+C). Surprisingly, in T3-treated animals, serum tT3 concentrations were even more increased (Fig. 3C) as compared with the two-week treatment (Fig. 2C). Subsequent RT-qPCR analysis of hepatic gene expression (Fig. 3D) revealed decreased mRNA amounts of Angt as well as increased transcript levels of Cd5l and Fbln1 on T4 treatment, with a similar trend for Cd5l (p = 0.09) and Fbln1 (p = 0.06) in the T3-treated mice. The positive controls Spot14 and Dio1 showed a fast and direct response as expected. Most interestingly, Cd5l expression was strongly and significantly correlated with serum tT4 (p = 0.0143) and tT3 (p = 0.0133) concentrations (Supplementary Fig. S2A+B), suggesting that this gene could, indeed, represent a suitable biomarker for hepatic TH action. In bone, Aspm, Cd5l, and Lum expression was reduced in the T3 group compared with the T4 and control groups (Fig. 3E).

OPEN IN VIEWER

FIG. 3.

Validation of target gene expression under mild or severe thyrotoxic conditions for a short treatment period. (A) Mouse study design for short-term thyrotoxicosis (n = 6). (B) Average serum tT4 concentrations of the respective groups. (C) Average serum tT3 concentrations of the respective groups. (D) Expression analysis of target genes as well as known TH responsive genes by RT-qPCR in the liver. (E) Determination of biomarker candidate gene expression by RT-qPCR in bone. Relative expression values are mean ± SEM of n = 6–8. Statistically significant differences between groups were determined with an unpaired two-tailed t-test and Welch's correction and are illustrated as *for p < 0.05, **for p < 0.01, ***for p < 0.001, and ****for p < 0.0001.

Circulating CD5L abundance in hypo- and hyperthyroid subjects

The product of the most promising target gene CD5L was tested for serum abundance in hypo- and hyperthyroid subjects selected from a cross-sectional study of thyroid patients by using a commercially available polyclonal antibody. The relative CD5L abundance was significantly different between these two groups (Fig. 4A). In addition, basal TSH, fT4, and fT3 were measured in the serum of these patients and correlated with relative CD5L abundance. Relative CD5L abundance showed a significant inverse correlation with basal TSH (Fig. 4B), a correlation that did not reach statistical significance with fT4 (Fig. 4C), and a strong significant positive correlation with fT3 concentrations (Fig. 4D).

OPEN IN VIEWER

FIG. 4.

Relative CD5L abundance in serum of patients with acute hypo- or hyperthyroidism by Western blot. (A) Relative CD5L abundance in hypo- and hyperthyroid patients. (B) Spearman correlation of serum CD5L abundance with basal serum TSH concentration. (C) Spearman correlation of serum CD5L abundance with serum fT4 concentration. (D) Spearman correlation of serum CD5L abundance with serum fT3 concentration. Values are mean ± SEM of n = 7. Statistically significant differences between groups were determined with an unpaired two-tailed t-test and Welch's correction and are illustrated as **for p < 0.01. fT3, free triiodothyronine; fT4, free thyroxine.

Target gene expression in TR mutants

To better characterize the underlying molecular mechanism and to identify TR isoforms involved in the regulation of these genes, TRα1^+/m and TRβ^−/− mutant mice with littermate controls were studied (38,39). TRα1^+/m mice, which have normal levels of serum TH, displayed an increased hepatic expression of Angt, Apoa1, Fetua, Pgrp2, and Dio1 compared with wild-type littermates, while the endogenously hyperthyroid TRβ^−/− mice had an increased expression of Camp and a decreased expression of Dio1, compared with heterozygous littermates (Fig. 5A). Remarkably, expression of Cd5l was not upregulated despite the elevated TH levels in TRβ^−/− mice, suggesting that its induction by TH depends on this receptor isoform. In spleen, expression of Camp was increased and Apod expression was decreased in TRα1^+/m (Fig. 5B). No changes in target gene expression in TRβ^−/− mice were detectable, although there was a tendency of increased transcript levels of Apoa1 and Apoc3 (Fig. 5B). In bones of TRα1^+/m mice, we observed an increased expression of Apod, Aspm, and Camp, while Lum mRNA levels were reduced compared with wild-type mice (Fig. 5C).

