Inadequately low serum levels of steroid hormones in relation to interleukin-6 and tumor necrosis factor in untreated patients with early rheumatoid arthritis and reactive arthritis

Authors


Abstract

Objective

To compare levels of steroid hormones in relation to cytokines and to study levels of cortisol or dehydroepiandrosterone (DHEA) in relation to other adrenal hormones in untreated patients with early rheumatoid arthritis (RA) and reactive arthritis (ReA) compared with healthy controls.

Methods

In a retrospective study with 34 RA patients, 46 ReA patients, and 112 healthy subjects, we measured serum levels of interleukin-6 (IL-6), tumor necrosis factor (TNF), adrenocorticotropic hormone (ACTH), cortisol, 17-hydroxyprogesterone (17-OH-progesterone), androstenedione (ASD), DHEA, and DHEA sulfate (DHEAS).

Results

RA patients had higher serum levels of IL-6, TNF, cortisol, and DHEA compared with ReA patients and healthy subjects, but no difference was noticed with respect to ACTH and DHEAS. However, in RA and ReA patients compared with healthy subjects, levels of ACTH, cortisol, ASD, DHEAS, and 17-OH-progesterone were markedly lower in relation to levels of IL-6 and TNF. Furthermore, the number of swollen joints correlated inversely with the ratio of serum cortisol to serum IL-6 in RA (RRank = −0.582, P = 0.001) and, to a lesser extent, in ReA (RRank = −0.417, P = 0.011). In RA patients, the mean grip strength of both hands was positively correlated with the ratio of serum cortisol to serum IL-6 (RRank = 0.472, P = 0.010). Furthermore, in these untreated patients with RA and ReA, there was a relative decrease in the secretion of 17-OH-progesterone, ASD, and DHEAS in relation to DHEA and cortisol. This indicates a relative predominance of the nonsulfated DHEA and cortisol in relation to all other measured adrenal steroid hormones in the early stages of these inflammatory diseases.

Conclusion

This study indicates that levels of ACTH and cortisol are relatively low in relation to levels of IL-6 and TNF in untreated patients with early RA and ReA compared with healthy subjects. The study further demonstrates that there is a relative increase of DHEA and cortisol in relation to other adrenal hormones, such as DHEAS. This study emphasizes that adrenal steroid secretion is inadequately low in relation to inflammation. Although changes in hormone levels are similar in RA and ReA, alteration of steroidogenesis is more pronounced in RA patients than in ReA patients.

With respect to chronic inflammatory diseases in humans, there is a reduction of cortisol relative to the degree of inflammation, as exemplified in African trypanosomiasis (1), Sjögren's syndrome (2), systemic lupus erythematosus (3), and polymyalgia rheumatica (4). This is particularly the case when the disease persists over a long period of time (weeks). With respect to rheumatoid arthritis (RA), patients have been described as having inappropriately low spontaneous and stimulated cortisol secretion levels (5–13). Another significant finding in patients with RA is the decrease in serum levels of dehydroepiandrosterone sulfate (DHEAS), an adrenal androgen (14–22). Investigators in our group have recently reported that an alteration of the hypothalamic–pituitary–adrenal axis (HPA axis) in RA has an impact on perpetuation of the disease (23).

In the present study, we wanted to extend our view of endocrine/immune relationships by comparing the levels of steroid hormones in patients with reactive arthritis (ReA) with those in RA patients. ReA and RA are two forms of inflammatory arthritis with common pathogenic features as well as dissimilarities (24, 25). On the one hand, bacteria or bacterial products may play an important role in stimulating and maintaining inflammation in both diseases (24, 25). On the other hand, RA is a chronic, lifelong, erosive disease, while ReA is usually self limiting and nonerosive. Although one can speculate that the triggering factors for these diseases may be similar, the reasons for the different courses of RA and ReA in their early phases remain largely unknown.

