A myelopoiesis gene signature in circulating leucocytes, exemplified by increased myeloperoxidase (MPO) and proteinase 3 (PR3) mRNA levels, has been reported in patients with active anti-neutrophil cytoplasm antibody-associated vasculitis (AAV), and to a lesser extent during remission. We hypothesized that this signature could predict disease relapse. mRNA levels of PR3, MPO, selected myelopoiesis transcription factors [CCAAT/enhancer binding protein α (CEBP-α), CCAAT/enhancer binding protein β (CEBP-β), SPI1/PU.1-related transcription factor (SPIB), spleen focus forming virus proviral integration oncogene, PU.1 homologue (SPI1)] and microRNAs (miRNAs) from patient and control peripheral blood mononuclear cells (PBMC) and polymorphonuclear cells (PMN) were analysed and associated with clinical data. Patients in stable remission had higher mRNA levels for PR3 (PBMC, PMN) and MPO (PBMC). PR3 and SPIB mRNA correlated positively in controls but negatively in patient PBMC. Statistically significant correlations existed between PR3 mRNA and several miRNAs in controls, but not in patients. PR3/MPO mRNA levels were not associated with previous or future relapses, but correlated with steroid treatment. Prednisolone doses were negatively linked to SPIB and miR-155-5p, miR-339-5p (PBMC) and to miR-221, miR-361 and miR-505 (PMN). PR3 mRNA in PBMC correlated with time since last flare, blood leucocyte count and estimated glomerular filtration rate. Our results show that elevated leucocyte PR3 mRNA levels in AAV patients in remission do not predict relapse. The origin seems multi-factorial, but to an important extent explainable by prednisolone action. Gene signatures in patients with AAV undergoing steroid treatment should therefore be interpreted accordingly.
Anti-neutrophil cytoplasm antibody (ANCA)-associated vasculitis (AAV) designates a group of diseases characterized by neutrophil-rich necrotizing vasculitis and the presence of autoantibodies specific for proteinase 3 (PR3) and myeloperoxidase (MPO) [1-3]. AAV predominantly affects the small vessels in the human body and can be classified into microscopic polyangiitis (MPA), granulomatosis with polyangiitis (GPA) and eosinophilic granulomatosis with polyangiitis (EGPA) . MPO-ANCA are more frequent in MPA and EGPA, and at least in North America and Europe PR3-ANCA occur more frequently in GPA patients . Without treatment these diseases often have a fatal course, but today remission can be achieved in a majority of cases with induction therapy consisting of cyclophosphamide or rituximab. However, remission does not mean cure and, despite widespread use of maintenance therapy, relapses are common. Long-term follow-up of large international multi-centre trials show that about half of the patients experience a relapse within 5 years . Some features at presentation are linked to an increased risk of relapse, such as PR3-ANCA positivity, preserved glomerular filtration rate and lung engagement, but on the individual level relapses cannot be predicted accurately .
When comparing AAV patients in remission with healthy controls, several parameters have been shown to differ. These include circulating cytokines , T cell subpopulations [9-11], chemokines , B cell subpopulations [13, 14], neutrophil subpopulations [15, 16] and circulating adhesion molecules .
Some of these parameters, such as interleukin (IL)-10 levels and proportions of regulatory B cell subpopulations, are actually higher during remission compared to flares and tend to ‘normalize’ along with increasing disease activity [8, 13]. These studies indicate that remission is not simply a state of no disease activity, but instead an active process keeping an underlying disease process under control. McKinney et al. described a transcription signature in CD8+ T cells that could predict the likelihood of relapse in AAV patients and that was enriched in genes involved in the IL-7 receptor pathway, T cell receptor signalling and those associated with T cell memory .
A feature of particular interest seen in patients with AAV is the presence of certain messenger RNA (mRNA) signatures in circulating leucocytes [18-21]. Yang et al. showed that leucocytes from active AAV patients expressed several genes normally expressed only during myelopoiesis (a myelopoiesis gene signature) . The highest fold change was found for elastase 2, but the signature also included the ANCA antigens PR3 and MPO. This gene signature was also present during remission, but to a much lesser extent. Cheadle and co-workers confirmed this work and found a correlation of the signature with disease activity and PR3 expression in peripheral blood mononuclear cells (PBMC) among GPA patients . We studied PR3 expression in isolated monocytes and found pronounced differences between stable AAV patients and controls, but did not see any correlation with disease activity .
