Accelerated Atherosclerosis in Systemic Lupus Erythematosus: Role of Fibroblast Growth Factor 23- Phosphate Axis.

PURPOSE: Despite management advances, accelerated atherosclerotic cardiovascular disease (ACVD) remains a major cause of morbimortality in systemic lupus erythematosus (SLE) patients; that is not fully explained by traditional risk factors. Fibroblast growth factor-23 (FGF23) is a bone-derived phosphaturic hormone with multiple klotho-dependent and independent effects, including promotion of atherosclerosis and vascular calcification, particularly in the context of chronic kidney disease. Increased circulating FGF23 was reported in SLE patients, particularly with lupus nephritis (LN); but its atherogenic role in these disorders was not explored. SUBJECTS AND METHODS: Three study groups of predominantly middle-aged females were categorized by the 2012 SLE International Collaborating Clinics (SLICC) criteria as SLE (without LN), LN, or controls matching for traditional CVD risk profile. Measures of SLE activity, damage, steroid therapy, and glomerular filtration rate were calculated. Fasting blood samples were checked for serum lipid profile, anti-DNA, urea, creatinine, uric acid, proteins, albumin, calcium, phosphorus, C3, C4, CRP, vitamin-D3, intact parathyroid hormone and FGF23 (iFGF23). By carotid ultrasonography, mean common carotid artery intima-media thickness (CC-IMT), plaque score (PS) and internal carotid resistive index (ICRI) were recorded.
RESULTS: CC-IMT, ICRI and serum iFGF23 differed along the study groups (LN>SLE>controls). In both SLE and LN patients, serum iFGF23 had a significant positive correlation with serum phosphorus, CC-IMT and PS. On multivariate analysis, the strongest predictor of increased CC-IMT was cumulative steroid dose in SLE and serum iFGF23 in LN patients. Most significant independent predictors of increased serum iFGF23 were hyperphosphatemia in SLE and proteinuria in LN patients.
CONCLUSION: FGF23-phosphate axis has a key role in accelerated ACVD in SLE patients. Serum phosphorus and iFGF23 should be included in ACVD risk profile assessment of these patients. Prospective studies shall define the role of dietary and/or pharmacologic control of hyperphosphatemia and proteinuria in reducing circulating iFGF23 and ACVD in them.

Introduction

Systemic lupus erythematosus (SLE) is a prototype autoimmune disease predominantly affecting middle-aged women, typically with multiple organ-system involvement, and a protracted course with remissions, exacerbation and cumulative tissue damage. Despite marked geographical disparities, a review of recent epidemiologic studies denoted a trend for increasing global prevalence.1 Although patients` life expectancy has improved in line with advances in diagnosis and treatment, it remains considerably lower compared with the general population.2 Accelerated atherosclerotic cardiovascular disease (ACVD) constitutes the major comorbidity and the leading cause of death, with a frightening risk of myocardial infarction and stroke disproportionately striking young female patients, who are otherwise known to have low ACVD risk.3 It has been long- recognized that this amplified risk cannot be fully accounted for by traditional (Framingham) risk factors, like hypertension, obesity and dyslipidaemia.4 The immunologically geared chronic inflammatory status that typify the disease is a major inducer of ACVD in itself.3 Corticosteroids, that have long constituted the backbone of effective anti-inflammatory therapy, are also atherogenic in the long-term.5 The inclusion of SLE diagnosis and corticosteroid use in the newly developed QRISK3 score has significantly improved its CVD risk prediction in SLE patients.6 Looking for novel, non-traditional, ACVD risk factors, with a focus on recently identified molecular biomarkers, is an evolving area of SLE research.7,8

There is now ample evidence for a link between bone disease and ACVD, both in the general population,9 and in patients with SLE and other autoimmune diseases.10,11 This disturbed bone- vascular axis may result from perturbations of vitamin D,12 parathyroid hormone (PTH),13 or fibroblast growth factor-23 (FGF23).14 Principally secreted from osteocytes and osteoblasts in response to phosphorus loading, FGF23 has been identified as the major phosphaturic hormone (phosphatonin), that decreases proximal renal tubular phosphate reabsorption, intestinal phosphorus absorption, and vitamin D activation.15 It was soon realized that FGF23 is a pleiotropic molecule, with a host of klotho-dependent and independent autocrine, paracrine and endocrine effects on almost all body tissues.16 Increased circulating FGF23 is now recognized as the earliest biochemical abnormality in chronic kidney disease – mineral bone disorder (CKD-MBD).17 Progressive renal impairment is paralleled with an exponential increase in circulating FGF23, approaching several thousands of normal in patients with end-stage renal disease.18