The question arises whether systemic inflammation similarly alters the HPA axis in RA and ReA. If one could detect a difference in pituitary and adrenal hormone levels, such a difference would have an impact on the early course of the disease, probably leading to perpetuation of inflammation. To our knowledge, there is only one study which focused on serum levels of adrenal hormones in ReA patients, and investigators in that study reported normal levels of cortisol, DHEAS, and pregnenolone sulfate (19). However, they did not study hormone levels in relation to inflammation or steroid hormone shifts, which can indicate preferential production of one hormone over another (4). This is important because cortisol favors a Th2 reaction (26–28), while adrenal androgens, such as DHEAS, DHEA, or androstenedione (ASD), support a Th1 reaction (for review, see ref. 29). Thus, hormone shifts can indicate a preponderance of one hormone over another, with possible endocrine immune consequences in the early course of these diseases.

It was the aim of this study to compare steroid hormone secretion in relation to proinflammatory cytokines in untreated patients with early RA and ReA versus healthy subjects. Furthermore, we wanted to study adrenal hormone shifts in both patient groups and compare them with those in healthy controls.

PATIENTS AND METHODS

Patients. We included 34 white patients (27 women, 7 men) newly diagnosed as having RA (duration of symptoms ≤12 months at entry) according to the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 revised criteria (30). The mean ± SEM duration of disease in the RA patients was 6.2 ± 0.7 months. The mean ± SEM age of RA patients was 52.9 ± 2.9 years (range 21–80 years). None of the RA patients had previously received disease-modifying antirheumatic drug (DMARD) therapy or oral glucocorticoids. Clinical variables of disease activity included the number of swollen joints, grip strength (in bars; average of both hands), and severity of pain using a visual analog scale (31).

We enrolled 46 white patients with ReA (17 women, 29 men) with a mean ± SEM disease duration of 3.5 ± 0.6 months. The mean ± SEM age of the patients was 31.8 ± 1.9 years (range 15–78 years). As with the RA patients, none of the ReA patients had previously received DMARDs or oral glucocorticoids. ReA had been caused by Yersinia enterocolitica in 20 patients (15 of whom were HLA–B27 positive). Twenty-six patients (22 of whom were HLA–B27 positive) presented with other causes of enteritis or urethritis, the etiologies of which were Chlamydia trachomatis in 5 (3 of whom were HLA–B27 positive), Salmonella enteritidis in 7 and Campylobacter jejuni in 3 (all HLA–B27 positive in both groups), and agents not known in 11 (9 of whom were HLA–B27 positive).

Simultaneously with the clinical examination, inflammatory activity was assessed in both disease groups by measuring the erythrocyte sedimentation rate (ESR). Blood for further determination of steroid hormone or cytokine levels was drawn between 9:00 AM and 12:00 PM, and serum was stored at −20°C in adequate aliquots.

Radiographs of the hands and feet of each RA patient were taken and read by an experienced radiologist who was unaware of the patient's clinical data. Radiographic changes were graded according to the Larsen index (32).

Healthy control subjects. One hundred twelve white healthy control subjects were recruited (mean ± SEM age 45.7 ± 1.6 years), and their health status was verified by means of a 33-item questionnaire as previously described (33). The questionnaire addressed known diseases in the past and at present, current symptoms of diseases, current medication, alcohol intake, smoking habits, family history, and surgical history. The questionnaire was adapted to the SENIEUR protocol (34). Our protocol established strict admission criteria for immunogerontologic studies in humans based on clinical information. Fifty-five of the subjects were men and 57 were women. Fertile women were not taking contraceptives. Blood was drawn between 10:00 AM and 12:00 PM, and serum and plasma were stored at −80°C in adequate aliquots. All subjects gave informed consent for further investigation of blood samples.

Subgroup analyses. Due to the different distributions of age and sex in the disease groups and in the group of healthy controls, subgroup analyses were carried out in order to correctly compare the different groups. The subgroups were matched according to age and sex. Table 1 summarizes the number of patients, their age and sex, and the duration of the diseases in the different subgroups.

Table 1. Basic characteristics of age- and sex-matched patients and healthy controls in the subgroup analyses*
RA patientsReA patientsHealthy controls
  • *

    RA = rheumatoid arthritis; ReA = reactive arthritis; NA = not applicable.