The hypothesis of the present study was that peripheral leucocyte expression of PR3 and/or MPO could serve as a marker of subclinical disease activity predicting future relapses. The aim was to exploit this parameter as proxy when screening for new biomarkers. Therefore, we measured PR3 and MPO levels in polymorphonuclear cells (PMN) and PBMC by quantitative real-time polymerase chain reaction (qPCR) in our patient cohort of GPA and MPA patients and in healthy controls. We also determined the levels of the myelopoiesis-regulating transcription factors CCAAT/enhancer binding protein α (CEBP-α), CCAAT/enhancer binding protein β (CEBP-β), spleen focus-forming virus proviral integration oncogene (PU.1 homologue) (SPI1) and SPI1/PU.1-related transcription factor (SPIB). To investigate the regulation/dysregulation of PR3 mRNA we measured levels of selected miRNAs, small round 22 nucleotide long non-coding RNAs that regulate the stability of mRNAs . Many genes are regulated by miRNA, and several miRNAs are critical for myelopoiesis. The mRNA and miRNA expression levels were then associated with clinical patient data.
Material and methods
Patients and controls
Patients with GPA and MPA visiting the out-patient clinics at the Departments of Rheumatology and Nephrology at the University Hospital in Linköping were asked to participate in the study. Patients were diagnosed using the Chapel Hill nomenclature and the distinction between MPA and GPA was made using the European Medicines Agency (EMA) algorithm [1, 2, 25], where patients with renal-limited AAV are classified as having MPA. Disease activity was recorded using the Birmingham Vasculitis Index version 03 (BVAS03). Clinical data, including medication and blood chemistry at the time of sampling, were obtained from patient files. Values for C-reactive protein (CRP) and blood cell counts were recorded only if taken on the day of sampling, while values for serum creatinine, ANCA and erythrocyte sedimentation rate were accepted from blood drawn up to 4 weeks from sampling in stable patients. Doses of methylprednisolone and betamethasone were converted to prednisolone according to 1 mg methylprednisolone = 1·25 mg of prednisolone and 1 mg betamethasone = 8 mg of prednisolone. After sampling, patients were followed prospectively and BVAS was assessed at each visit. Relapse was defined as a BVAS of ≥2. Patients were followed for up to 23 months until December 2012. Blood was also drawn from healthy controls; only sex and age were recorded.
PMN and PBMC were separated using either a Percoll separation protocol (modified from ) or a Polymorphprep/Lymphoprep separation protocol. The cell separation method did not affect intracellular mRNA/miRNA levels, which was shown by comparing the two methods in a subgroup of the samples (data not shown). In the Percoll separation protocol the blood was centrifuged at 1500 g for 15 min and the buffy coat was collected. A discontinuous Percoll (GE Healthcare Bio-Sciences, Uppsala, Sweden) density gradient was prepared by gently overlaying 5 ml 63% Percoll solution on top of 5 ml 72% Percoll solution. The buffy coat was layered over the Percoll gradient and centrifuged at 490 g for 25 min. PMN and PBMC were collected and washed twice in phosphate-buffered saline (PBS) by centrifugation at 300 g for 5 min. Thereafter, the cells were counted and stored in RNAlater® (Ambion, Austin, TX, USA) at −20°C. In the Polymorphprep/Lymphoprep protocol the blood was layered over 3 ml Polymorphprep (Fresenius Kabi Norge AS, Oslo, Norway) and 2 ml Lymphoprep (Fresenius Kabi Norge AS) and centrifuged at 480 g for 35 min. PMN, together with some adjacent red blood cells (RBC), and PBMC were collected and washed in 1 vol. 0·45% NaCl + PBS at 400 g for 10 min. Thereafter, RBC were lysed by adding 10 ml NH4Cl lysis buffer and after 5–10 min at room temperature the samples were centrifuged at 300 g for 5 min. For PMN the lysis step was repeated once. Finally, the cells were washed in PBS (centrifuged at 300 g for 5 min), counted and stored in RNAlater® at −20°C. The purity of the PBMC fraction was analysed by flow cytofluorometry and revealed < 1% contamination with neutrophils in the patient and the control group.