Epidemiologic studies have inked increased circulating FGF23 in subjects with normal renal function with both early functional alterations (impaired forearm flow-mediated dilatation)14 and established carotid ultrasonographic (US) changes (increased intima-media thickness [IMT] and plaque presence)19,20 of atherosclerosis in the community. Similar associations were confirmed in CKD patients.21,22 Therefore, serum FGF23 could improve the power of Framingham risk score to predict increased carotid IMT.23 FGF23 may induce/augment some traditional atherosclerosis risk factors, such as hypertension,24 dyslipidaemia,25 insulin resistance and obesity.26 Of more importance, however, is the association of FGF23 with a host of non-traditional ACVD risk factors such as chronic inflammation,27 hypovitaminosis D,28 and vascular calcification.29 These latter factors become particularly prominent in presence of CKD. Increased FGF23 is also a significant predictor of CKD occurrence30 and progression.31

Two small studies reported significantly higher serum FGF23 in SLE patients compared with controls, with a still higher level in patients with lupus nephritis (LN).32,33 The inflammatory reaction of active lupus upregulates FGF23 production by osteocytes.34 The development of LN is conductive for a further rise of circulating FGF23, possibly induced by hyperparathyroidism,35 hyperphosphataemia,36 impaired renal clearance,37 and klotho deficiency.38 Therefore, LN patients typically have in parallel significantly higher level of FGF23 and higher burden of ACVD, compared with SLE patients without nephritis.33 FGF23 may thus prove to be an important mediator of accelerated ACVD in SLE patients, with a particularly more prominent role in patients with LN. Controlling the circulating level of this key molecule may then become a novel therapeutic approach to reduce cardiovascular morbimortality in SLE patients. Of note, however, studies exploring the cardiovascular risk of circulating FGF23 in SLE patients are lacking.39

Materials and Methods

Study Design and Participants

This was a cross-sectional, case control study, conducted in accordance with the Declaration of Helsinki and approved by the institutional medical ethics committee. Between March and November 2018, eligible participants providing written informed consent were recruited from the outpatient clinics of the Medical Research Institute, Alexandria University, Egypt, and subjected to full clinical evaluation and medical records review, followed by laboratory and imaging studies (Figure 1).

Figure 1

Study flowchart.

Diagnosis of SLE was based on the 2012 SLE International Collaborating Clinics (SLICC) criteria.40 LN was defined by persistent proteinuria > 0.5gm/day in the context of SLE.41 SLE disease activity was assessed by SLE Disease Activity Index (SLEDAI),42 whereas the extent of established damage was assessed by SLICC Damage Index.43 Average steroid dose (Av_S) was calculated as the average equivalent prednisone dose in mg/day. Cumulative steroid dose (Cum_S) was obtained by multiplying Av_S by disease duration in years. Steroid pulses were defined as short-term (3–5 days) intensification of basic steroid dose, mostly given parenterally during in-hospital admission. Traditional atherosclerosis risk factors (age decade, male sex, hypertension, dyslipidaemia, obesity, treatment for these conditions, and presence of ACVD in different vascular beds) were defined by standard criteria and given one point each. The total score (termed Second Manifestations of ARTerial disease, or SMART score) provided a validated semiquantitative estimate of the burden of traditional ACVD risk factors.44,45

We excluded patients with diabetes mellitus, morbid obesity, smoking (current or past), pregnancy, estimated glomerular filtration rate (eGFR) < 30 mL/min/1.73m2, age > 50 years or insufficient data. Finally, 68 subjects (age 20–50 years, 9 males) were triaged into 3 study groups:

G1 (Controls) (N = 20): selected on individual basis to be matching with the other groups in age, sex and SMART score.

G2 (SLE) (N = 20): having SLE without LN.

G3 (LN) (N = 28): having LN.

Laboratory Studies46

After 12 hours fast, blood samples were drawn into EDTA tubes (for complete blood count) and serum separator tubes that were immediately transported and centrifuged, keeping serum at −80 OC until batch analysis was made for levels of serum low-density lipoproteins (LDL), high-density lipoproteins (HDL), triglycerides (TGs), anti-double-stranded DNA (anti-DNA), complement components C3 and C4, C-reactive protein (CRP), urea, creatinine, uric acid, total proteins, albumin, calcium, phosphorus, vitamin D3, intact parathyroid hormone (iPTH) (3rd generation assay), and intact FGF23 (iFGF23) using commercially available kits according to manufacturers` instructions. Morning void urine was used for complete urinalysis. Twenty-four hours urine collections were used for quantitation of proteinuria. eGFR was calculated from stable serum creatinine using CKD-epidemiology collaboration (CKD-EPI) equation.47

Carotid Duplex Ultrasonography48

It was done by an experienced sonographer blinded to the patients` data, using Acuson X300 US System, Siemens Healthineers (formerly Siemens Healthcare), USA. The patient was placed in a supine position with a slight head tilt away from the examined side. A linear-array transducer (VF10-5) with a frequency of 8 MHz was used to scan the carotid system on both sides and calculate the following:

Plaque Score (PS):49 A carotid plaque was defined as either a focal protrusion of the intima-media layer encroaching into the arterial lumen (protuberant plaque), or a diffuse thickening of the intima-media layer measuring > 1.5 mm (diffuse plaque). These were looked for thoroughly in three distinct segments (distal common carotid, bulb, and proximal internal carotid arteries). The total number of plaque-bearing segments on both sides represented the PS, which ranged 0 to 6.