OPEN IN VIEWER

FIG. 5.

Validation of candidate gene expression in tissues of TRα1^+/m and TRβ^−/− mice compared with wild-type animals. (A) Expression analysis of target genes as well as the known TH responsive genes by RT-qPCR in the liver of TR mutant mice compared with wild-type animals (n = 5–7). (B) Determination of target gene expression in the spleen of TR mutant mice compared with wild-type animals (n = 5–12). (C) Determination of target gene expression in the bone of TR mutant mice compared with wild-type animals (n = 5–7). Values are mean ± SEM of n = 3–12. Statistically significant differences between groups were determined with an unpaired two-tailed t-test and Welch's correction and are illustrated as *for p < 0.05, **for p < 0.01, and ***for p < 0.001. TR, TH receptors.

Target gene expression in primary bone cells

As 6 of the 16 initially identified target genes were expressed in bone, including Cd5l and Lum, which were regulated under acute T3 treatment or in TR mutant mice, we used primary bone cell culture, differentiated osteoblasts, from mouse mesenchymal stem cells (Fig. 6A) or osteoclasts from mouse hematopoietic stem and progenitor cells (HSPCs) (Fig. 6B) to identify the cell type expressing the gene and to test for a direct cell-type specific regulation. The gene expression analysis of primary mouse osteoblasts revealed a reduction of Cd5l and Lum mRNAs in cells deficient for TRβ, while only a tendency for decreased Cd5l expression or increased Lum expression in cells carrying a mutation in TRα1 was detectable (Fig. 6C). These genes were also affected in primary mouse osteoclasts where Cd5l transcript amounts were increased in cells lacking TRβ, while those of Lum only showed a tendency toward increase in these cells (p = 0.16). Conversely, in primary osteoclasts from TRα1^+/m mice, Cd5l expression was unchanged and Lum expression increased (Fig. 6D). Surprisingly, different TH concentrations did not induce any alterations in the expression of Cd5l and Lum (Fig. 6E), suggesting that those observed in the different animal models are rather the result of pre-existing endogenous dysregulations in these cell types caused by the genetic manipulations, and do not result from direct regulation via these TRs.

OPEN IN VIEWER

FIG. 6.

Determination of target gene expression in primary osteoblast or osteoclast culture. (A) Representative picture of primary osteoblast cell culture on day 14 after beginning of differentiation (40 × magnification). (B) Representative picture of primary osteoclast cell culture on day 5 after beginning of differentiation (40 × magnification). (C) Determination of target gene expression in primary mouse osteoblasts of TRα1^+/m and TRβ^−/− mice compared with wild-type littermates. (D) Determination of target gene expression in primary mouse osteoclasts of TRα1^+/m and TRβ^−/− mice compared with wild-type littermates. (E) Determination of Cd5l and Lum expression in wild-type osteoblasts and osteoclasts under hypo-, eu-, or hyperthyroid culture conditions. Values are mean ± SEM of n = 3–5. Statistically significant differences between groups were determined with an unpaired two-tailed t-test and Welch's correction and are illustrated as *for p < 0.05.

