A retrospective case control study identifies peripheral blood mononuclear cell albumin RNA expression as a biomarker for non-alcoholic fatty liver disease

Abstract

Purpose

Non-alcoholic fatty liver disease (NAFLD) improves after bariatric surgery. The aim of this study was to determine whether peripheral blood mononuclear cell albumin gene expression was related to NAFLD and whether albumin (ALB) and alpha fetoprotein (AFP) expression could be detected in whole blood and visceral adipose tissue.

Methods

Using a retrospective case control study design, RNA isolated from peripheral blood mononuclear cells from patients prior to undergoing bariatric surgery was used for pooled microarray analysis. Quantitative polymerase chain reaction (QPCR) was used to analyze whole blood and visceral adipose tissue. Liver histology was obtained via intra-operative biopsy and clinical data extracted from the electronic health record.

Results

The albumin (ALB) gene was the second most up-regulated found in microarray analysis of peripheral blood mononuclear cell RNA from patients with hepatic lobular inflammation versus normal liver histology. Transcript levels of ALB were significantly different across those with normal (n = 50), steatosis (n = 50), lobular inflammation (n = 50), and peri-sinusoidal fibrosis (n = 50) liver histologies, with lobular inflammation 3.9 times higher than those with normal histology (p < 0.017). Albumin expression levels decreased in 11/13 patients in paired samples obtained prior to and at 1 year after Roux-en-Y gastric bypass surgery. ALB expression could be detected in 23 visceral adipose tissue samples obtained intra-operatively and in 18/19 available paired whole blood samples. No significant correlation was found between ALB expression in visceral adipose tissue and whole blood RNA samples. Alpha fetoprotein expression as a marker of early hepatocytic differentiation was detected in 17/17 available VAT RNA samples, but in only 2/17 whole blood RNA samples.

Conclusion

Albumin RNA expression from blood cells may serve as a biomarker of NAFLD. Albumin and alpha fetoprotein appear to be ubiquitously expressed in visceral adipose tissue in patients with extreme obesity.

Introduction

Metabolic and bariatric surgery, e.g., Roux-enY gastric bypass (RYGB), is associated with an amelioration of a number of obesity-related co-morbidities, including non-alcoholic fatty liver disease (NAFLD) [1], which has become a leading cause of cryptogenic cirrhosis [2] and indication for liver transplantation [3, 4]. The physiological mechanisms involved in the beneficial effects of bariatric surgery on NAFLD are not yet well delineated, although a number of molecular mechanisms have been proposed [5, 6]. Because most patients achieve significant metabolic improvements after surgery, such molecular changes appear to mediate some degree of cellular repair or regeneration in damaged organs. A significant amelioration in NAFLD liver histology occurs after metabolic and bariatric surgery, with reported improvements in steatosis, inflammation (steatohepatitis), and fibrosis [7], consistent with significant enhancements in hepatic function [8]. Activation of resident hepatic stem and/or progenitor cells has also been implicated as part of the response of the liver in NAFLD [9].

In the course of characterizing the transcriptomic profiles of circulating peripheral blood mononuclear cells (PBMCs), the expression of the albumin (ALB) gene was unexpectedly identified as one of the most differentially expressed transcripts in patients with histologically documented steatohepatitis. Because cells with hepatocytic lineage differentiation have been detected using PCR circulating in the blood of patients with various types of cirrhosis [10] and similar types of cells can be mobilized with exogenous growth factors and detected by flow cytometry in patients with NAFLD [11], we expanded our analysis to include visceral adipose tissue (VAT) and alpha fetoprotein (AFP) as a marker of early hepatocytic differentiation [12, 13]. We hypothesized that albumin expression was originating from cells with hepatic differentiation that were co-migrating in the PBMC fraction and that they emanated from VAT. Our objectives were to determine whether PBMC ALB expression could serve as a biomarker for NASH and whether ALB and AFP could be detected in VAT. Expression of these genes in unexpected tissues and cell types may serve as biomarkers of NAFLD and potentially provides evidence in support of hepatic repair.

Materials and methods

Patients and samples

Blood samples along with associated clinical data were obtained from patients enrolled in the Bariatric Surgery Programs at the Geisinger Clinic Center for Nutrition and Weight Management [14] and the Metabolic and Bariatric Surgery Program of the Temple University Health System. All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The Institutional Review Boards of Geisinger Clinic and Temple University approved the research.