Mean Common Carotid Artery Intima-Media Thickness (CC-IMT):50 The distal segment of the common carotid artery on each side was scanned with the probe in 3 directions (anterior, lateral, and posterior). Each time, a short clip was saved in which the intima-lumen and the media-adventitia interfaces on the arterial far wall were clearly delineated. Some clearly captured image frames were then analysed by the Syngo Arterial Health Package (AHP), an automatic edge detection software, to calculate the mean intima-media thickness along a one-centimetre plaque-free arterial segment. The composite mean CC-IMT was calculated by averaging the 6 readings (3 from each side).

Internal Carotid Artery Resistive Index (ICRI):51 The insonation angle (between transducer beam and axis of blood flow) was kept <60O and the pulse repetition frequency was adjusted to prevent aliasing while maximizing sensitivity and waveform size. Aided by colour flow mapping, the sampling gate was placed completely within the internal carotid flow. The spectral velocity-time curve was recorded while the patient holds breath. The ICRI was calculated as:[(peak-systolic flow velocity – end-diastolic flow velocity) divided by (peak-systolic flow velocity)] (Pourcelot’s equation, measured from the same waveform). Two readings were obtained on each side and the average of the four readings was recorded.

Test-retest variabilities of the ultrasonographic measurements were generally < 5%.

Statistical Methods

Data were analysed using SPSS software package version 20 (SPSS Inc., Chicago, Illinois, USA). Categorical data were expressed as absolute numbers (percentages) and compared by Chi-square or Fisher exact test. Continuous data were tested for normality using Shapiro Wilk test. Parametric data were presented as mean ± SD and compared by analysis of variance (ANOVA). Non-parametric data were presented as median (interquartile range) and compared by Mann–Whitney U or Kruskal–Wallis H-test. Post-hoc analysis was performed for pairwise comparisons. Correlations were tested by the Spearman’s rank correlation coefficient. Models for multivariate linear regression analysis were constructed separately in G2 and G3, introducing clinically relevant predictors with unadjusted P < 0.08. Significance was judged at the 5% level.

Results

The three study groups were comparable regarding sex, age, frequency of hypertension and dyslipidaemia, SMART score, white blood cells, serum TGs, HDL, uric acid, and PS (Table 1). Compared with the other two groups, controls had significantly higher haemoglobin, serum C4 and vitamin D3, and significantly lower serum anti-DNA, CRP, and iPTH. SLE (G2) patients had significantly higher serum proteins and significantly lower platelets than the other two groups. LN patients had significantly higher proteinuria, serum phosphorus and ICRI and significantly lower serum calcium than the other two groups (Figure 2). Compared with controls, LN patients had significantly higher serum urea and CC-IMT (Figure 3). Compared with SLE (G2), LN patients had significantly higher serum LDL and creatinine and significantly lower eGFR and serum albumin. Serum C3 and iFGF23 showed significant changes between the 3 groups (Figure 4). Compared with SLE (G2), LN patients had significantly higher Cum_S and SLEDAI (Table 2).

