Acute-on-chronic liver failure (ACLF) is a syndrome of liver failure manifested by acute jaundice and coagulopathy on the basis of chronic liver disease.¹ Predicting the prognosis of ACLF plays a crucial role for the treatment of the disease which associate with a high risk of mortality. At present, there are some predictors to assess prognosis of this disease clinically, such as the Model for the CLIF Consortium ACLF (CLIF-C ACLF) score, the Model for End Stage Liver Disease (MELD) score, and artificial liver support system-prognosis model (APM).^2-4 Future study is still needed to evaluate the prognostic biomarkers of liver failure due to the limitations of the above indicators.

Previous studies have shown that antigenic changes in cells can be recognized by the immune system of patients, which may lead to the appearance of circulating autoantibodies.⁵ To detect autoantibodies in human serum, as an important indicator for the diagnosis of autoimmune diseases, have been widely used in clinical research and practice.⁶ Immunoproteomics is a new technology for screening and identifying disease-related antigens, with unique technical advantages.⁷ Much work has been done in our previous research on the identification of tumor-associated antigens as biomarkers in hepatocellular carcinoma with the technology of immunoproteomics.⁸ The highly specific autoantibody response of systemic autoimmune diseases usually predicts the biological phenotype of the disease, which indicating that autoantibodies have important clinical significance and diagnostic value. Studies have shown that autoantibodies are not only found in the serum of patients with immune diseases, but also in cancer, virus hepatitis, and liver failure.^9-11

Rapid development of hepatocellular necrosis in the progression of liver failure may lead to abnormal exposure of certain protein components or abnormal protein expression positions in hepatocytes. These proteins or antigenic components which are abnormally exposed during liver failure are called liver failure-associated antigens (LFAAs), which are similar with the tumor-associated antigens. The immune system of patients with liver failure can recognize these abnormally expressed proteins as foreign proteins to produce an immune response that induces the production of anti-LFAA autoantibody.^12,13 The expression of anti-LFAA autoantibody in liver failure patients with different prognosis will be different. Therefore, serum anti-LFAA autoantibodies, which can be easily detected clinically, are expected to predict the prognoses of patients with liver failure. An immunoproteomic approach was applied to screen out specific LFAAs related to different prognosis of ACLF with 60 samples in our previous study.

In this study, 82 serums of ACLF patients were used to investigate the frequency and titer of anti-LFAA autoantibody with the technology of enzyme-linked immunosorbent assay (ELISA). At the same time, we compared the different expressions of LFAAs in ACLF, chronic hepatitis B (CHB), liver cirrhosis (LC), and normal population to verify whether anti-LFAA autoantibodies can be used as biomarkers to evaluate the prognosis of ACLF.

1. Study design and patients

In this study, we included 60 CHB patients, 60 hepatitis B virus (HBV)-related cirrhosis patients, and 82 ACLF patients admitted to Beijing Youan Hospital from August 2019 to October 2021. Another cohort of 60 serums of ACLF patients was used to screen out the different anti-LFAA autoantibodies in ACLF patients in our previous screening study. In this validation cohort study, 82 ACLF patients were followed up for 3 months after the diagnosis of ACLF. Twenty-four normal human serum (NHS) samples were obtained from healthy people in the same period, excluded systemic diseases and liver diseases.

The Ethics Committee of Beijing Youan Hospital, Capital Medical University approved the study (approval number: [2019]013). Informed consent forms have been signed by all patients in this study.

2. Criteria

The entry criteria are based on the “Guidelines for acute-on-chronic liver failure” formulated by consensus recommendations of the Asian Pacific Association for the Study of the Liver in 2019.¹⁴ ACLF is an acute hepatic insult manifesting as jaundice (serum bilirubin ≥5 mg/dL [85 µmol/L]) and coagulopathy (international normalized ratio ≥1.5 or prothrombin activity <40%) complicated within 4 weeks by clinical ascites and/or encephalopathy in a patient with previously diagnosed or undiagnosed chronic liver disease/cirrhosis, and is associated with high 28-day mortality.