Total RNA (including miRNA) was extracted from the samples using the mirVana miRNA Isolation Kit (Ambion), according to the manufacturer's protocol. Briefly, ethanol was added to the samples and they were passed through a glass-fibre filter, which immobilizes the RNA. The filter was washed and then the RNA was eluted with preheated nuclease-free water. The samples were DNase-treated with the DNA-free kit (Ambion) and then stored at −70°C. The concentration and purity of the RNA was determined by a NanoDrop 1000 spectrophotometer (version 3·8; Thermo Fisher Scientific, Waltham, MA, USA). The RNA integrity was analysed by agarose gel electrophoresis, and some of the samples were assessed further with Agilent 2100 BioAnalyzer (Agilent Technologies, Santa Clara, CA, USA).
qPCR for myelopoiesis genes and transcription factors
To prepare cDNA from RNA isolated from PBMC or PMN the high-capacity cDNA reverse transcription kit (Applied Biosystems, Grand Island, NY, USA) was used and the reverse transcription run in a Mastercycler Personal (Eppendorf, Hamburg, Germany) with the program parameters: 10 min 25°C, 120 min 37°C, 5 min 85°C.
qPCR was run on cDNA of the myelopoiesis genes PRTN3 (TaqMan Gene Expression Assay: Hs01597752_m1; Applied Biosystems) and MPO (Hs00924296_m1), of the transcription factor genes CEBP-α (Hs00269972_s1), CEBP-β (Hs00270923_s1), SPIB (Hs00162150_m1) (PU.1-related) and SPI1 (Hs02786711_m1) (PU.1) as well as of the reference gene glyceraldehyde 3-phosphate dehydrogenase (GAPDH) (Hs99999905_m1). PCR reactions were prepared initially in triplicate and then distributed as 3 × 10 μl reactions onto a 384-well plate by a BioMek 2000 automatic pipetting robot (Beckman Coulter, Brea, CA, USA). Each 10 μl reaction contained 2·5–10 ng cDNA, 0·5 μl 20× TaqMan Gene Expression Assay (see above), 5 μl 2× TaqMan Universal Master Mix II with UNG (cat no. 4440038; Applied Biosystems) in nuclease-free water. The PCR was run in a 7900HT real-time PCR system (Applied Biosystems) using the standard run with the parameters: 2 min 50°C, 10 min 95°C, 40 cycles 15 s 95°C, 1 min 60°C.
qPCR results were analysed using SDS version 2·4 and RQ Manager version 1·2.1 software (both from Applied Biosystems) with automatic baseline setting and a manual setting of the threshold value at 0·1 (The threshold value is the baseline value subtracted from the ratio of reporter fluorescence and reference dye fluorescence. It is placed within the exponential phase of the PCR amplification curve.). Individual replicates were excluded based on the standard settings of the software for ‘a well is empty’, ‘a well is not amplified’ and ‘large mean squared error’. Further, if the standard deviation of the replicates' Ct values was >1·2, the replicate with the highest deviation was excluded. For each sample a Ct value relative to the reference gene GAPDH (ΔCt = Ct, sample–Ct, GAPDH) was calculated.
GeneChip miRNA 2·0 arrays
The FlashTag biotin HSR RNA labelling kit for Affymetrix GeneChip miRNA arrays (Genisphere, Hatfield, PA, USA) was used in accordance with the manufacturer's instructions to add a poly(A) tail and biotin-labelled 3DNA dendrimers for signal amplification to about 0·5 μg total RNA (four controls, four GPA patients). The biotin labelling was confirmed by enzyme-linked oligosorbent assay (ELOSA) (Genisphere). Then, the labelled RNAs were hybridized onto individual Affymetrix GeneChip miRNA 2·0 arrays (Affymetrix, High Wycombe, UK), according to the manufacturer's instructions. The GeneChip miRNA 2·0 array covers all sequences from the miRBase v15, i.e. it contains 15 644 probe sets of 131 organisms. For Homo sapiens a total of 1105 mature miRNAs and 1105 pre-miRNAs are included on the chip. The chips were washed and stained using a Fluidics Station 450, and scanned in an Affymetrix Scanner 3000 using GeneChip Operating Software (GCOS) and the Affymetrix GeneChip protocols. The efficiency of the miRNA runs was verified by checking the intensities of spike oligos that were included in the FlashTag biotin HSR labelling kit using the Affymetrix miRNA QCTool version 1·1.1.
Analysis of GeneChip miRNA 2·0 array results
Affymetrix CEL files were imported into Affymetrix miRNA QCTool version 1·1.1 and robust multi-array averaging (RMA) global background correction was performed. Then, the raw data were normalized using quantile normalization over the entire array, according to the manufacturer's guidelines. The data were exported to Microsoft Excel and only the human miRNAs were analysed further by calculating the coefficient of variation (CV) and logarithmic fold changes in the four patients compared to the four healthy controls. The miRNAs were finally sorted according to high fold change, high intensity, low CV and true detection. The data were also analysed using the ‘Affy’ package for miRNA microarray data analysis from the R-Bioconductor software suite (http://www.bioconductor.org).