Table 1

Comparison Between the 3 Study Groups

Parameter	Controls (Cont) (N = 20)	Lupus Only (SLE) (N = 20)	Lupus Nephritis (LN) (N = 28)	Statistical Test	P value (Post- hoc Differences)
Male	3 (15%)	2 (10%)	4 (14.3%)	Fisher exact test	0.57
Female	17 (85%)	18 (90%)	24 (85.7%)
Age (Years)	35.56 ± 7.5	31.15 ± 7.8	36.11 ± 9.3	ANOVA	0.127
HTN	6 (30%)	3 (15%)	11 (39.3%)	Chi square	0.19
No HTN	14 (70%)	17 (85%)	17 (60.7%)
DL	8 (40%)	6 (30%)	15 (53.6%)	Chi square	0.524
No DL	12 (60%)	14 (70%)	13 (46.4%)
SMART Score	5 (4–6)	4.5 (4–5)	4 (4–6)	KWH	0.743
S. LDL (mg/dL)	121.89 ± 21	108.6 ± 15.2	130.96 ± 33.1	ANOVA	0.016* (LN > SLE)
S. TGs (mg/dL)	170.5 (151–188.3)	160.5 (136–171.3)	155 (120–198.3)	KWH	0.311
S. HDL (mg/dL)	44 (41.3–49.8)	44.5 (42.8–47)	45 (41.8–49.3)	KWH	0.828
Haemoglobin (gm/dL)	13.54 ± 0.9	9.87 ± 1.9	10.18 ± 1.4	ANOVA	< 0.001** (Cont > Others)
WBCs (k/uL)	7 (4.5–7.5)	7.8 (3.7–9.5)	6.9 (5.5–9.2)	KWH	0.627
Platelets (k/uL)	288.5 (263.5–309)	217 (192.3–273.5)	297 (245–320.5)	KWH	0.018* (SLE < Others)
S. Anti-DNA (IU/mL)	37 (28.3–43.5)	71 (48.8–102.8)	135 (69.5–210.3)	KWH	< 0.001** (Cont < Others)
S. C3 (mg/dL)	132 ± 29.7	93.85 ± 24.4	56.83 ± 26.4	ANOVA	0.001** (LN < SLE < Cont)
S. C4 (mg/dL)	48.67 ± 9.3	21.33 ± 9.1	18.89 ± 9.9	ANOVA	< 0.001** (Cont > Others)
S. CRP (mg/L)	3.2 (2.4–4)	5.5 (4–7)	7 (5.8–8.3)	KWH	< 0.001** (Cont < Others)
S. Urea (mg/dL)	23.5 (22–30)	31 (27.8–44.5)	48 (29.8–80.3)	KWH	< 0.001** (LN > Cont)
S. Creatinine (mg/dL)	0.9 (0.8–1.1)	0.8 (0.7–1)	1.1 (0.9–1.6)	KWH	0.011* (LN > SLE)
eGFR (mL/min/1.73)	89.2 (66.9–103.3)	99.6 (75.2–120.7)	70.2 (44.7–95.8)	KWH	0.006** (LN < SLE)
S. Uric Acid (mg/dL)	5.2 (4.2–6.3)	4.7 (4.3–5.4)	5.7 (4.8–7.2)	KWH	0.237
Proteinuria (gm/day)	0.24 (0.2–0.3)	0.15 (0.1–0.3)	2.5 (0.86–4.78)	KWH	< 0.001** (LN > Others)
S. Proteins (gm/dL)	6.4 (5.8–7.2)	7.3 (6.6–7.6)	6.6 (6.2–7)	KWH	0.007** (SLE > Others)
S. Albumin (gm/dL)	3.4 (3.2–3.7)	4 (3.4–4.2)	3.2 (2.7–3.7)	KWH	< 0.001** (LN < SLE)
S. Calcium (mg/dL)	10 (9.5–10.5)	10.1 (9–10.6)	8.8 (8.3–8.9)	KWH	< 0.001** (LN < Others)
S. Phosphorus (mg/dL)	3.73 ± 0.6	3.74 ± 0.6	4.53 ± 0.8	ANOVA	0.001** (LN > Others)
S. Vitamin D3 (ng/mL)	34.5 (29.6–40.8)	12.5 (11–15.8)	12 (10.9–13.1)	KWH	< 0.001** (Cont > Others)
S. iPTH (pg/mL)	46.5 (32.5–51.8)	107.5 (84–121.3)	105 (92.3–120)	KWH	< 0.001** (Cont < Others)
S. iFGF23 (pg/mL)	108.5 (92.4–152.2)	409.5 (313–483.2)	771.7 (579.2–900)	KWH	< 0.001** (LN > SLE > Cont)
PS	0 (0–0)	0 (0–1)	0 (0–1)	KWH	0.427
CC-IMT (mm)	0.47 (0.4–0.5)	0.517 (0.464–0.577)	0.541 (0.425–0.661)	KWH	0.01* (LN > Cont)
ICRI	0.61 (0.6–0.6)	0.62 (0.58–0.63)	0.66 (0.61–0.71)	KWH	0.001** (LN > Others)

Notes: Data are expressed as mean + SD or median (interquartile range), *Significant (P < 0.05), **Highly significant (P < 0.01).

Abbreviations: HTN, hypertension; DL, dyslipidemia; S, serum; LDL, low-density lipoproteins; TGs, triglycerides; HDL, high-density lipoproteins; WBCs, white blood cells; k, X1000; CRP, C-reactive protein; eGFR, estimated glomerular filtration rate; iPTH, intact parathyroid hormone; iFGF23, intact fibroblast growth factor-23; PS, Plaque score; CC-IMT, common carotid intima-media thickness; ICRI, internal carotid resistive index; ANOVA, analysis of variance; KWH, Kruskal–Wallis H-Test.

Figure 2

Comparison of serum phosphorus in the 3 study groups. **Highly significant (P < 0.01).

Figure 3

Comparison of common carotid intima-media thickness in the 3 study groups. *Significant (P < 0.05).

Figure 4

Comparison of serum iFGF23 in the 3 study groups. *Significant (P < 0.05), **Highly significant (P < 0.01).

Table 2

Comparison of Lupus-Related Parameters in SLE Patients with and without Lupus Nephritis (LN)

Parameter	SLE (without LN) (N = 20)	LN (N = 28)	P (Mann Whitney U-Test)
Lupus Duration (Y)	3 (2–4.3)	3.5 (2.4–6)	0.194
Average Steroid Dose	22.5 (18.8–30)	30 (20–30)	0.272
Cumulative Steroid Dose	60 (40–105)	90 (60–165)	0.049*
Steroid Pulses	1 (0–1.3)	1 (0–2)	0.217
SLEDAI Activity Index	6.5 (4–9.3)	10 (6–12)	0.046*
SLICC Damage Index	0 (0–1.3)	1 (0–2)	0.628

Notes: Average Steroid Dose: Expressed as average equivalent prednisone dose in mg/day. Cumulative Steroid Dose: Average steroid dose multiplied by disease duration in years. Steroid Pulses: Total number of pulses given throughout the disease history; a pulse is defined as an exceptionally high steroid dose given for 3–5 days. *Significant (P < 0.05).