Cd5l expression in BMDMs

As the main source of CD5L are hepatic Kupffer cells and other tissue residing M1 macrophages (50,51), we hypothesized that the effect of TH on hepatic gene expression would be on liver-residing macrophages rather than epithelial cells, circulating immune cells, or hepatocytes, which did not express high levels of Cd5l as assessed by qPCR analysis in sorted cell populations (Supplementary Fig. S2C, D). Therefore, BMDMs were differentiated from HSPCs by treatment with M-CSF for six days, followed by an LPS+IFNγ-induced polarization for 24 hours into M1 BMDM (Fig. 7A). To test which TR isoform could be involved, HSPCs from TRα1^+/m and TRβ^−/− mice were also differentiated into M1 BMDM and compared with wild-type BMDM Cd5l expression. BMDM from animals carrying a mutation in Thra encoding TRα1 exhibited an 82% reduced expression of Cd5l, whereas BMDM lacking TRβ showed a similar expression as wild-type cells (Fig. 7B). Surprisingly, T3 itself did not affect the expression of Cd5l in wild-type BMDMs (Fig. 7C), suggesting that the changes in the TR mutant mice are likely caused by general alterations in BMDM formation capacity in vitro but do not result from direct regulation via TH. We, therefore, tested whether the BMDMs would be generally capable of inducing Cd5l expression by using liver X receptor (LXR) and retinoid X receptor (RXR) agonists, which are known transcriptional inducers of Cd5l (52 –54). Treating BMDM with LXR agonist T0901317 (T1317) and RXR agonist 9-cis retinoic acid (9cRA) resulted in increased Cd5l transcript levels in all groups independently of T3 in the culture medium, demonstrating that the BMDM of all animal models are capable of inducing Cd5l expression (Fig. 7D). We, therefore, hypothesized that the effect of T3 on Cd5l mRNA expression would be indirect through hepatocytes, which would be supported by the fact that the increase in gene expression in the intact animal depends on TRβ, although in macrophages TRα1 represents the predominating TR (55). To test whether a factor secreted by hepatocytes might regulate Cd5l expression, we exposed BMDMs to medium primed by T3-treated primary hepatocytes for 48 hours (Fig. 7E). To control for effects that could be caused by a general depletion of medium ingredients through the cellular metabolism, we used medium primed by T3-treated human hepatic stellate cell line LX-2 for 48 hours as a negative control. In contrast to the LX-2 medium, the hepatocyte primed medium was, indeed, capable of increasing Cd5l expression in mouse M1 BMDM in both TRα1^+/m and wild-type mice, but not in TRβ^−/− BMDMs (Fig. 7F)—an effect detectable across all tested TH concentrations (Supplementary Fig. S2E–G). The effect of medium primed by hepatocytes of TRβ^−/− mice was comparable to that of wild-type mice (Supplementary Fig. S2H). Analyses of pro- and anti-inflammatory macrophage polarization markers did not reveal changes in M1 marker profile of BMDMs from TR mutant mice compared with wild-type mice, but M2 marker Il10 expression was increased in BMDMs from TRα1^+/m mice, indicating a more M2-like polarization, although Argi1 expression was not affected (Fig. 7G). These findings demonstrate that a TRβ-dependent hepatocyte-derived factor is required to induce Cd5l expression in BMDMs. The positive regulation of Cd5l by TH, which can be detected in serum proteome and clinical samples, liver transcriptome and cell culture, therefore may provide an integrated readout of TH action on the liver-macrophage crosstalk.

OPEN IN VIEWER

FIG. 7.

Determination of target gene expression in primary BMDM. (A) Representative picture of primary M1 BMDM (40 × magnification). (B) Cd5l expression in M1 BMDM under euthyroid conditions in wild-type animals compared with TRα1^+/m and TRβ^−/− mice. (C) Cd5l expression in M1 BMDM under hypo-, eu-, or hyperthyroid conditions in wild-type animals compared with TRα1^+/m and TRβ^−/− mice. (D) Fold change of Cd5l expression in euthyroid M1 BMDMs compared with euthyroid M1 BMDM with LXR and RXR agonist treatment for 24 hours in wild type, TRα1^+/m, and TRβ^−/− mice. (E) Representative picture of primary hepatocytes 48 hours after seeding (40 × magnification). (F) Cd5l expression in M1 BMDM after 24 hours incubation with LX-2 or primary hepatocyte conditioned medium with 1 nM T3. (G) Gene expression of M1 and M2 marker genes in BMDMs of wild type, TRα1^+/m, and TRβ^−/− mice after 24 hours of incubation with LX-2 or primary hepatocyte conditioned medium with 1 nM T3. Values are mean ± SEM of n = 3–6. Determination of statistically significant source of variation on Cd5l expression of conditioned medium across different T3 concentrations (data not shown) was performed by a two-way ANOVA and is indicated for the medium by ^$$for p < 0.01 and for the T3 concentration by *for p < 0.05. Determination of a statistically significant source of variation on M1/M2 marker gene expression of conditioned medium across different genotypes was performed by a two-way ANOVA and is indicated for the genotype by ^###for p < 0.001. Statistically significant differences between groups were determined with an unpaired two-tailed t-test and Welch's correction and are illustrated as *for p < 0.05, **for p < 0.01, and ****for p < 0.001. BMDM, bone marrow-derived macrophages; RXR, retinoid X receptor; LXR, liver X receptor.