Patient samples were grouped by the major classes of liver histology in NAFLD, i.e., normal, steatosis, lobular inflammation, and fibrosis as defined by the NASH Clinical Research Network histological feature scoring system [15], that were selected from over 2000 patients consecutively enrolled from 2005 to 2013 who underwent bariatric surgery and intra-operative liver biopsy at Geisinger Clinic, a large integrated health system in north central Pennsylvania. Clinical data were obtained from the electronic health record as previously described [14] and included age, gender, race, BMI, alcohol use, diabetes, use of any type of 2 diabetes (T2D) medication, insulin use, statin use, hemoglobin A1c, AST, ALT, total cholesterol, LDL, HDL, triglycerides, and NASH activity score (NAS) calculated as the unweighted sum of the scores for steatosis (0–3), lobular inflammation (0–3), and ballooning (0–2). Cases were selected to maximize sample size for each histological group yet minimize differences in clinical parameters. Blood was collected into BD Vacutainer® CPT™ Cell Preparation Tubes (Becton, Dickinson and Company, Franklin Lakes, New Jersey) from which peripheral blood mononuclear cells (PBMCs) were processed per the manufacturer’s protocol [16], as well as into lavender EDTA anti-coagulated tubes for whole blood analysis, in the pre-operative period at least 1 month prior to surgery. Liver biopsies were obtained intraoperatively as previously described [17]. Visceral adipose tissue was taken from the gastrocolic omentum, along the greater curvature of stomach and preserved immediately and directly in RNAlater (Thermo-Fisher, Waltham, MA) and stored at − 80 °C until analysis. Gene expression in tissues taken at the beginning and end of bariatric surgery shows differences [18]; therefore, the timing of VAT biopsies was standardized [19].

RNA isolation

Total RNA was isolated using the RNeasy total RNA isolation kit (Qiagen, Valencia, CA) according to manufactor’s protocol and quantitated using a Nanodrop ND-1000 spectrophotometer (Thermo Scientific, Wilmington, DE, USA).

Microarray analysis

Total RNA isolated from PBMCs from 43 individuals with normal liver histology and 47 with histological evidence of lobular inflammation was pooled for analysis on the Human GeneChip Exon 1.0ST Array from Affymetrix (Santa Clara, CA, USA), prepared, and analyzed according to the manufacturer’s protocol, and the Affymetrix Transcriptome Analysis Console (TAC) 2.0 was used to compute the expression levels.

Quantitative real-time PCR

QPCR assays were performed using TaqMan® RNA-to-Ct™ 1-Step Kit (Life Technologies, Carlsbad, CA, USA) according to the manufacturer’s protocol. Pre-designed TaqMan assays for human albumin (targets, Hs00609411_m1 and Hs00910225_m1), GAPDH (endogenous control, Hs02786624_g1 and Hs99999905_m1), and AFP (target, Hs01040598_m1) were purchased from Life Technologies (Life Technologies, Carlsbad, CA, USA). Assay results were analyzed using Life Technologies StepOne Software Version 2.3. The expression level was calculated as the value of 40–delta C_t (C_t target–C_t endogenous control) and the fold change relative to HepG2 RNA was calculated using the formula: fold = 2–ΔΔCt (ΔΔCt = ΔCt test sample–ΔCt control sample). All results were then normalized to the lowest detectable RNA sample.

Statistical analysis

All statistical analyses were performed using GraphPad Prism 8 (GraphPad Software; La Jolla, CA). For simple comparisons between two groups, a two-sample Student’s t test was used. Repeated measures analysis of variance (ANOVA) with a Bonferroni post hoc test was used for comparison across histological groups. A p ≤ 0.05 was considered statistically significant.

The sample size for the discovery microarray analysis was based on available samples and a goal to keep the contribution to expression of any one individual to the final result to less than 2.5%, i.e., at least 40 samples per group. The sample sizes for the replication study were based on availability of samples and resources for conducting analyses. Bias was accommodated by matching for clinical characteristics and through adjusting for baseline differences in sex, race, type 2 diabetes medications, hemoglobin A1c, HDL, triglycerides, ALT, and AST for the ANOVA and pair-wise comparisons of QPCR results.