Abbreviations: SLEDAI, systemic lupus erythematosus disease activity index; SLICC, SLE International Collaborating Clinics damage index.

In both G2 and G3, serum iFGF23 had a statistically significant positive correlation with serum phosphorus, CC-IMT and PS; and these two US parameters significantly correlated with each other (Table 3, Figure 5). Serum iFGF23 had a statistically significant negative correlation with eGFR (in G2), and a statistically significant positive correlation with proteinuria (in G3). Serum phosphorus had a statistically significant positive correlation with CC-IMT and PS (in G2), and a statistically significant negative correlation with eGFR (in G3). Cum_S had a statistically significant positive correlation with CC-IMT in G2; this correlation was reversed and of weaker significance in G3.

Table 3

Statistical Correlations Between Some Parameters in SLE Patients with and without Lupus Nephritis (LN)

		eGFR	Proteinuria	S. Phosphorus	S. iFGF23	Cum_S	PS	CC-IMT
eGFR	r	SLELN	− 0.334	− 0.153	− 0.446	− 0.56	− 0.245	− 0.297
	P		0.15	0.519	0.049*	0.01*	0.299	0.203
Proteinuria	r	− 0.139	SLELN	− 0.184	0.244	0.479	0.245	0.152
	P	0.48		0.437	0.301	0.033*	0.297	0.523
S. Phosphorus	r	−0.458	0.34	SLELN	0.513	0.295	0.552	0.539
	P	0.014*	0.076		0.021*	0.207	0.012*	0.014*
S. iFGF23	r	− 0.31	0.479	0.383	SLELN	0.414	0.675	0.568
	P	0.108	0.01*	0.044*		0.069	0.001**	0.009**
Cum_S	r	− 0.234	0.087	0.115	− 0.12	SLELN	0.263	0.689
	P	0.23	0.659	0.56	0.543		0.263	0.001**
PS	r	0.315	0.295	− 0.054	0.431	− 0.281	SLELN	0.485
	P	0.102	0.128	0.784	0.022*	0.147		0.03*
CC-IMT	r	0.057	0.338	− 0.06	0.422	− 0.396	0.478	SLELN
	P	0.774	0.078	0.76	0.025*	0.037*	0.01*

Notes: *Significant (P < 0.05), **Highly significant (P < 0.01). Correlations in G2 (SLE without LN, N = 20) and G3 (LN, N = 28) are shown in right upper and left lower halves of the table, respectively.

Abbreviations: S, serum; eGFR, estimated glomerular filtration rate; iFGF23, intact fibroblast growth factor 23; Cum_S, cumulative steroid dose; PS, plaque score; CC-IMT, common carotid intima-media thickness.

Figure 5

Correlation of serum iFGF23 with common carotid intima-media thickness (upper) and plaque score (lower) in SLE patients (groups 2 and 3 together). **Highly significant (P < 0.01).

In the regression analysis, the strongest predictor of increased CC-IMT was Cum_S in G2 and serum iFGF23 in G3. The strongest predictor of increased serum iFGF23 was hyperphosphatemia in G2 and proteinuria in G3 (Table 4).

Table 4

Multivariate Linear Regression Analysis for Predictors of Increased CC-IMT and S. iFGF23 in SLE Patients with and without Lupus Nephritis (LN)

A: Outcome: Increased CC-IMT
G2: SLE Patients (N = 20)			G3: LN Patients (N = 28)
Model R2		0.614	Model R2		0.339
Model Significance		0.029*	Model Significance		0.042*
Predictors	β Coefficient	Adjusted P	Predictors	β Coefficient	Adjusted P
Cum_S	0.833	0.05*	S. iFGF23	0.442	0.09
S. iFGF23	0.195	0.475	Cum_S	− 0.053	0.792
S. Phosphorus	0.043	0.86	Dyslipidaemia	0.163	0.43
Haemoglobin	− 0.420	0.089	Proteinuria	0.099	0.64
Dyslipidaemia	− 0.364	0.214
Age	− 0.308	0.292
*Significant (P < 0.05).
B: Outcome: Increased S. iFGF23
G2: SLE Patients (N = 20)			G3: LN Patients (N = 28)
Model R2		0.611	Model R2		0.463
Model Significance		0.005**	Model Significance		0.028*
Predictors	β Coefficient	Adjusted P	Predictors	β Coefficient	Adjusted P
S. LDL	0.072	0.73	SMART Score	0.233	0.277
S. Phosphorus	0.397	0.049*	Dyslipidaemia	0.231	0.209
S. Creatinine	0.276	0.197	Proteinuria	0.466	0.026*
Cum_S	0.314	0.121	S. Creatinine	− 0.046	0.828
			Av_S	0.059	0.762
			S. Phosphorus	0.011	0.954
*Significant (P < 0.05), **Highly significant (P < 0.01).