All patients should be excluded from the following criteria: (1) clearly diagnosed autoimmune diseases; (2) with other serious active physical and mental diseases, including uncontrolled primary lung, heart, vascular, kidney, metabolic and neurological diseases, digestive diseases, immunodeficiency diseases or combined malignant tumors, etc.; (3) patients during pregnancy or lactation.

3. Methods

1) Cell culture and extraction

Hepatocellular carcinoma cell line (HepG2) was donated by the Artificial Liver Laboratory of Beijing Youan Hospital. Refer to previous literature for specific culture methods and experimental reagents and instruments.¹¹

2) Indirect immunofluorescence assay

Indirect immunofluorescence assay was performed on Hep-2 cell matrix slides. The serum diluted to 1:40 with phosphate-buffered saline (PBS) pH 7.4, then incubated with the slides at ambient temperature for half an hour, and then washed thoroughly. After that, the slides were incubated with a goat anti-human IgG secondary antibody conjugated with fluorescein isothiocyanate at ambient temperature for 20 minutes and washed thoroughly with PBS. Then, added a drop of mounting agent containing 4,6-diamidino-2-phenylindole. Nikon ECLIPSE Ti (Tokyo, Japan) was used to examine the slides as mentioned before.¹¹

3) Two-dimensional gel electrophoresis analysis

To obtain an atlas of the proteins in HepG2 cells, total proteins from HepG2 cells were separated by two-dimensional gel electrophoresis and then transferred onto nitrocellulose membranes (Millipore, Burlington, MA, USA). HepG2 cells were lysed briefly with 1 mL lysis buffer (40 mM Tris, 1 mM ethylenediaminetetraacetic acid Na2, 2 M thiourea, 7 M urea, 4% CHAPS, 1% dithiothreitol, total volume 10 mL), and the protein supernatant was centrifuged and harvested at 12,000× rpm for 30 minutes at 4℃ after standing at 4℃ for 2 hours. Protein was purified through 2-D clean up kit (GE, Boston, MA, USA), the purified protein dissolved in hydration solution (Bio-Rad, Hercules, CA, USA) could be stored at –80℃ or directly used in isoelectric focusing after quantification. The protein concentration was determined by a 2D quantification kit (GE, Boston). For one-dimensional gel electrophoresis analysis, 130 µL protein solution containing 250 µg protein was mixed with a hydration solution containing a trace of bromophenol blue and then applied to pH 3–10, 7 cm isoelectric focusing strip (purchased from Bio-Rad). Isoelectric focusing was carried out at an electric current of 50 mA per gel, 50 V for 12 hours (hydration), 250 V for 30 minutes (desalination), 1,000 V for 1 hour (desalination) 4,000 V for 3 hours (boost voltage), 4,000 V 7 hours (focus). The strips after one-dimensional gel electrophoresis were stored at –80°C in time or used for the next dimensional gel electrophoresis analysis after equilibrating twice with an equilibration buffer containing 2% dithiothreitol. In the second dimensional electrophoresis, the protein in strips were electrophoresed on 12% sodium dodecyl sulfate–polyacrylamide gels and transferred onto nitrocellulose membrane for Western blot analysis. The spots points were saved by scanning.

As mentioned above, proteins were separated by two-dimensional gels and transferred onto nitrocellulose membrane. Next, the primary and secondary antibody were incubated in turn, and the immune response spots were detected by enhanced chemiluminescence kit. See the previous literature for details.¹¹

4) Gel scan analysis

On the gel stained with Coomassie brilliant blue, the Western blot reaction map was scanned with the Image Scanner, and the image was compared and analyzed with image master 7.0 to identify the differentially reactive protein spots between the alpha-fetoprotein (AFP)-negative hepatocellular carcinoma patient serum and the normal control. Parameters for selecting the point are "smooth" to 3, "min area" to 65 and "salience" to 250. By comparing the Western blot reaction map and the gel map, a protein spot that matches the differential reaction protein spot in the Western blot is found on the parallel gel.