Rowttests from the package ‘genefilter’ was used to calculate t-statistics to find miRNAs expressed differentially in patients compared to healthy controls. For different patient group comparisons, different statistical significance cut-off levels were set. The minimum stringency was P < 0·05. To control for multiple comparisons, Benjamini–Hochberg (BH) false discovery rate (FDR) correction was performed on the miRNAs after t-tests.
Selection of miRNAs
The miRNAs were selected in a multi-step process using results from the miRNA array experiments. Candidate miRNAs were selected based on maximum disparity between patients and controls in order to find deregulated miRNAs and maximum similarity to find miRNAs suitable for normalization. miRNAs were also selected based on results from measurements of plasma miRNAs in another study . The selection process resulted in the list presented in Supplementary Table 1.
qPCR for miRNA
The RNA samples were reverse transcribed into cDNA using the TaqMan microRNA (miRNA) reverse transcription kit (Applied Biosystems), according to the manufacturer's protocol. Specific reverse transcription (RT) primers from different Taqman miRNA assays (cat. no. 4440887; Applied Biosystems) were also included (for details see Supplementary Table 2). Each 15 μl reaction consisted of 5 μl (10 ng) total RNA, 7 μl Master Mix (100 mM dNTPs, 50 U/μl MultiScribe reverse transcriptase, 10× reverse transcription buffer, 20 U/μl RNase inhibitor and nuclease-free water) and 3 μl of 5× RT primer. The reverse transcription was performed using a thermocycler (Mastercycler personal; Eppendorf, Hamburg, Germany) with the program: 30 min at 16°C, 30 min at 42°C and 5 min at 85°C. The cDNA was stored at −20°C until further use.
qPCR was performed on cDNA from both PMN and PBMC using a 7900HT fast real-time PCR system (Applied Biosystems) with absolute quantification and different TaqMan miRNA Assays (cat. no. 4440887; Applied Biosystems) as well as Taqman universal PCR Master Mix II with UNG (Applied Biosystems). Each TaqMan miRNA assay contains one tube with a miRNA-specific RT primer, used in the RT-reaction, and one tube for the qPCR reaction with a mix of a miRNA-specific forward PCR primer, a specific reverse PCR primer and a miRNA-specific TaqMan minor groove binder (MGB) probe. Reactions of 10 μl, including 0·67 μl cDNA, 5 μl Master Mix, 0·5 μl miRNA assay and 3·8 μl nuclease-free water, were transferred onto a 384-well plate by an automated pipetting robot (BioMek 2000; Beckman Coulter). The plate was loaded into the instrument and run with the thermal cycling conditions 50°C for 2 min, 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. Each sample was run in triplicate and a standard curve was included on each plate, as well as a positive control and a no template control. For each miRNA studied, all samples from the same cell type were run on the same plate.
The results of the qPCR run were analysed in the software package SDS version 2·4 (Applied Biosystems). The analysis settings were automatic baseline and a manual threshold value set to 0·1. The data were transferred to Microsoft Office Excel 2007 and the mean of the triplicates as well as standard deviation (s.d.) were calculated. Samples with s.d. > 1 Ct were excluded. A median normalization with 11 stably expressed miRNAs was performed for PMN, and for PBMC the Ct-values were normalized with miRNA RNU48. In addition, as an alternative normalization for PBMC, median normalization with four stably expressed miRNAs was performed, giving comparable results to those achieved using normalization with RNU48.
Statistical comparisons between patients and controls as well as graphs were performed with GraphPad Prism 5 and IBM SPSS statistics version 20. For correlations between myelopoiesis genes and transcription factors, Pearson's product–moment correlation was calculated, while for correlations involving miRNAs the non-parametric Spearman's correlation was applied.
The study was approved by the regional ethical review board in Linköping and all patients and controls provided written informed consent.