Abbreviations: CC-IMT, common carotid intima-media thickness; Cum_S, cumulative steroid dose; S, serum; iFGF23, intact fibroblast growth factor 23; LDL, low-density lipoproteins; Av_S, average steroid dose.

Discussion

This study confirms previous reports that serum FGF23 is elevated in SLE patients (more so in LN),32,33 and generates a conceptual framework for its atherogenic role in these patients. We excluded subjects with exaggerated ACVD risk profile because of diabetes mellitus, obesity, advanced renal disease or age. Controls (G1) were individually selected to match with lupus patients (G2,3) in the traditional ACVD risk profile, so that inter-group differences are largely determined by the lupus disease and related factors. We used CKD-EPI equation for eGFR calculation as it performed best in SLE patients, compared with other creatinine-based formulas.52 Subclinical atherosclerosis was evaluated in the carotid arteries by 3 integrative US measures, of proven benefit for quantitation of atherosclerosis extent,53 and progression,54,55 as well as prediction of future CVD events in SLE patients.56 IMT is a sensitive early marker of generalised atherosclerosis.57 Its measurement at the distal common carotid artery offers easier accessibility and higher accuracy and reproducibility, over other carotid segments.58 However, IMT may increase because of media thickening as a function of age and increased blood pressure, rather than atherosclerosis.59 Therefore, the measurement of established focal atherosclerotic protuberances by PS was intended to further improve the quantitation of subclinical atherosclerosis and CVD risk.60 The simplified PS score we adopted may be less prone to assessment errors, and has repeatedly proved useful for quantification of carotid atherosclerosis in SLE patients and assessment of its progression.54,61 Carotid US included a measurement of ICRI, a potentially more sensitive measure for subtle early functional vascular alterations,62 with comparable cardiovascular morbimortality predictor power to the standard IMT.51

Among these predominantly middle-aged females with relatively low traditional ACVD risk profile, both CC-IMT and ICRI showed a graded rise along the 3 study groups (LN > SLE > Controls). LN patients had significantly higher CC-IMT compared with controls, and significantly higher ICRI compared with the other 2 groups. These results conform with the previous reports that SLE patients suffer an increased burden of subclinical atherosclerosis3,5,7,13 and that this burden is much significantly amplified and to a large extent determined, by the presence of LN.63–66 Contrary to some previous studies,67,68 we did not find significant differences between the study groups regarding PS, which may be explained by the relatively young age, low overall ACVD risk and low plaque occurrence, in the study subjects. Serum iFGF23 varied the same way as CC-IMT and ICRI (LN > SLE > Controls), with significant differences between the 3 study groups, also in line with the few previous reports.32,33

Our study revealed multiple possible correlates for the increased ACVD risk in SLE patients. Some correlates encompassed all SLE patients (G 2,3); and some others were particularly distinguished among LN patients. Correlates common for all SLE patients were mainly indicative of the overwhelming immunologically driven, chronic inflammatory nature of the disease; and have been reported by previous studies, like increased serum anti-DNA,69 and CRP,70 and depressed serum C3 and C4.63 Clearly, the lupus-related parameters included in Table 2 also tackle SLE patients as whole; and may constitute an indirect measure of the immune-inflammatory disease pathogenesis, which is thought to progress in parallel with atherogenesis.71 Hypovitaminosis D was also common to all SLE (G2,3) patients, a nearly universal finding in previous studies, that is mainly ascribed to insufficient sun exposure and drug-induced accelerated vitamin D catabolism.12,72–74 Somewhat unexpectedly, PTH was significantly increased in all SLE patients (including those without LN). Increased circulating PTH and its association with ACVD in SLE patients was only recently reported,13 and may be the result of concomitant hypovitaminosis D.75,76

As shown in Table 1, LN patients have, in comparison with other SLE patients, the added risks of proteinuria, renal impairment, hyperphosphatemia and a higher FGF23 elevation. In community studies, both proteinuria,77–80 and GFR decline,81–84 have significant, additive, dose-dependent associations with subclinical atherosclerosis burden and CVD events. These associations were firmly reproduced in studies of LN patients.63–66 Clearly, the development of LN engenders a more aggressive lupus phenotype, evidenced in the present cohort by having significantly lower serum C3, and significantly higher LDL, SLEDAI, and Cum_S, compared with SLE (G2) patients.