RA patients vs. healthy controls
 No. of subjects3453
 Age, mean ± SEM (range) years52.9±2.9(21–80)54.9±1.4(37–75)
 No. of women/no. of men (% women/% men)27/7(79/21)39/14(74/26)
 Duration of arthritis, mean ± SEM months6.2±0.7NA
ReA patients vs. healthy controls
 No. of subjects4660
 Age, mean ± SEM (range) years31.8±1.9(15–78)32.5±1.1(18–48)
 No. of women/no. of men (% women/% men)17/29(37/63)29/31(48/52)
 Duration of arthritis, mean ± SEM months3.5±0.6NA
All groups compared
 No. subjects222182
 Age, mean ± SEM (range) years45.1±3.0(21–74)41.3±2.7(28–78)44.2±1.4(24–65)
 No. of women/no. of men (% women/% men)15/7(68/32)10/11(48/52)40/42(49/51)
 Duration of arthritis, mean ± SEM months6.4±1.04.2±1.2NA

Laboratory parameters. Several adrenal hormones were considered in order to detect major adrenal pathways of steroidogenesis. Figure 1 demonstrates the steroid cascades in the adrenal gland. We used radioimmunometric assays for the quantitative determination of serum levels of cortisol (detection limit 10 nmoles/liter; Coulter Immunotech, Marseilles, France) and ASD (detection limit 0.14 nmoles/liter; DPC Biermann, Bad Nauheim, Germany). Immunometric enzyme immunoassays were used to measure serum levels of 17-hydroxyprogesterone (17-OH-progesterone) (detection limit 0.3 nmoles/liter; IBL, Hamburg, Germany), DHEAS (detection limit 130 nmoles/liter; IBL), DHEA (detection limit 0.13 nmoles/liter; Diagnostic Systems Laboratories, Webster, TX), interleukin-6 (IL-6) (high sensitivity Quantikine, detection limit 0.2 pg/ml; R&D Systems, Minneapolis, MN), and tumor necrosis factor (TNF) (high sensitivity Quantikine, detection limit 0.2 pg/ml; R&D Systems).

Figure 1.

Schematic diagram demonstrating the biosynthesis of important adrenal hormones. Lines with bars demonstrate the inhibitory effects of indicated mediators (transforming growth factor β1 [TGFβ], interleukin-1β [IL-1], tumor necrosis factor [TNF]). Lines with arrows indicate stimulatory effects of interleukin-6 (IL-6). Broken line indicates pathway to aldosterone, with several intermediate steps. Numbers indicate enzymes (1 = 20,22-desmolase; 2 = 3β-hydroxysteroid dehydrogenase; 3 = 21α-hydroxylase [P450c21]; 4 = 11β-hydroxylase [P450c11]; 5/6 = 17α/17,20-hydroxylase[lyase] [P450c17]; 7 = sulfotransferase for dehydroepiandrosterone [DHEA] → DHEA sulfate [DHEAS] and sulfatase for DHEAS → DHEA). Adrenocorticotropic hormone (ACTH) stimulates the 20,22 desmolase (enzyme 1) to start the synthesis of adrenal hormones. P450c21 (enzyme 3) converts 17-hydroxyprogesterone (17-OH-progesterone) to 11-deoxycortisol, and P450c11 (enzyme 4) converts 11-deoxycortisol to cortisol. 17α/17,20-hydroxylase (enzymes 5 and 6) convert pregnenolone to DHEA and progesterone to androstenedione. TNF can inhibit the secretion of cortisol from adrenal cells due to an inhibition of P450c21 (see ref. 43), and IL-6 increases cortisol secretion (see refs. 44 and 45).

Using a sensitive enzyme immunoassay for adrenocorticotropic hormone (ACTH) (detection limit 0.1 pmoles/liter; Sangui BioTech, Santa Ana, CA, via IBL), we were able to demonstrate a highly significant interrelationship (R = 0.646, P< 0.000001 from 112 healthy subjects [ 33]) between ACTH levels measured in serum and ACTH levels assayed in plasma with the following regression equation: ACTH (plasma) = 7.2508 + 1.8707 × ACTH (serum). In the present study, we measured ACTH in the serum samples obtained from healthy controls and RA and ReA patients using this enzyme immunoassay, and we calculated plasma ACTH levels using the formula above. Calculated plasma levels of ACTH are reported in the Results. Intraassay and interassay coefficients of variation were <10% in all tests.