A total of 67 patients, 44 with GPA and 23 with MPA, were included into the study. The median age at sampling was 68 years, and the median disease duration at the day of sampling was 6·7 years (Table 1). The patients had previously experienced a mean number of 1·25 relapses, and the median time since last flare (relapse or time of diagnosis) was 35·7 months. Most patients had ongoing therapy against vasculitis, 72% (n = 48) received oral steroids, 54% (n = 36) received disease-modifying anti-rheumatic drugs (DMARDs) and only 13% (n = 9) of the patients received no immunosuppressive therapy. In 43% (n = 29) of the patients the latest induction therapy had been rituximab. At sampling, six patients had active disease while the remaining 61 were in complete remission (BVAS = 0). Thirty-six healthy control subjects (17 male/19 female) with a median age of 44 years at sampling took part in the study.
Table 1. Patient characteristics at time of sampling.
GPA versus MPA
*A patient with p-ANCA at diagnosis in 1997 and no available positive enzyme-linked immunosorbent assay (ELISA) results. Healthy controls: male/female, 17/19; age (years) median (IQR), 44 (27–53). eGFR: estimated glomerular filtration rate; ESR: erythrocyte sedimentation rate; GPA: granulomatosis with polyangiitis; IQR: interquartile range; MMF: mycophenolate mofetil; MPA: microscopic polyangiitis; ANCA: anti-neutrophil cytoplasm antibody; MPO: myeloperoxidase; PR3: proteinase 3.
mRNA levels for the PR3 gene PRTN3 were elevated in both PBMC (mean fold change 3·2, P < 0·001) and PMN (fold change 1·8, P = 0·012). MPO mRNA was also elevated in PBMC (fold change 1·5, P = 0·04), but not in PMN [fold change 1·0, not significant (n.s.)]. mRNA levels for PR3 and MPO correlated significantly with each other in both cell types in patients (r = 0·78 and 0·64, respectively) as well as in controls (r = 0·83 and 0·77, respectively) (Table 2).
Table 2. Significant correlations between mRNA for the proteinase 3 (PR3) gene, PRTN3, and selected transcriptions factors as described in Methods.
To investigate the regulation of PR3 mRNA we measured mRNA levels for a selected number of transcription factors known to have critical roles in myelopoiesis. In PMN we found, as expected, that mRNA for PR3 correlated with SPI1 (PU.1) and SPIB in both patients and controls (Table 2). PR3 mRNA in PBMC from controls exhibited a positive correlation with mRNA for SPIB. However, in patients this correlation was negative.
Correlations between miRNAs and mRNA for PR3 or MPO
In controls we found several statistical correlations between PRTN3 mRNA levels and miRNA levels measured by qPCR, three positive correlations in PBMC and five in PMN, as well as four negative correlations in PMN (Table 3). None of these correlations were present in patients, and no other significant correlation was seen.
Table 3. Correlation between PRTN3 mRNA and selected miRNAs.
The finding regarding the MPO gene mRNA was similar to that of PR3, with five significant correlations in PBMC controls (three in PMN controls) and no significant correlations in AAV patients (data not shown).
Correlation between mRNA for PR3 or MPO and clinical data
The six patients with active disease had higher levels of mRNA for PR3 and MPO in PBMC compared to patients in remission (fold change 10·3 versus 2·8 and 4·7 versus 1·3), but the difference was statistically significant only for MPO (P = 0·06 and P = 0·04). In PMN the mRNA levels were not statistically different.
During follow-up, 15 of the 61 patients who were in remission developed a relapse. When comparing patients eventually developing relapse with those who did not, no difference was seen in mRNA for PR3 or MPO. However, blood sampling and mRNA level analysis was performed only once and not at the time of relapse. Those who subsequently developed a relapse had a 2·8-fold change for PR3 mRNA in PBMC compared to 2·9 for those who remained in remission until the end of follow-up or death (n = 2). The corresponding figures in PMN were 2·5 and 1·6. Similarly, there did not seem to be any correlation with previous relapse tendency and PR3 mRNA levels. When the patients in remission were divided into thirds based on PR3 mRNA levels, patients in the highest group had a mean number of relapses of 1·6 compared to 1·2 (n.s.) for the patients in the lowest group (Table 4).
Table 4. Clinical features and the proteinase 3 (PR3) gene; 61 anti-neutrophil cytoplasm antibody-associated vasculitis (AAV) patients in remission were divided into three groups based on mRNA levels for PRTN3.
While previous or subsequent relapses were not associated significantly with PR3 mRNA levels, several other of the clinical features were. High PR3 mRNA in PBMC was associated with reduced estimated glomerular filtration rate (eGFR), high blood leucocyte count, short time since last episode of active disease and male sex. Conversely, there was no detectable correlation to ANCA positivity or general inflammation, as measured by erythrocyte sedimentation rate (ESR) or C-reactive protein (data not shown).