An intriguing finding is that Cum_S had a statistically significant positive correlation with CC-IMT in G2 (SLE only) that remained marginally significant (P = 0.05) in the adjusted model, whereas this correlation was reversed (still significant but somewhat weaker) in G3 (LN patients). Corticosteroids in SLE represent a double-edged sword.85 It is possible that their pro- atherogenic effects (insulin resistance, dyslipidaemia, obesity, and hypertension) dominated in the absence of renal disease.5 LN, as a hallmark of more aggressive disease, determined a greater need for the immunosuppressive and anti-inflammatory steroid actions, imparting a better benefit/risk ratio and a net anti-atherogenic action.86 Possibly also the steroid dose was not precisely tailored for the different patient needs, being relatively overdosed in absence, and underdosed in presence, of LN; or there may be endogenous individual variations that determine the patient`s response to these medications.85

The FGF23/klotho axis is the principal regulator of phosphorus metabolism.87 Tight regulation of serum phosphorus is essential since hyperphosphatemia, and even minor elevations of serum phosphorus within the normal range, have been increasingly linked in community studies with impaired vasoreactivity,88 subclinical atherosclerosis,89,90 vascular calcification,91 cardiovascular events,92,93 and mortality.94 Similar strong associations were also found in CKD patients.95–97 Growing evidence for cardiovascular phosphorus toxicity led some authors suggest it the “new cholesterol”.98 Surprisingly, serum phosphorus was rarely reported in SLE patients. Similar to the present study, one study reported serum phosphorus to be significantly higher in LN patients compared with lupus patients without nephritis and controls.99 There is evidence that increased serum phosphorus in LN patients would augment further systemic inflammation, vascular calcification and faster ACVD.100–103 Another study found serum phosphorus to be marginally elevated in SLE patients with atheromatous plaques or increased arterial IMT compared with patients lacking these signs, but the differences were insignificant.13 In the present study, serum phosphorus in G2 (SLE only) had a significant correlation with CC-IMT and PS, although it was within normal range (< 4.5 mg/dL) in all but one patient. A Finnish population study found a significant direct correlation between dietary phosphorus intake and CC-IMT.90 Unawareness about excessive dietary phosphorus consumption is a recently recognised global health problem,104 particularly afflicting less-privileged societies habitually ingesting highly absorbable phosphorus-rich food additives.105

Normally, increased secretion of the counter-regulatory phosphaturic hormone, FGF23, occurs in a robust response to phosphorus loading.15,106 This adaptive FGF23 rise (typically not exceeding 200 pg/mL), utilizing klotho-dependent pathways, generates a controlled phosphaturic response destined to prevent hyperphosphatemia and subsequent perturbations of the bone-vascular axis (bone disease, atherosclerosis, and vascular calcification). A maladaptive FGF23 surge, largely signalling through klotho-independent pathways, approaches much higher values but fails to prevent hyperphosphatemia and rather produces off target effects like hypovitaminosis D, bone rarefaction, ACVD, and myocardial hypertrophy.107,108 A full-blown picture of this maladaptive FGF23 increase occurs in CKD, when klotho deficiency, low GFR, and hypovitaminosis D generate a state of progressive FGF23 resistance;109 leading in advanced CKD to a marked rise of both serum phosphorus and FGF23 in a characteristic strong direct correlation.110,111 A hallmark for the disturbed FGF23-phosphorus axis in SLE patients (G2 and 3) in the present study was the rise of both biomarkers and the presence of a significant direct correlation between them. The maladaptive FGF23 rise was not restricted to LN patients; it involved SLE patients without evidence of renal disease as well. Indeed, G2 patients (SLE only) had a significantly increased serum iFGF23 over controls (exceeding the typical level of adaptive increase and matching with a previous study),32 and an even stronger direct correlation between serum iFGF23 and phosphorus compared to this correlation in LN patients. In multivariate analysis, serum phosphorus persisted in G2 as the most (and only) significant independent predictor of increased serum iFGF23. We thus infer that the maladaptive FGF23 hypersecretion can occur in SLE patients as an effect of the lupus phenotype itself, independently of renal disease. At least two salient, renal function independent, features of SLE can synergistically reduce klotho gene expression and induce significant FGF23 resistance and hypersecretion, namely, chronic inflammation,27,34,112,113 and hypovitaminosis D.114,115 The latter might also result in increased serum PTH which provides another stimulus for FGF23 hypersecretion,116 a possibility supported by the finding of increased serum PTH in SLE patients in the present study and a previous one.13 The issue of FGF23 resistance in SLE requires a separate study.

Previous studies built a consensus that ACVD in SLE mainly (but by no means exclusively) develops in conjunction with LN, with or without renal impairment.63–66 Consistent with this, we found the median CC-IMT in SLE patients, with or without LN, increased over controls; but the difference was statistically significant only with LN patients. Furthermore, we found a significantly increased ICRI in LN patients over the other two groups, denoting a more generalized vascular pathology conductive for atherogenesis in LN patients.62 Serum iFGF23 had a statistically significant positive correlation with CC-IMT and PS, both in G2 and G3 (Table 3), and a similar correlation with ICRI in G2 (not shown). Regression analysis in LN patients revealed that the strongest independent predictor of increased CC-IMT was the increased serum iFGF23; and that proteinuria was the strongest (and the only significant) independent predictor of increased serum iFGF23 in these patients. Therefore, increased FGF23 may, at least in part, explain the strong association between proteinuria and ACVD observed in LN patients;63–66 an association that has been reported as well in type 2 diabetic patients,117 and even in the healthy population.118 The explanation for FGF23 rise in proteinuric nephropathies was based on results of the KNOW-CKD119 and subsequent studies.120,121 Proteinuria decreases the biologic activity of FGF23 on renal tubules, independent of renal function.120 Albuminuria induces a state endoplasmic reticulum stress in renal tubular cells, leading to downregulation of klotho production.122 This leads to FGF23 hypersecretion that fails to correct hyperphosphatemia (more so if renal function is impaired as well). In these cases, an enormous ACVD risk burden would result from the combined effects of increased FGF23, hyperphosphatemia, hypovitaminosis D, and renal impairment.123 Identification of this multifactorial model, with the focal role of the proteinuria-hyperphosphatemia axis, is an important addition to our understanding of the ACVD in SLE patients.