5) In-gel digestion

The protein spots were excised from two-dimensional gels stained and decolorized by washing three times in 200 µL aliquots of 50 mM ammonium bicarbonate in 50% (v/v) acetonitrile for 15 minutes each time. The gel pieces were dried in a Speed Vac Vacuum, and then rehydrated at 4℃ for 15 minutes in 3 to 5 µL digestion solution containing 0.01 mg/mL modified sequence-grade trypsin and 25 mM ammonium bicarbonate. Then the same volume of trypsin-free digestion solution was added to keep the gel pieces moist during the digestion process. The digestion was stopped with 1% trifluoroacetic acid for 15 minutes after incubating overnight at 37°C. Peptides were extracted by 20 µL 0.1% trifluoroacetic acid for 0.5 hour and then by 20 µL 0.1% trifluoroacetic acid/50% acetonitrile for 0.5 hour.

6) Peptide mass fingerprinting by MALDI-TOF-MS

See the previous literature for details.¹¹

7) Enzyme-linked immunosorbent assay

ELISA technology was performed for the test of the frequency of LFAAs autoantibodies in 82 ACLF patients. The fumarate hydratase (FH) protein was coated on a 96-well plate at a dilution concentration of 0.5 µg/mL at 4°C overnight. Then, the supernatant was discarded, and 10% fetal calf serum was added to block the response at 37°C for 1 hour. The test serum of the primary antibody was diluted to 1:100. PBS was used as a blank control. A 100 µL volume of the above liquid was added to each well, then incubated at 37°C for 40 minutes. The plate was washed three times with phosphate buffered saline with tween-20. The secondary antibody horseradish enzyme labeled goat anti-human IgG was diluted to 1:10,000 with PBS. Incubate 100 µL of the secondary antibody for 30 minutes at 37℃ per well. Then wash the plate three times with phosphate buffered saline with tween-20. A 100 µL of chromogenic solution was added to react at 37℃ for 10 minutes per well. Then 0.05 mL of 2 M sulfuric acid was added to each well to stop the reaction. Read the optical density (OD) value with a microplate reader at 450 nm.

The mean OD value of the 24 NHS samples plus three standard deviations was designated as the cutoff value. Each microtiter plate was included with 10 NHS samples of serum and their average OD value was used to normalize all OD values to the standard mean of total normal samples. Repeat the test twice for each sample.

4. Statistical methods

SPSS 25.0 software (IBM Corp., Armonk, NY, USA) was used for data processing and statistical analysis. When the measurement data were approximately normal distribution, it was expressed as mean±standard deviation. An independent sample t-test was used for baseline comparison between the two groups. Multivariate analysis of variance with post hoc analysis was used in analyzing the comparisons between multiple groups. The chi-square test was used for comparison between count data groups. A p<0.05 means the difference is statistically significant.

1. The prognosis of the 82 ACLF patients

Outcomes of the 82 ACLF patients were followed up for 3 months after the diagnosis. Fifty-six patients included in the good prognosis group, which survived without liver transplantation at 3 months after diagnosis of ACLF. This part of patients accounted for 68.29% of the total. Twenty-six patients (31.71% of the total) included in the poor prognosis group, which died within 3 months of ACLF diagnosis or underwent liver transplantation.

2. Clinical characteristics of the 82 ACLF patients

The clinical characteristics of ACLF patients on baseline in the good prognosis and poor prognosis groups are shown in Table 1. Sixty cases of ACLF patients are HBV-related liver failure. The acute insult of HBV-related liver failure was due to HBV replication in 52 cases and infection, alcohol consumption, or drugs in eight cases. Twenty-two cases of alcoholic liver disease, drug-induced liver disease or other chronic liver disease were deteriorated by infection or variceal bleeding. We compared the laboratory findings of the patients between the two groups. Among them, AFP, MELD score, and prothrombin time of the poor prognosis group were remarkably higher than those of the good prognosis group (all p<0.05). There were no significant differences between the two groups in sex, age, transaminase (aspartate transaminase and alanine transaminase), total bilirubin, albumin, platelets, and creatinine.