Calculation of hormone/cytokine ratios and hormone ratios. In order to demonstrate the relationship between serum levels of hormones and serum levels of IL-6 or TNF, a hormone/cytokine ratio was calculated (units were nmoles/liter divided by pg/ml). We used the simple ratio to express the hormone–cytokine relationship which is based on recent investigations by Tsigos et al (35). Those investigators have demonstrated that following subcutaneous injection of different doses of IL-6, there is a linear interrelation between serum levels of IL-6 and levels of serum cortisol or plasma ACTH in the range of 5–100 pg IL-6/ml of serum (maximum IL-6 levels in our study were 75.9 pg/ml in RA patients and 94.5 pg/ml in ReA patients).

In order to demonstrate the shift from one serum hormone to another serum hormone, the molar ratio of these hormones was calculated (given without units, similar to the descriptions in refs. 36–38). This procedure detects a hormonal shift through one or two adrenal enzyme steps which can demonstrate a preponderance of an adrenal pathway (Figure 1): cortisol/17-OH-progesterone for the pathway through P450c21 and P450c11 in the direction of cortisol, ASD/17-OH-progesterone for the 17,20-lyase (second reaction of the P450c17) in the direction of ASD, DHEA/ASD for the 3β-hydroxysteroid dehydrogenase in the direction of DHEA, and DHEAS/DHEA for the combined reaction of the DHEA sulfotransferase and the DHEAS sulfatase in the direction of DHEAS.

Statistical analysis. In order to compare means in two different groups, the Mann-Whitney signed rank test was used (SPSS/PC for Windows, version 10.0.5; SPSS, Chicago, IL). Overall comparison of means in three different subgroups was carried out using the Kruskal-Wallis test (SPSS). Investigation of an interrelation between two variables was done using Spearman rank correlation analysis. In the figures, the linear regression line is given together with the Spearman rank correlation coefficient (RRank) and the respective rank correlation P value. Pvalues less than 0.05 were considered significant; means are always given ± SEM (mean and SEM in figures). If necessary, Bonferroni adjustment was used.

RESULTS

Serum cytokines in patients and healthy controls. In patients with RA and ReA compared with the respective control groups, serum levels of IL-6 and TNF were markedly higher (Table 2). In the overall comparison of age-matched and sex-matched subjects, serum levels of IL-6 and TNF were highest in RA patients, next highest in ReA patients, and lowest in healthy subjects (Figure 2). This indicates a much stronger inflammatory state in patients than in healthy subjects.

Figure 2.

Serum levels of cortisol (nmoles/liter), 17-OH-progesterone (17OHP; nmoles/liter), DHEA (nmoles/liter), IL-6 (pg/ml), and TNF (pg/ml) in 82 healthy subjects (open bars), 21 patients with reactive arthritis (ReA; cross-hatched bars), and 22 patients with rheumatoid arthritis (RA; solid bars). Values are the mean and SEM. P values above the bars indicate significance of the overall comparison of the three subgroups, by Kruskal-Wallis analysis. The P value above the thin line indicates significance of the comparison of RA patients and ReA patients. See Figure 1 for other definitions.

Table 2. Subgroup comparison of RA or ReA patients with healthy subjects*
 RA patients vs. healthy subjectsReA patients vs. healthy subjects
RA patients (n = 34)Healthy subjects (n = 53)ReA patients (n = 46)Healthy subjects (n = 60)
  • *

    Values are the mean ± SEM. P= 0.05 (with Bonferroni correction for 17 comparisons, P= 0.0029). RA = rheumatoid arthritis; ReA = reactive arthritis; IL-6 = interleukin-6; TNF = tumor necrosis factor; ACTH = adrenocorticotropic hormone; 17-OH-progesterone = 17-hydroxyprogesterone; ASD = androstenedione; DHEA = dehydroepiandrosterone; DHEAS = DHEA sulfate.

