A wide scan of plasma proteins demonstrates thioredoxin reductase 1 as a potential new diagnostic biomarker for hepatocellular carcinoma

Figures & data

Figure 1. Flowchart of included patients. The outpatient group included 284 patients, from which the protein discovery cohort (n = 172) was derived from. The remaining patients in the outpatient group were combined with patients from the liver resection group to form the protein verification cohort (n = 160)*. Lastly, the protein verification cohort was divided into a training set and a validation set. *19 patients (9 HCC and 10 cirrhosis) from the protein discovery cohort were also included in the protein verification cohort.

Table 1. List of the 20 proteins with the highest fold-change in plasma levels in patients with HCC compared to those with cirrhosis. The proteins are sorted by decreasing values of fold-change in the SBA and PEA platforms, respectively.

Protein		Protein
SBA	Fold-change	PEA	Fold-change
GNA12†	1.65	KIM1†	2.32
HES4†	1.43	FGF21†	2.01
NCOA1†	1.42	HAOX1†	1.75
ISX†	1.41	CCL20†	1.70
BAIAP2L2	1.41	MMP12†	1.69
TXNRD1†	1.40	CES1†	1.62
AFP†	1.39	IL8†	1.59
ITPKA	1.38	ACE2	1.57
TPX2†	1.38	ADGRG1	1.56
WWTR1	1.37	HGF†	1.52
HSPG2	1.36	IL6†	1.50
HSPD1†	1.36	PRSS2	1.47
PPP2R2C	1.36	MIC-A/B	1.39
C6orf223	1.36	BNP	1.36
BRD4	1.35	PTN	1.36
FAM83D	1.35	CCL18	1.34
S100A9†	1.35	SLAMF7	1.33
PLEKHN1	1.35	FCGR2A	1.31
FBXL18	1.35	TRAIL-R2	1.31
STAT1	1.34	TNFSF14	1.30

† Denotes proteins selected for further protein verification analysis.

Table 2. Clinical and laboratory data of patients in the protein verification cohort (n = 160).

	HCC (n = 82)		Cirrhosis (n = 78)		p-value
	Median/number (%)	IQR	Median/number (%)	IQR
Age (years)	71	65.7 − 74.6	62	55.1 − 69.3	<0.001
Sex (male)	67 (82%)		49 (63%)		0.01
Cirrhosis	51 (62%)		78 (100%)		<0.001
Child Pugh					0.06
- A (or no cirrhosis)	59 (72%)		42 (54%)
- B	18 (22%)		28 (36%)
- C	5 (6%)		8 (10%)
MELD	9	7 – 11	9.5	7 – 13	0.19
Etiology					<0.001
- Viral hepatitis	18 (22%)		13 (17%)
- Alcohol	25 (30%)		35 (45%)
- NAFLD	16 (20%)		14 (18%)
- Autoimmune	0 (0%)		6 (8%)
- Other	8 (10%)		10 (13%)
Blood platelets (10^9/L)	198	141 – 259	127.5	92 − 187.5	<0.001
Plasma bilirubin (µmol/L)	10	6 – 16	17	9.25– 27	<0.001
Plasma albumin (g/L)	34	31 – 37	33	29 – 38	0.27
Plasma PT-INR	1.1	1.0 − 1.2	1.2	1.1 − 1.4	0.02
Ascites					0.02
- None	64 (78%)		45 (58%)
- Mild/moderate	12 (15%)		23 (29%)
- Severe	6 (7%)		10 (13%)
Sum of tumor diameters (cm)	5.5	3.0 − 9.5	–	–
Number of tumors	1	1 – 3	–	–
BCLC
- stage 0	8 (10%)		–
- stage A	38 (46%)
- stage B	20 (24%)
- stage C	13 (16%)
- stage D	3 (4%)

MELD: model of end stage liver disease; NAFLD: non-alcoholic fatty liver disease; PT-INR: prothrombin time international normalized ratio.

Table 3. Plasma levels of the analyzed protein biomarkers in patients with HCC and cirrhosis respectively, in the protein verification cohort (n = 160).