We also detected correlations between PR3 mRNA and treatment. Patients who previously received rituximab had higher PR3 mRNA levels (compared to patients who received cyclophosphamide), as did patients with ongoing prednisolone therapy. Figure 1a illustrates the relationship between different regimens of maintenance therapy and PR3 mRNA levels. Patients treated with mycophenolate, azathioprine or methotrexate exhibited similar PR3 mRNA levels. However, in each group (apart from the azathioprine group), patients on prednisolone had higher levels compared to those not receiving oral steroids. The difference was statistically significant only within the group without any DMARD (P = 0·0057 for PR3 mRNA and P = 0·016 for MPO mRNA). Also, when patients were divided into groups based on previous rituximab therapy there was a significant difference between those on and off steroids within the rituximab-treated group (P = 0·011 for PR3 mRNA and P = 0·029 for MPO mRNA) (Fig. 1b and data not shown). Figure 1c depicts the relationship between prescribed daily steroid dose and PR3 mRNA levels. Even though most patients were on small or moderate doses of prednisolone, the differences were statistically significant compared to those without steroids. Furthermore, a dose–response relationship between prednisolone dose and PR3 mRNA levels was seen. A similar pattern was found for MPO mRNA levels, but a statistically significant difference was found only between patients receiving no prednisolone and those receiving doses above 7·4 mg/day (Fig. 1d).
There was also a significant negative relationship between time since last flare (i.e. time between last flare and blood sampling) and PR3 mRNA levels (rho = −0·38, P = 0·003). The relationship is depicted in Fig. 2a. All patients with recent flares were on prednisolone; the shortest time since flare for a prednisolone-free patient was 18 months. Among patients without prednisolone there was no correlation between time since last flare and PR3 mRNA levels. The negative correlation was found only in patients receiving prednisolone (rho = −0·34, P < 0·05). As expected, prednisolone doses were reduced gradually during stable remission [< 18 months, 10·3 (± 16·6) mg/day, n = 18; 18–36 months, 2·7 (± 2·6) mg/day, n = 16; >36 months, 2·1 (± 2·7) mg/day, n = 33].
PR3 mRNA levels also correlated with renal function (rho = −0·40, P < 0·01). Patients with low eGFR had higher PR3 mRNA levels, as shown in Fig. 2b. The negative correlation was present only in patients treated with prednisolone (rho = −0·42; P < 0·01). However, no correlation was found between prednisolone dose and renal function. In patients without this treatment there was no statistical correlation with eGFR (rho = −0·07).
There was a strong correlation between PR3 mRNA in PBMC and leucocyte count in patients. This correlation was seen both in patients treated with prednisolone and untreated patients (rho = 0·43; P < 0·05 and rho = 0·60; P < 0·05, respectively) (Fig. 2c). Leucocyte count was also associated with prednisolone dose (rho = 0·42, P = 0·0019). The leucocyte count was not measured in controls.
Steroid doses, myeloid genes, transcription factors and miRNAs
In PBMC the mRNA levels of the myeloid genes PRTN3 and MPO correlated with the daily steroid dose (Table 5). Additionally, SPIB was found to be correlated negatively with steroid dose in both PBMC and PMN. Steroid doses correlated only with levels of a few miRNAs. Using non-parametric testing two miRNAs reached significance in PBMC and three in PMN as shown in Table 6. All correlations were in the same direction; greater steroid doses correlated with lower levels of the respective miRNA. No statistically significant differences for mRNA levels of the transcription factors (CEBP-α, CEBP-β, SPIB, SPI1) were found between patients in remission and patients with active disease, either in PBMC or in PMN. When we compared the levels of the above transcription factors in patients receiving steroids and those not receiving immunosuppressive treatment, we found that in prednisolone-treated patients SPIB was decreased significantly in PBMC (mean fold change 2·8, P < 0·0001) and SPI1 was reduced significantly in PMN (mean fold change 1·1, P = 0·041).
Table 5. Correlation between daily doses of prednisolone and myelopoiesis genes as well as transcription factors.
In this study we confirm previous findings regarding up-regulation of mRNA levels for the PR3 and MPO genes in peripheral blood leucocytes [18, 22]. This reflects a general up-regulation of genes normally expressed during myelopoiesis, and has been referred to as a myelopoiesis signature. This expression signature has been reported to vary with disease activity, but can also be found in many, but not all, patients in remission. In our study, up-regulation was found in both PBMC and PMN but was most pronounced in PBMC. This is in line with previous reports, including our own, regarding the PR3 mRNA expression in isolated monocytes .