  • A direct comparison of RA and ReA patients with their respective control groups is not possible due to significantly different age and sex (these comparisons are depicted in Figures 2 and 3 and Table 3).

  • P ≤ 0.001 versus respective group of healthy subjects, by Mann-Whitney signed rank test.

  • §

    P = 0.005 versus respective group of healthy subjects, by Mann-Whitney signed rank test.

Cytokine levels, pg/ml
 IL-621.0 ± 3.42.0 ± 0.210.0 ± 2.71.4 ± 0.1
 TNF5.8 ± 0.31.8 ± 0.13.6 ± 0.21.8 ± 0.1
Hormone levels
 Plasma ACTH, pmoles/liter5.1 ± 0.64.9 ± 0.34.6 ± 0.14.9 ± 0.3
 Cortisol, nmoles/liter401 ± 29288 ± 27396 ± 30346 ± 31
 17-OH-progesterone, nmoles/liter1.7 ± 0.33.2 ± 0.32.4 ± 0.24.7 ± 0.3
 ASD, nmoles/liter6.0 ± 1.15.2 ± 0.47.7 ± 0.710.5 ± 0.8
 DHEA, nmoles/liter28.7 ± 3.411.9 ± 1.336.8 ± 3.820.0 ± 1.6
 DHEAS, nmoles/liter4,717 ± 5243,644 ± 2516,500 ± 4696,298 ± 335
Ratios
 ACTH/IL-60.8 ± 0.33.1 ± 0.32.2 ± 0.44.3 ± 0.4
 ACTH/TNF1.0 ± 0.23.0 ± 0.21.8 ± 0.43.1 ± 0.2
 Cortisol/IL-6120 ± 52170 ± 20179 ± 37285 ± 28
 Cortisol/TNF76 ± 7175 ± 17137 ± 20217 ± 21
 Cortisol/17-OH-progesterone315 ± 33113 ± 12222 ± 2889 ± 10
 Cortisol/DHEA20.7 ± 2.841.5 ± 6.415.8 ± 2.029.5 ± 5.1
 ASD/17-OH-progesterone4.0 ± 0.72.2 ± 0.23.7 ± 0.3§2.7 ± 0.3
 DHEA/ASD6.9 ± 1.02.5 ± 0.35.3 ± 0.42.1 ± 0.1
 DHEAS/DHEA187 ± 16495 ± 73220 ± 16540 ± 121

Serum hormone levels in patients and healthy controls. Compared with the respective control group, serum levels of cortisol and DHEA were markedly higher in RA patients, while levels of 17-OH-progesterone were significantly lower (Table 2). This latter result was also seen in ReA patients, which indicates a shift to cortisol in relation to 17-OH-progesterone in both patient groups (Table 2). In ReA patients compared with healthy subjects, serum levels of cortisol were similar, while DHEA levels were elevated (Table 2). In the overall comparison of age-matched and sex-matched subjects, serum levels of cortisol and DHEA were highest in RA patients, followed by those in ReA patients and healthy subjects (Figure 2). This indicates a much stronger inflammatory response of the adrenal gland in patients compared with healthy subjects.

Interestingly, mean ± SEM ACTH plasma levels, DHEAS serum levels, and ASD serum levels were similar in the subgroup analysis (Table 2) and in the overall analysis, as follows: for ACTH, 4.8 ± 0.7 pmoles/liter in RA patients versus 4.8 ± 0.3 pmoles/liter in ReA patients versus 4.9 ± 0.2 pmoles/liter in healthy subjects (P not significant [NS]); for DHEAS, 5,492 ± 728 nmoles/liter in RA patients versus 5,124 ± 584 nmoles/liter in ReA patients versus 4,843 ± 284 nmoles/liter in healthy subjects (P NS); for ASD, 7.0 ± 1.5 nmoles/liter in RA patients versus 6.3 ± 0.9 nmoles/liter in ReA patients versus 7.9 ± 0.6 nmoles/liter in healthy subjects (P NS).