	HCC (n = 82)		Cirrhosis (n = 78)		Difference
Protein	Median	IQR	Median	IQR	p-value	AUC
Proteins with increased plasma levels in HCC compared to cirrhosis patients
DCP (ng/ml)	251	(90–2331)	38	(20–105)	<0.001	0.786
AFP (ng/ml)	7.7	(3.7–163)	4.1	(2.7−5.5)	<0.001	0.729
FGF21 (pg/ml)	516	(203–1077)	198	(88–334)	<0.001	0.717
TXNRD1 (ng/ml)	4.9	(3.0−6.4)	2.9	(2.1−4.4)	<0.001	0.692
AFPL3 (%)	0.0	(0.0−5.6)	0.0	(0.0−0.0)	<0.001	0.663
KIM1 (pg/ml)	31	(5.0–98)	18	(5.0–53)	0.008	0.648
Fuc-AGP (µg/ml)	2.6	(1.4−5.5)	1.6	(1.3−2.8)	0.006	0.630
S100A9 (pg/ml)	270	(180–426)	199	(156–306)	0.028	0.601
CES1 (ng/ml)	70	(40–127)	49	(35–87)	0.047	0.591
Proteins with increased plasma levels in cirrhosis compared to HCC patients
GP73 (ng/ml)	35	(26–52)	47	(33–68)	0.003	0.636
FGF19 (ng/ml)	0.31	(0.25−0.47)	0.44	(0.27−0.61)	0.007	0.626
MMP12 (pg/ml)	15	(11–16)	16	(11–20)	0.017	0.609
HES4 (ng/ml)	0.33	(0.22−0.58)	0.43	(0.30−0.61)	0.035	0.597
Proteins with no significant differences
HSP60 (pg/ml)	612	(374–1122)	423	(337–733)	0.366	0.610
OPN (pg/ml)	1722	(1012−2401)	1353	(938–1953)	0.067	0.585
IL8 (pg/ml)	13	(7.8–26)	11	(6.6–23)	0.216	0.557
HAOX1 (ng/ml)	8.9	(3.8–32)	10	(4.1−108)	0.267	0.551
HGF (pg/ml)	280	(191–436)	299	(193−522)	0.457	0.534
IL6 (pg/ml)	4.8	(2.6–10)	4.3	(2.1−9.0)	0.540	0.529
CCL20 (pg/ml)	112	(55–239)	105	(64–169)	0.756	0.514
NCOA1 (ng/ml)^a	0.47	(0.17−0.69)	0.45	(0.20−0.64)	0.656	0.535
GNA12 (ng/ml)^b	0.59	(0.29−1.0)	0.65	(0.31−1.2)	0.600	0.532
ISX (ng/ml)^b	20	(15–26)	15	(14–22)	0.931	0.533
TPX2 (ng/ml)^c	4.0	(3.7−4.1)	3.7	(3.4−3.8)	0.548	0.640

AUC: area under the curve; IQR: interquartile range.

^a57 HCC and 19 cirrhosis samples analyzed.

^b40 HCC and 53 cirrhosis samples analyzed.

^c5 HCC and 5 cirrhosis samples analyzed.

Table 4. The AUC values of the two-, three- and four-protein combination with the highest AUC in the cut-off model employed on the training set (n = 120), with values of the validation set (n = 40) shown in parenthesis. The panel to the right shows the AUCs when sex and age are included in the model.

Parameters	AUC for parameters	Specificity at 90% sensitivity	AUC for parameters + sex & age	Specificity at 90% sensitivity
AFP, DCP	0.799 (0.732)	48% (35%)	0.835 (0.825)	51% (55%)
AFP, DCP, AFPL3%	0.822 (0.752)	48% (35%)	0.861 (0.830)	54% (55%)
AFP, DCP, TXNRD1, S100A9	0.840 (0.776)	59% (33%)	0.859 (0.836)	55% (50%)

Figure 2. ROC curves for the combined score using the logistic regression method. The combination of DCP and AFP yielded an AUC of 0.844 (a). Addition of TXNRD1 increased the AUC to 0.878 (b), while addition of sex and age increased the AUC to 0.914 (c). The highest AUC (0.920) was achieved when all parameters were combined (d).

Table 5. The AUC values of different protein combinations in the logistic regression model employed on the training set (n = 120), with values of the validation set (n = 40) shown in parenthesis. The panel to the right shows the AUCs when sex and age are included in the model. AFP and DCP are included in every multiprotein combination being the proteins with the highest AUC, and with AFPL3%, FGF21 and TXNRD1 added in various combinations.

Parameters	AUC for parameters	Specificity at 90% sensitivity	AUC for parameters + sex & age	Specificity at 90% sensitivity
AFP, DCP	0.844 (0.770)	55% (25%)	0.914 (0.882)	74% (55%)
AFP, DCP, AFPL3%	0.844 (0.765)	55% (25%)	0.912 (0.880)	74% (60%)
AFP, DCP, FGF21	0.846 (0.763)	56% (32%)	0.915 (0.882)	74% (55%)
AFP, DCP, TXNRD1	0.878 (0.855)	66% (65%)	0.920 (0.895)	81% (68%)
AFP, DCP, AFPL3%, FGF21	0.846 (0.763)	56% (32%)	0.916 (0.887)	76% (55%)
AFP, DCP, TXNRD1, FGF21	0.876 (0.850)	67% (65%)	0.922 (0.902)	84% (70%)
AFP, DCP, AFPL3%, TXNRD1	0.877 (0.858)	66% (65%)	0.923 (0.900)	81% (75%)
AFP, DCP, AFPL3%, TXNRD1, FGF21	0.873 (0.865)	64% (70%)	0.925 (0.897)	87% (73%)

Figure 3. Plasma concentrations of proteins included in either the cut-off or log-reg models (AFP, DCP, TXNRD1, FGF21, S100A9 and AFPL3%) between patients with cirrhosis, early HCC (BCLC stage 0 or A) and advanced HCC (BCLC stage B to D). All proteins were significantly increased when comparing patients with cirrhosis to those with HCC BCLC stage B to D, but only AFPL3%, TXNRD1, FGF21 and DCP demonstrated a significant difference between patients with cirrhosis and early HCCs with BCLC stage 0 or A.

Data availability statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.