Results

Our initial approach used Affymetrix GeneChip® Human Exon 1.0 ST Arrays on human RNA isolated from PBMCs from patients with extreme obesity collected in the pre-operative period at least 2 weeks before the initiation of a liquid diet administered prior to bariatric surgery. Liver biopsies were obtained intra-operatively for histological analysis. We used a pooled approach to decrease costs as a screen to nominate individual genes for further study. PBMC RNA samples with associated histological data from 43 individuals with normal liver histology and from 47 with lobular inflammation, i.e., steatohepatitis, were pooled for analysis so that any individual sample would not contribute more than 2.5% to the total expression result. No differences were found between groups for age, gender, race, BMI, alcohol use, diabetes, any type 2 diabetes (T2D) medication, insulin use, statin use, hemoglobin A1c, AST, ALT, total cholesterol, LDL, HDL, or triglycerides (Table 1). Each pool was analyzed on 3 GeneChips. Surprisingly, ALB was identified as the second most up-regulated gene (greater than 20-fold) in PBMC RNA from those with lobular inflammation relative to the PBMCs of patients with normal liver histology (Table 2).

Table 1.

Patient demographics and clinical characteristics

		Microarray			QPCR					Post-Op
		Normal	Inflammation	p value	Normal	Steatosis	Inflammation	Fibrosis	p value
Sample size (n)		43	47		97	151	56	106		13
Age, years	Mean (SD) [range]	48.4 (10.4) [27, 69]	46.6 (10.7) [24, 68]	0.434	48.0 (12.5) [22, 72]	46.8 (10.4) [19, 69]	47.0 (12.1) [23, 71]	47.9 (9.6) [25, 72]	0.828	43.2 (13.6) [27, 66]
Gender	Female, % (n)	84% (n = 36)	85% (n = 40)	0.856	89% (n = 86)	81% (n = 122)	88% (n = 49)	71% (n = 75)	0.0060	85% (n = 11)
	Male, % (n)	16% (n = 7)	15% (n = 7)		11% (n = 11)	19% (n = 29)	13% (n = 7)	29% (n = 31)		15% (n = 2)
Race	White, % (n)	100% (n = 43)	100% (n = 47)	NA	95% (n = 92)	100% (n = 151)	100% (n = 56)	97% (n = 103)	0.010	100% (n = 13)
	Black, % (n)	0% (n = 0)	0% (n = 0)		5% (n = 5)	0% (n = 0)	0% (n = 0)	3% (n = 0)		0% (n = 0)
BMI, kg/m2	Mean (SD) [range]	48.5 (8.7) [35.0, 70.3]	49.4 (8.1) [35.1, 74.3]	0.589	48.3 (8.5) [35.6, 82.3]	49.5 (9.6) [35.0, 97.5]	49.4 (9.0) [35.1, 74.3]	48.6 (7.8) [35.1, 71.5]	0.730	50.5 (7.8) [36.1, 60.2]
Alcohol use	Yes, % (n)	44% (n = 17)	43% (n = 17)	0.922	29% (n = 24)	39% (n = 55)	40% (n = 21)	40% (n = 40)	0.398	58% (n = 5)
	No, % (n)	56% (n = 22)	57% (n = 23)		71% (n = 58)	61% (n = 86)	60% (n = 31)	60% (n = 60)		42% (n = 7)
	Unknown	N = 4	N = 7		N = 15	N = 10	N = 4	N = 6		N = 1
Diabetes	Yes, % (n)	30% (n = 13)	38% (n = 18)	0.421	31% (n = 30)	37% (n = 56)	34% (n = 19)	46% (n = 49)	0.136	38% (n = 5)
Any T2D medication	Yes, % (n)	47% (n = 20)	51% (n = 24)	0.666	34% (n = 33)	52% (n = 78)	50% (n = 28)	61% (n = 65)	0.0014	54% (n = 7)
Insulin use	Yes, % (n)	9% (n = 4)	6% (n = 3)	0.705	13% (n = 13)	19% (n = 29)	7% (n = 4)	14% (n = 15)	0.175	8% (n = 1)
Statin use	Yes, % (n)	40% (n = 17)	34% (n = 16)	0.589	34% (n = 33)	40% (n = 60)	38% (n = 21)	42% (n = 44)	0.718	15% (n = 2)
Hemoglobin A1c	Mean (SD)	6.2 (1.1)	6.3 (1.3)	0.713	5.9 (0.9)	6.3 (1.2)	6.4 (1.3)	6.8 (1.6)	0.0001	6.4 (1.9)
AST	Mean (SD)	24.6 (9.1)	26.5 (9.4)	0.353	25.6 (13.2)	24.8 (9.4)	25.4 (9.4)	34.2 (17.7)	<0.0001	23.1 (6.0)
ALT	Mean (SD)	28.0 (13.2)	31.8 (14.8)	0.194	27 (19.3)	28.7 (14.7)	30.7 (12.6)	41.2 (23.7)	<0.0001	28.7 (13.0)
Total cholesterol	Mean (SD)	183.9 (35.0)	188.4 (35.7)	0.545	179.4 (43.1)	182.5 (35)	186.1 (31.9)	188.1 (41.4)	0.363	184.1 (34.5)
LDL	Mean (SD)	107.6 (31.0)	106.6 (34.1)	0.884	101.5 (38.5)	107.6 (31.3)	103.7 (31.9)	104 (32.7)	0.733	106.9 (24.2)
HDL	Mean (SD)	45.9 (8.9)	49.3 (11.8)	0.131	49.7 (10.7)	46.1 (11.3)	47.1 (11.6)	43 (11.1)	0.0006	48.7 (24.2)
Triglycerides	Mean (SD)	151.9 (59.3)	162.4 (73.7)	0.459	140.8 (75.2)	146 (67.2)	180.7 (105.5)	219.6 (143)	<0.0001	175.6 (169.6)
NAS score	Mean (SD)	1.4 (0.7)	3.5 (1.2)	<0.0001	0.0 (0)	1.3 (0.6)	3.6 (1.2)	3.8 (1.6)	<0.0001	2.1 (1.9)