Table 1 . Clinical Characteristics of the Acute-on-Chronic Liver Failure Patients with Different Prognoses.

Clinical characteristics	Good prognosis (n=56)	Poor prognosis (n=26)	χ2/z	p-value‡
Male sex	36 (64.3)	21 (80.8)	2.28	0.13
Age, yr	47.09±14.02	50.32±12.20	–1.22	0.22
Alanine transaminase , U/L	217.86 (28.75–307.25)	177.44 (38.50–215.50)	–0.55	0.58
Aspartate transaminase, U/L	209.50 (65.25–215.75)	138.32 (68.50–194.00)	–0.26	0.79
Albumin, g/L	30.98 (27.03–33.70)	31.39 (28.30–34.65)	–0.57	0.57
Creatinine, µmol/L	79.94 (47.25–80.80)	71.91 (42.50–69.00)	–1.20	0.23
Platelets, ×109/L	117.16 (59.00–173.25)	95.88 (69.50–113.50)	–0.91	0.36
Prothrombin time, sec*
≤20	20	3	3.86	<0.05
>20	35	22
Total bilirubin, µmol/L
≤171	15	2	2.86	0.09
>171	41	24
Alpha-fetoprotein, ng/mL
≤25	29	4	8.33	<0.05
>25	27	22
MELD score*,†
≤24	27	6	4.47	<0.05
>24	28	19
Etiology
HBV/alcohol/drug/others	41/7/4/4	19/4/2/1

Data are presented as number (%), mean±SD, or median (interquartile range)..

MELD, Model for End Stage Liver Disease; HBV, hepatitis B virus..

*Regarding the prothrombin time indicator, there is one missing data for each of the good and poor prognosis groups. Due to the lack of prothrombin time data, one data is missing in each of the two groups of MELD scores; ^†MELD score: R=3.8ln[bilirubin (mg/dL)]+11.2ln (international normalized ratio)+9.6ln[creatinine (mg/dL)]+6.4; ^‡p<0.05 was considered statistically significant..

3. Identification of immunoreactive proteins in prognostic evaluation of ACLF

The total number of ACLF patients is 142 in this study. Two different cohorts of serums were used for screening (n=60) and validation (n=82) studies with completely different research methods. In previous screening cohort study, 60 serums of ACLF patients were used to screen out the different anti-LFAA autoantibodies in ACLF patients with different prognoses. Indirect immunofluorescence assay and Western blot analysis were performed in the screening study. Serums from 19 patients with positive autoantibodies were screened out by indirect immunofluorescence assay method (Fig. 1), which is a preliminary qualitative method for identifying autoantibody-positive serum. Among the serum of these autoantibody-positive patients, five serums from patients with good prognosis and five serums from patients with poor prognosis were selected and mixed respectively for further screening of differential proteins. Then we applied two-dimensional electrophoresis to screen for differentially expressed proteins related to ACLF prognosis. The protein spots on the membrane were matched and compared with the equivalent protein spots on the original two-dimensional gel by three repeated tests. The results indicated that there were 17 immunoreactive protein spots differentially expressed in the serum of ACLF patients with poor prognosis or good prognosis (Fig. 2). After obtaining the peptide mass fingerprint, we used the Mascot search engine to search the SWISS-PROT and NCBI databases to identify the protein. Fifteen of 17 protein spots were identified by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry, as shown in Table 2. FH was one of the identified LFAAs (Fig. 3).

Table 2 . Summary of Identified Protein Spots by MALDI-TOF-MS.