Relationship of hormones to cytokines in patients and healthy controls. In the subgroup analysis, the ratios of plasma ACTH/serum IL-6, plasma ACTH/serum TNF, serum cortisol/serum IL-6, and serum cortisol/serum TNF were significantly lower in both patient groups compared with healthy subjects (Table 2). This was corroborated by the overall analysis (results shown for serum cortisol in Figure 3). This indicates that levels of serum cortisol and plasma ACTH were low in relation to those of serum IL-6 and serum TNF in RA and ReA patients compared with healthy subjects. A very similar significant reduction of the adrenal hormone in relation to IL-6 was demonstrated for 17-OH-progesterone (data shown) and for ASD, DHEA, and DHEAS in RA and ReA patients (data not shown).

Figure 3.

Ratios of serum cortisol levels to levels of serum cytokines (IL-6 and TNF) in 82 healthy subjects (open bars), 21 ReA patients (crosshatched bars), and 22 RA patients (solid bars). Values are the mean and SEM. P values above the bars indicate significance of the overall comparison of the three subgroups, by Kruskal-Wallis analysis. The P value above the thin line indicates significance of the comparison of RA patients and ReA patients. See Figures 1 and 2 for definitions.

In order to judge the possible clinical relevance of low serum cortisol levels in relation to serum IL-6, we calculated the correlations between clinical variables and the ratio of serum cortisol to serum IL-6. In both patient groups, the ESR correlated negatively with this ratio (for RA patients, RRank = −0.502, P = 0.006; for ReA patients, RRank = −0.612, P = 0.001). Furthermore, this ratio correlated negatively with the number of swollen joints in both patient groups (Figure 4). This was corroborated by the positive correlation of the mean grip strength of both hands with this ratio, which was only assessed in RA patients (RRank = 0.472, P = 0.010). In addition, in RA patients, the Larsen score tended to correlate inversely with the ratio of serum cortisol to serum IL-6 (RRank = −0.352, P = 0.056). However, the subjective marker of joint pain did not correlate with the ratio of serum cortisol to serum IL-6 (data not shown).

Figure 1.

Interrelationship of the number of swollen joints and the ratio of serum cortisol (in nmoles/liter) to serum IL-6 (in pg/ml) levels in RA patients and ReA patients. Shown are the linear regression line as well as the Spearman rank correlation coefficient (RRank) and its P value. See Figures 1 and 2 for other definitions.

With respect to the ratio of serum 17-OH-progesterone to serum IL-6, we found the following correlations with ESR (for RA patients, RRank = −0.545, P = 0.001; for ReA patients, RRank = −0.638, P = 0.001), swollen joints (for RA patients, RRank = −0.476, P = 0.005; for ReA patients, RRank = −0.452, P = 0.006), mean grip strength of both hands (assessed only in RA patients, RRank = −0.622, P < 0.001), and Larsen score (assessed only in RA patients, RRank = −0.229, PNS). These results indicate that a more severe disease was present in patients with RA and ReA when serum levels of cortisol or 17-OH-progesterone were low in relation to serum levels of IL-6.

Major adrenal hormone shifts in patients and healthy controls. Another important aspect of systemic inflammation are hormone shifts of adrenal and gonadal hormones which may influence the inflammatory state in the joints (23). In the subgroup analysis, serum cortisol levels were elevated in relation to 17-OH-progesterone (Table 2), which was confirmed in the overall analysis (Table 3). This ratio was highest in RA patients, next highest in ReA patients, and lowest in healthy subjects (Table 3), which indicates a hormone shift to cortisol at the expense of 17-OH-progesterone. Interestingly, this hormone shift was inversely correlated with the ratio of serum TNF to serum IL-6 in RA patients, but not in ReA patients, which indicates that high levels of TNF in relation to IL-6 probably inhibit this enzyme step (Figure 5).

Figure 5.

Interrelationship of the ratio of TNF to IL-6 and the ratio of cortisol to 17-OH-progesterone in RA patients and ReA patients. Shown are the linear regression line as well as the Spearman rank correlation coefficient (RRank) and its P value. Ratios are given without units. NS = not significant (see Figures 1 and 2 for other definitions).