Table 2.

Affymetrix array results for the ten most up-regulated genes in PBMCs from patients with lobular inflammation versus normal histology

Fold change	Gene symbol	Gene name
21.38	DDR2	Discoidin domain receptor tyrosine kinase 2
20.56	ALB	Albumin
19.46	RAD50	RAD50 homolog
18.79	LRCH4	Leucine-rich repeats and calponin homology (CH) domain containing 4, Sin3A-associated protein, 25 kDa
16.64	POTEA	POTE ankyrin domain family, member A
14.59	FSTL4	Follistatin-like 4
13.84	SP3	Sp3 transcription factor,
13.39	RGPD5	RANBP2-like and GRIP domain containing 5,
13.1	MAPK11	Mitogen-activated protein kinase 11
12.58	LARP1B	La ribonucleoprotein domain family, member 1B, FOS-like antigen 1

Quantitative real-time PCR (QPCR) was then performed to measure the mRNA transcript levels of ALB relative to GAPDH in separate individual PBMC RNA (Fig. 1) samples from 410 individuals with normal (n = 97), steatosis (n = 151), lobular inflammation (n = 56), and peri-sinusoidal fibrosis (n = 106) liver histologies (Table 1). Differences were found across groups for gender, race, any type 2 diabetes (T2D) medication, hemoglobin A1c, AST, ALT, HDL, and triglycerides. For the 106 in the fibrosis group, the breakdown by fibrosis stage was 1a = 48% (n = 51), 1b = 8% (n = 9), 1c = 16% (n = 17), 2 = 18% (n = 19), 3 = 6% (n = 6), and 4 = 4% (n = 4). Adjusting for baseline differences in sex, race, any T2D medications, hemoglobin A1c, AST, ALT, HDL, and triglycerides did not impact the results, with the overall ANOVA p = 0.0005 and pairwise differences of steatosis versus normal p = 0.167, inflammation versus normal p = 0.0017, and fibrosis versus normal p = 0.0003. The PBMC ALB transcript levels in patients with lobular inflammation on liver biopsy were 3.22 times higher than those with normal histology (p = 0.004) and 3.2 times higher in those with peri-sinusoidal fibrosis on biopsy (p < 0.0001). For a subgroup of 13 individuals, PBMC samples were collected both preoperatively and at 1 year after RYGB (Table 1). ALB expression decreased in all but two patients (Fig. 2), consistent with an improvement in NAFLD.