Based on the present work and current knowledge, we constructed a simplified conceptual framework (Figure 6) for the role of FGF23 in the pathogenesis of ACVD in SLE patients. Although not all inclusive, the factors numbered one through ten may be regarded as the key players in this paradigm. Factors in the upper (yellow) pane are nearly always present in all SLE patients since the disease beginning. Inflammation is meant to denote the SLE disease process itself, as well as markers of disease activity and treatment; which continue to have a significant impact on ACVD risk and long-term outcome.3,5,6 Factors in the lower (gold) pane occur mainly or exclusively in patients with LN, when all risk factors become significantly amplified. FGF23 lies at the focal point of intersection of several self-perpetuating (vicious) cycles. For example, increased circulating FGF23 further exaggerates hypovitaminosis D, which is present in most SLE patients.28,72–74 Hypovitaminosis D exaggerates inflammation12,124 and hyperparathyroidism75,76 which independently complete two vicious circles for FGF23 hypersecretion. We can likewise describe another vicious circle in which proteinuria causes hyperphosphatemia and FGF23 hypersecretion;120,121 then FGF23 fails to correct hyperphosphatemia which even continues to progress causing further renal damage,14,30,31 and resistance to the renoprotective effects of angiotensin converting enzyme inhibitors,125 and low protein diet.126 Unless one or more of these vicious circles is interrupted, FGF23 levels would continue to rise inexorably with increased burden of ACVD and further CKD progression. We propose that breaking these cycles to control circulating FGF23 level (for example by controlling hyperphosphatemia and proteinuria) might provide novel approaches for reduction of ACVD risk in SLE patients. It should be noticed, however, that directly targeting the FGF23 itself by neutralizing antibodies in a rat model of CKD-MBD led to a dose-dependent increase in serum phosphorus, aortic calcification, and mortality.127 Therefore, well-designed prospective studies shall define the optimal circulating FGF23 range offering the best compromise between adaptive and maladaptive effects and resulting in the best measures of long-term outcome.

Figure 6

Simplified schema for the role of FGF23 in atherogenesis in SLE patients. Factors in the upper (yellow) part are typical early findings in all (inflammation) or a large proportion of (hypovitaminosis D and dyslipidaemia) SLE patients, whereas the other factors (lower gold part) are either dependent on (proteinuria and renal impairment), or largely determined by, the presence and severity of LN.

To the best of our knowledge, this is the first study to explore the role of FGF23 and related parameters in subclinical atherosclerosis in SLE patients with or without LN. We acknowledge the limitations of the relatively small sample size and the cross-sectional design of the study; so that a cause-effect relationship between the study parameters cannot be readily inferred. The simplified PS we adopted does not account for the multiplicity of plaques in one arterial segment or variations in their area and size, as offered by more elaborate scores.49 Subclinical atherosclerosis assessed in only one vascular bed might not sufficiently reflect its generalized burden. Better exploration of FGF23 actions would have required assessment of dietary phosphorus intake, urinary phosphorus excretion and circulating klotho level.

Conclusion

The FGF23-phosphate axis has a key role in accelerated ACVD in SLE patients. A maladaptive FGF23 hypersecretion coupled with renal tubular resistance to its phosphaturic action operates several self-perpetuating cycles leading to progressive hyperphosphatemia, hyperparathyroidism, hypovitaminosis D and ACVD. Most significant independent predictors of FGF23 hypersecretion were hyperphosphatemia and proteinuria, in SLE patients without and with LN, respectively. Serum phosphorus and iFGF23 should be included in the ACVD risk profile assessment of SLE patients. Dietary and/or pharmacologic control of hyperphosphatemia and proteinuria can be feasible therapeutic targets to reduce circulating FGF23 and ACVD burden in SLE patients. Large-scale prospective studies are needed to define the circulating FGF23 target range in SLE/LN patients providing the best compromise between its adaptive and maladaptive effects, and to assess the effects of reduction of serum phosphorus, iFGF23 and proteinuria on ACVD progression and CVD events in SLE/LN patients. Dietary vigilance for phosphorus-rich food additives may be a readily feasible intervention that can be broadly promulgated in less-privileged communities.