Table 3. Overall comparison of RA patients, ReA patients, and healthy subjects after matching for age and sex*
RatiosRA patients (n = 22)ReA patients (n = 21)Healthy subjects (n = 82)
  • *

    Values are the mean ± SEM. See Table 2 for definitions.

  • P ≤ 0.001 for the comparison of all three subgroups, by Kruskal-Wallis analysis. P= 0.05 (with Bonferroni correction for 17 comparisons, P= 0.0029).

Cortisol/17-OH-progesterone307 ± 44272 ± 5196 ± 9
Cortisol/DHEA19.9 ± 3.420.5 ± 3.636.2 ± 5.0
ASD/17-OH-progesterone4.3 ± 1.03.7 ± 0.42.3 ± 0.2
DHEA/ASD6.1 ± 0.84.8 ± 0.52.2 ± 0.2
DHEAS/DHEA191 ± 17233 ± 26543 ± 94

Another important shift of adrenal steroid hormones is directed toward an increase of serum DHEA in relation to ASD and DHEAS (Tables 2 and 3). This shift indicates that the 3β-hydroxysteroid dehydrogenase reaction and the combined reaction of the DHEA sulfotransferase and DHEAS sulfatase may be inhibited in these inflammatory diseases. The inhibition is more marked in RA patients than in ReA patients (Table 3). Thus, the ratio of cortisol to DHEA necessarily remains similar in the three groups (Table 3). A further important step is the second step of the P450c17 reaction from 17-OH-progesterone to ASD which has been suspected to be inhibited in chronic systemic inflammatory diseases (4, 39). However, in these RA and ReA patients with early disease compared with healthy subjects, serum levels of ASD were significantly higher in relation to 17-OH-progesterone (Tables 2 and 3), which does not indicate a blockade of the latter enzyme.

DISCUSSION

This study used ratios of hormones to cytokines in order to judge whether secretion levels of cortisol in untreated patients with early RA and early ReA were inadequate in comparison with those of age- and sex-matched healthy controls. We found that plasma ACTH and serum cortisol levels were markedly lower in relation to serum IL-6 and TNF levels in RA and ReA patients compared with healthy subjects. A high ESR and a high number of swollen joints were inversely correlated with the ratios of cortisol or 17-OH-progesterone to IL-6 in RA and ReA patients. Although changes in hormone levels and ratios are similar in RA and ReA, alteration of the steroidogenesis is more pronounced in RA patients than in ReA patients.

It has been previously demonstrated that serum cortisol levels are normal or somewhat higher in untreated patients with RA compared with controls (5–13). This was particularly evident in a study by Crofford et al, who investigated circadian variation of cortisol levels in patients with early RA compared with healthy subjects and found them to be very similar in the two groups (9). Simultaneously, the adrenal androgen DHEAS level was markedly lower in RA patients, which certainly depends on prior glucocorticoid treatment (14–22). It was thought that cortisol secretion remains high at the expense of adrenal DHEAS secretion (40, 41). Findings in patients with Crohn's disease, ulcerative colitis, polymyalgia rheumatica, progressive systemic sclerosis, and systemic lupus erythematosus indicated that this reaction of the adrenal gland is a general response to systemic inflammation independent of the underlying disease (for review, see ref. 23). In the present study, we found higher serum levels of cortisol and the nonsulfated DHEA in RA and ReA patients, which indicates a stronger inflammatory response of the adrenal glands in these patients compared with healthy controls. However, serum levels of ASD or DHEAS did not differ among the three groups. We believe that production of DHEAS was not inhibited in the patients evaluated here because these patients had early disease and were never treated with glucocorticoids.

Normal plasma levels of ACTH and normal serum levels of cortisol in the presence of systemic inflammation is indicative of an inadequate HPA axis response to systemic inflammation (23, 40, 42). In the present study, we aimed to demonstrate directly the failure of ACTH and cortisol secretion in relation to serum IL-6 and TNF levels in untreated patients with early RA and ReA. We used ratios of serum levels of cortisol to IL-6 or TNF levels because there is a linear interrelationship between serum levels of IL-6 and plasma ACTH or serum cortisol levels in the range of 5–100 pg of IL-6/ml of serum in healthy subjects (35). Since our patients had IL-6 serum levels that were <100 pg/ml, we believe that this ratio is an adequate way to judge cortisol serum levels in relation to systemic inflammation.