Approximately 50% of patients with AAV relapse within 5 years, despite widespread use of maintenance therapy. Today we have a limited possibility to predict who is going to be affected by a disease flare and when it will happen. We hypothesized that elevated levels of mRNA for PR3 might be an indication of risk of relapse, which could be useful in the development of biomarkers predicting relapses in AAV. Unfortunately, the results of the present study show that elevated mRNA for PR3 is not linked to relapses. There was no difference in the number of subsequent relapses among those with elevated mRNA for PR3 compared to those with normal amounts. This was true for both PBMC and PMN. Furthermore, elevated PR3 mRNA was not linked to factors known to predict relapse, such as previous relapses, positive PR3-ANCA, GPA phenotype and non-elevated serum creatinine levels .
When exploring the regulation of mRNA levels for PR3 and MPO, we examined levels of selected transcription factors. We found a positive correlation with SPIB in PBMC in healthy controls but a negative correlation in AAV patients. In a similar manner, we found significant correlations between mRNA levels of PR3 and MPO with several miRNAs in healthy controls, but these correlations were attenuated or totally absent in patients. This deregulation of mRNA levels in patients might be explained by epigenetic events in myelopoietic precursor cells, a phenomenon that has been reported to occur in AAV . An alternative explanation could be an exogenous factor, such as a pharmacological agent.
When PR3 mRNA levels were associated with clinical data from the time of sampling, we found several statistically significant correlations. We found correlations with renal function (eGFR), leucocytosis, time since last flare of disease and medication. The most striking association was with ongoing low-dose prednisolone therapy. Overall, there was a clear dose–response relationship between prednisolone and mRNA levels for PR3 in PBMC and the difference was already statistically significant, with doses around 5 mg/day (for MPO at doses >7·4 mg/day). Patients were treated with a variety of DMARDs, but within each treatment regimen (apart from azathioprine) those with concomitant steroid therapy had higher PR3 mRNA levels compared to those without any steroids. One way in which glucocorticoids (GCs) may affect gene expression is via glucocorticoid response elements (GREs). However, Basic Local Alignment Search Tool (BLAST) searches in the published gene sequences of PRTN3, MPO, SPIB and SPI1 identified no GREs. GCs act not only at the transcription level via GREs, but have a variety of other genomic and non-genomic effects that may account for the findings in our study [29-31]. SPIB and SPI1 were found to regulate the glucocorticoid receptor promoter in lymphoblast cells . An involvement of SPIB in chemosensitivity involving prednisone treatment has been shown in a study on acute lymphoblastic leukaemia in which leukaemia cells of chemosensitive patients showed higher levels of SPIB mRNA than in chemoresistant patients . The intricate effect of steroids on gene regulation is highlighted by our finding that prednisolone-treated AAV patients showed lower SPIB mRNA levels than patients without such treatment [34-37].
Patients previously given induction therapy with rituximab had higher PR3 mRNA levels compared to those who had received induction therapy with cyclophosphamide instead. Within the rituximab-treated group there was a significant difference between patients with and without prednisolone, and consequently rituximab therapy does not explain the correlation between prednisolone and elevated mRNA levels for PR3. Furthermore, patients treated with rituximab as induction therapy also received higher prednisolone doses compared to patients with cyclophosphamide induction therapy (data not shown).
We found a negative correlation between eGFR and mRNA levels for PR3 in PBMC. When we divided the patients into two groups based on prednisolone therapy we found a negative correlation only in those receiving prednisolone.
Furthermore, there was a negative correlation between time since last flare and mRNA levels for PR3. As it is common practice to reduce prednisolone therapy slowly over time, this relationship could be a consequence of differences in prednisolone doses administered. However, we cannot rule out a relationship going in the other direction, but in patients without prednisolone the correlation was very weak (rho = −0·13) and not statistically significant.
Prednisolone is well known to induce leucocytosis early after start of treatment and is usually very pronounced when large doses are given. However, leucocytosis persists and can also be seen during long-term treatment with small doses. The mechanisms accounting for steroid-induced leucocytosis have not, to our knowledge, been studied recently at the molecular level, and most reports on this subject stem from the previous century [38-40]. Steroid-induced leucocytosis consists mainly of PMN, but monocytes are also increased. The numbers of other cell types, such as eosinophiles, are decreased. The effect of high-dose methylprednisolone on myelopoiesis has been described previously (for review see ).