Fig. 1.

QPCR of albumin (ALB) transcripts in PBMC mRNA relative to GAPDH from 410 individuals with normal (n = 97), steatosis (n = 151), lobular inflammation (n = 56), and peri-sinusoidal fibrosis (n = 106) liver histologies with ANOVA (*p = 0.0005) and pairwise comparisons (**p = 0.0017, ***p = 0.0003) indicated

Fig. 2.

Levels of ALB expressing cells decrease 1 year after bariatric surgery

With evidence that ALB gene expression could be detected in the PBMC fraction of blood, we sought to determine whether expression could be detected in a separate cohort using an independent assay, and in WB, a less burdensome sample to process, and in VAT, an abundant source of mesenchymal stem cells whose characteristics appear to be favorable for developing cellular-based therapies for liver disease [20]. In a group of 22 patients (17 females) with a mean age of 42, of whom 36% were either African American or Hispanic American, on whom paired WB and VAT samples were available, ALB expression could be detected in 23/23 samples (Table 3), with one much higher than the others and most within several-fold of each other. In the paired whole blood (WB) samples, ALB expression could be detected in 18/19 individuals. No significant correlation was found between ALB expression in VAT and WB RNA samples. We then measured AFP expression in VAT and WB as a marker of early hepatocytic differentiation. AFP was detected in 17/17 VAT RNA samples but only 2/17 paired WB samples (Table 3).

Table 3.

Relative ALB/GAPDH and AFP/GAPDH expression results in VAT and WB

ALB			AFP
Sample	VAT	WB	VAT	WB
1	1.14	12.59	3.40	-
2	1.00	8.75	1.68	-
3	1.67	5.33	1.70	-
4	2.05	18.28	2.08	-
5	2.29	11.40	4.51	-
6	1.63	6.45	2.26	-
7	1.94		6.40
8	1.08	3.11	1.59	1.38
9	2.05	6.08	1.00	-
10	3.08	8.24	2.41	-
11	2.32	12.20	4.26	-
12	2.09	-	10.76	-
13	4.85	4.41		-
14	1.28	4.87	1.27	-
15	1.95		9.89
16	2.31	5.66		-
17	4.13	19.55		-
18	86.94	5.42	3.76	-
19	4.40	2.25	3.21	-
21	2.66	1.00		1.00
22	1.47	4.10	1.91	-
23	6.43

Discussion

Metabolic and bariatric surgery can improve a number of metabolic conditions including NAFLD [21–37]. The underlying mechanisms by which the surgery mediates these favorable metabolic effects are complex but appear to include the mobilization of progenitor/stem cells that contribute to the mechanism of improvement of end-organ function. We found that albumin expression in PBMCs was related to NASH and that ALB and AFP expression could be detected in VAT. These findings are consistent with the detection of cells with hepatocytic differentiation circulating in the blood of patients with various types of cirrhosis and with VAT serving as a reservoir for progenitor-like cells with hepatocytic differentiation.

Our finding of circulating cells that express hepatocyte markers is consistent with the detection of cells with hepatocytic differentiation circulating in the blood of patients with various types of cirrhosis [10]. Similar types of cells can be mobilized with exogenous growth factors and detected by flow cytometry in patients with NAFLD [11]. These data are commensurate with the hypothesis that the inflammatory process of NAFLD/NASH generates signals for the mobilization of progenitor/stem cells for hepatic repair [38, 39]. Activation of resident hepatic stem and/or progenitor cells has also been implicated as part of the response of the liver in NAFLD [9], which some have regarded as a hepatic wound healing response to lipotoxicity in which progenitor/stem cells are recruited in order for dead hepatocytes to be replaced [40].