The ratios of plasma ACTH to serum IL-6 or TNF and those of serum cortisol to serum IL-6 or TNF were significantly lower in both patient groups compared with healthy subjects. Moreover, the ratio of serum cortisol to serum IL-6 was positively correlated with mean grip strength (assessed in RA patients) and negatively correlated with ESR and number of swollen joints in both patient groups. These results indicate the presence of a more severe disease in RA and ReA patients when serum cortisol levels are low in relation to serum IL-6 levels.

In order to further judge specific enzymatic steps for steroidogenesis, we investigated hormone ratios in RA and ReA patients and in healthy controls (see Figure 1). On the one hand, these ratios indicate that there is a shift to cortisol at the expense of 17-OH-progesterone and an increase of ASD at the expense of 17-OH-progesterone in RA and ReA patients compared with healthy subjects. On the other hand, DHEA levels were higher in relation to ASD and DHEAS levels, which indicates inhibition of 1) the 3β-hydroxysteroid dehydrogenase and 2) the conversion from/to DHEAS (see Figure 1). Taken together, the data indicate that there is a shift to cortisol at the expense of 17-OH-progesterone and ASD in ReA patients, but this is more pronounced in RA patients. Furthermore, there is a shift to DHEA at the expense of ASD and DHEAS. Thus, levels of the two hormones cortisol and DHEA remain relatively high in these inflammatory diseases in relation to those of other steroid hormones. This indicates that in RA and ReA, there is no blockade on the level of the 17,20-lyase as suspected in earlier studies in other chronic inflammatory diseases where only DHEAS levels were measured (4, 39). We now believe that it is really important to measure all hormones of the different enzymatic pathways at the same time point in order to get the full picture of changes in steroidogenesis (see Figure 1).

Finally, it was demonstrated that changes in hormone levels, hormone ratios, and hormone-to-cytokine ratios were more pronounced in RA than in ReA. It seems as though ReA patients have an intermediate response pattern relative to RA patients and healthy controls. However, a completely different reaction was not observed. We speculate that the intermediate response pattern in ReA is due to the intermediate serum levels of proinflammatory cytokines, such as IL-6 and TNF. This suggests that the levels of IL-6 and TNF may be important factors for the changes in steroidogenesis. However, there was an interesting observation in RA which was not seen in ReA: serum cortisol levels were high in relation to 17-OH-progesterone when serum TNF levels were low in relation to IL-6. Since TNF can inhibit the secretion of cortisol from adrenal cells due to an inhibition of P450c21 (43), and IL-6 increases cortisol secretion (44, 45) (Figure 1), this particular ratio may determine the secretion of cortisol in relation to 17-OH-progesterone. Interestingly, the interrelationship was only observed in RA. It may be too early to speculate whether this enzymatic step is more sensitive to inhibition by local TNF in RA than in ReA. Changes of local adrenal cytokine levels in the course of rheumatic diseases may be a relevant factor for the alteration of steroidogenesis (46–48).

In conclusion, in ReA patients and even more so in RA patients compared with healthy controls, this study demonstrated that serum levels of cortisol and other adrenal hormones are low in relation to IL-6 and TNF levels. Inadequate secretion of cortisol in relation to IL-6 was associated with a more severe form of arthritis. Due to systemic inflammation, hormonal shifts occur which lead to a preponderance of DHEA and cortisol at the expense of DHEAS, ASD, and 17-OH-progesterone. In RA compared with ReA, this study further points toward an interesting enzymatic step between 17-OH-progesterone and cortisol (P450c21 or P450c11) which may be more sensitive to the influence of TNF. Whether a mild functionally relevant defect in the genes encoding P450c21 or P450c11 may play a role has to be the subject of further studies.

Ancillary