The correlation between leucocytosis and mRNA for PR3 in PBMC was also observed in patients not treated with prednisolone, indicating that multiple factors affecting leucocytosis will also affect mRNA for PR3.
In this study we have not investigated the cell type responsible for the increase in PR3 mRNA seen in our PBMC fraction. Monocytes are known to contain PR3, and as we have previously found an increased PR3 mRNA level in isolated monocytes, we assume that these cells are the main source of PR3. However, a subset of neutrophils has been described in SLE patients [42-45] that co-purified with the PBMC fraction during Ficoll/Hypaque density centrifugation. These low-density granulocytes (LDGs) could be distinguished and isolated from the true PBMC by use of surface markers. LDGs may be responsible for the high levels of neutrophil-specific genes in PBMC preparations seen in some SLE patients. The percentage of LDGs has been reported to be as high as 17% . LDGs have never been described in AAV, and in our previous studies on monocyte subpopulations in AAV we did not detect any contaminations of neutrophils in that range [23, 46]. The PBMC fraction of patients and controls in our present study was more than 99% pure according to flow cytofluorometry analysis. However, we cannot rule out the possibility that small numbers of high-producing LDGs account for the bulk of the PR3 mRNA expression in this study. The LDGs are thought to harbour pathogenic potential in SLE, such as increased propensity for neutrophil extracellular trap (NET) formation . This would not be consistent with our findings of a dose–response relationship with prednisolone treatment.
When we looked at the association between prednisolone and miRNAs we found only a few correlations in our miRNA subset, and they were all negative. Among those was miR-155, which is considered necessary for normal immune function and for regulation of transcription factors CEBP-β and PU.1 [47-49]. Also, proinflammatory effects of miR-155 and a role in autoimmunity have been described [50, 51]. Down-regulation of miR-155 by GCs was shown previously to occur in lipopolysaccharide (LPS)-induced macrophage inflammatory response in mice via the glucocorticoid receptor (GR) and the nuclear factor (NF)-κB pathway , but recently also in activated human CD4+ T cells . GCs not only influence miRNA transcription but, in turn, the GR expression also is regulated by miRNAs . It is, however, necessary to stress that these correlations we discovered were found after multiple testing and, if true, could reflect differences in cell subpopulations as well as changes in intracellular levels within distinct cell types.
We have studied mainly patients in remission; only a few patients in our cohort had active disease at the time of sampling and our study is therefore not powered for patients with active disease. Consequently, we cannot draw any conclusions regarding the origin of the myelopoiesis pattern during flares of AAV. Our results indicate that mRNA of PR3 in PBMC is dependent upon many factors; in remission exogenous prednisolone seems to be a major player, making it difficult to analyse other factors. Patients with active disease had in most cases received high doses of prednisolone for several days prior to blood sampling. Furthermore, we cannot exclude that PR3 mRNA levels rise only shortly before relapse. To elucidate this possibility and to detect possible fluctuations in PR3 mRNA levels, repeated patient blood sampling would be required. For our study we only used one blood sample per patient, which allows us to conclude that PR3 levels in patients who subsequently relapse are not generally high, and that the PR3 level is therefore no predictor of relapse. A limitation of our study is that it does not include any disease controls. This is due to the fact that our study was initially not intended as a pharmacological study. Therefore, our findings may not be applicable to other diseases treated with steroids. More research is needed before firm conclusions can be drawn. Until then, the possibility of an effect of steroids on myelopoiesis genes should be considered when interpreting gene signatures of AAV patients undergoing steroid treatment.
In conclusion, mRNA for PR3 is elevated in peripheral blood leucocytes, especially in PBMC, in patients with AAV in remission. Our data are compatible with the notion that a large part of this is caused by pharmacological intervention, most importantly treatment with prednisolone.
We would like to acknowledge the support of the Swedish Rheumatism Association, the Ingrid Asp Foundation, the Swedish Kidney Association, and the Östergötland County Council Research Foundation. We would further like to express our thanks to Marianne Petersson for patient care and to all patients and healthy blood donors who gave their consent for the use of their data in this research project. For linguistic help we thank Professor Wayne Jones.
The authors have no financial conflicts of interest.
P. E. and M. S. designed the study. T. K., M. W., C. S. and S. B. performed research. T. K., M. W., C. S., S. B., P. E and M. S. analysed the data. T. K., M. W., C. S., P. E and M. S. wrote the paper.