Histologic improvement of NASH and fibrosis in patients with extreme obesity after bariatric surgery has been documented in a number of studies. A recent systematic review and meta-analysis [41] found that steatosis and steatohepatitis improved after bariatric surgery in a majority of patients, with improvements in liver fibrosis in 30% of patients, though some patients can develop new NAFLD or experience worsening [42]. Bariatric surgery is associated with several major metabolic effects that may influence NAFLD including substantial weight loss, improvements in glucose metabolism, and a reduction in adipose-related inflammatory activity [7]. A molecular mechanism implicated in the metabolically ameliorative effects of bariatric surgery includes bile acids [6]. Bile acids have been shown to induce the differentiation of mesenchymal stem cells to the hepatocyte lineage [43] and to recruit vasculogenic progenitor cells to circulate [44]. However, bariatric surgery causes complex effects, including inducing major changes in cytokines and other signaling molecules [45]; thus, a variety of molecular mechanisms may be involved in eliciting cellular repair mechanisms which may have a significant impact on NAFLD. For example, weight regain after bariatric surgery has been associated with less improvement in NAFLD [46], though this is likely also associated with changes in glucose homeostasis, inflammation, and bile acid metabolism. How these changes may impact cellular repair mechanisms is not known. More studies on the mechanism by which bariatric surgery improves NAFLD are needed.

The tissue of origin of putative hepatic progenitor/stem cells includes adipose tissue and bone marrow, since both originate from the embryonic mesenchyme, share similar cell populations, and have both been used for harvesting mesenchymal stem cells [47]. Bone marrow is the most studied source of adult progenitor/stem cells, although human white adipose tissue has been reported to contain a significantly greater number of mesenchymal stem cells than bone marrow per unit weight [48], comprising as much as 1–3% of cells in the tissue [49, 50]. Adipose tissue is thus an abundant reservoir of mesenchymal stem cells [51] and has been used as a source of exogenously administered mesenchymal stem cells for therapeutic applications [52], but whether they can be mobilized endogenously after bariatric surgery is not known. Mesenchymal-type adipose-derived stromal/stem cells have been the subject of much recent investigation for a variety of translational applications in tissue repair and regenerative medicine, including for cardiovascular disease and hepatic cirrhosis, primarily due to their high abundance in adipose tissue, an accessible and abundant source, and for their multi-lineage pluripotency [53] and can yield a high expression of hepatic differentiated tissue via various in vitro methods [54]. Few studies have investigated stem cells from VAT likely because of the barriers to ethically obtaining the tissue from human volunteers. We used the opportunity of bariatric surgery to safely obtain samples for analysis.

We used expression of albumin (ALB) and alpha-fetoprotein (AFP) as markers for hepatic differentiation [12, 13], which are highly liver specific, and have previously been used as markers for liver committed progenitor cells mobilized in the peripheral blood of humans and mice [38]. However, we did not detect differential expression of AFP in PBMCs in the discovery microarray analysis. This may not be surprising given the results of the individual sample analysis which detected AFP expression in only 2/19 WB samples, but in all 19 VAT samples. Detection of expression in VAT but not WB may reflect differentiation of the hepatocyte-like cells. AFP is a marker for a relatively immature hepatic phenotype, whereas ALB expression reflects differentiated hepatocyte functioning [55]. Release of stem cells from adipose tissue into the circulation may not occur until a certain degree of differentiation is achieved. Isolation of the expressing cells and single cell analysis using techniques such as RNA sequencing will likely be required in order to characterize the expression.

Another possibility is that the cells expressing ALB and AFP were from the liver. Ballooning degeneration and hepatocyte cell death is a prominent part of NAFLD [15]. It is not known whether such hepatocyte cell fragments circulate, though if they do, it is possible that they are a component of the PBMC fraction. Their levels would also likely correlate with disease severity, thus would be increased with steatohepatitis and decrease following surgery. The detection of ALB and AFP in VAT could be from contaminating blood cells though we found no correlation between VAT and WB. Hepatocyte-derived extracellular vesicles have also been implicated in NAFLD [56], although it is not clear whether circulating extracellular vesicles would be collected in the PBMC fraction. It is also possible that ALB and/AFP are normally or abnormally expressed by a cell component of WB or VAT, rather than a recruited progenitor/stem cell. Larger and more detailed studies will be needed to determine the origin and potential significance of blood ALB expression in NAFLD.

Conclusion

ALB RNA expression from blood cells may serve as a biomarker of NAFLD that may be reflective of cellular repair following bariatric surgery. ALB and AFP also appear to be ubiquitously expressed in VAT in patients with extreme obesity. The precise role and cellular origin for these hepatocyte-specific markers has yet to be defined.