Spot No.	Identified proteins	Accession no.	Theory pi/MW	No. of peptides	Score	Sequence coverage, %
0	Fumarate hydratase	NP_006316.1	54830	29	208	29
1	GTP-binding nuclear protein Ran isoform 1	NP_006316.1	24579	18	285	45
2	Triosephosphate isomerase isoform1	NP_000356.1	26938	29	835	83
3	Keratin 10	AAH34697.1	59020	28	295	29
4	Annexin A2	EAW77587.1	32600	27	171	55
5	Peroxiredoxin-6	NP_004896.1	25133	16	279	44
6	Heat shock protein 27	AAA62175.1	22427	6	128	17
7	Actin	NP_001605.1	42108	33	876	73
8	Hemoglobin subunit alpha	AQN67653.1	20190	6	123	29
9	Mutant beta-globin	AAL68978.1	16098	8	165	35
11	Poly(rC)-binding protein 1	NP_035995.1	37987	20	216	35
827	Heat shock 70	NP_004125.3	73920	34	486	37
925	Keratin1	NP_006112.3	66170	23	189	30
926	T-complex protein 1 subunit beta isoform 1	NP_006422.1	57794	28	292	41
1021	AKR1B1	CAG29347.1	36231	25	259	46

MALDI-TOF-MS, matrix-assisted laser desorption/ionization time-of-flight mass spectrometry..

Figure 1. Indirect immunofluorescence assay results of 60 acute-on-chronic liver failure (ACLF) patients (400×). Indirect immunofluorescence assay technology was used to screen the autoantibody-positive serum of ACLF patients with a good prognosis or poor prognosis. The 19 serum samples positive for autoantibodies were identified.
FITC, fluorescein isothiocyanate; DAPI, 4,6-diamidino-2-phenylindole.

Figure 2. Two-dimensional gel electrophoresis results of 17 specific proteins that were differentially expressed in good and poor prognosis acute-on-chronic liver failure patients. Spot number 0 represents fumarate hydratase.

Figure 3. Mascot score histogram of fumarate hydratase protein. Peptide mass fingerprinting was used to search the SWISS-PROT and NCBI database via the Mascot search engine.

4. Frequency and titer of autoantibody against FH in serum from patients with ACLF

In the validation cohort study, 82 serums of ACLF patients were used to verify the frequency of anti-LFAA autoantibody with the technology of ELISA. By the verification of ELISA, we found that the expression of FH autoantibody showed the most significant difference in ACLF patients with good or poor prognoses compared to other LFAAs autoantibodies. The titer of anti-FH/Fumarate hydratase protein autoantibodies in the serum of ACLF patients (1.26±0.80) was significantly higher than that of NHS (0.49±0.19, p<0.001) (Fig. 4). More importantly, the frequency (83.9% vs 61.5%, p=0.019) and titer (1.41±0.85 vs 0.94±0.56, p=0.017) (Table 3, Fig. 4A) of anti-FH/Fumarate hydratase protein antibodies in the serum of patients with good prognosis was much higher than that of patients with poor prognosis. The cutoff value of serum anti-FH/Fumarate hydratase protein to differentiate between ACLF patients with good or poor prognosis calculated by the Youden index was 1.28. The FH was compared with traditional MELD scores and APM model using the area under the receiver operating characteristic (AUROC) (Fig. 5A). Receiver operating characteristic curve showed that anti-FH/Fumarate hydratase protein had the largest AUROC, superior to MELD scores and APM model.

Table 3 . Frequency of Anti-FH/Fumarate Hydratase Protein Autoantibodies in Serum.

Group	No.	Frequency of anti-FH/Fumarate hydratase protein autoantibodies, %	p-value
Normal control	24	0	<0.001*
CHB	60	6.7	0.022†
LC	60	10.0	<0.001‡
ACLF	82	76.8	0.019§
Good prognosis	56	83.9	<0.001Ⅱ
Poor prognosis	26	61.5	0.001¶

FH, fumarate hydratase; CHB, chronic hepatitis B; LC, liver cirrhosis; ALCF, acute-on-chronic liver failure..

Multivariate analysis of variance with post hoc analysis was used to analyze the comparisons among multiple groups; p<0.05 was considered statistically significant..

*Normal control group vs ACLF group; ^†CHB group vs ACLF group; ^‡LC group vs ACLF group; ^§Good prognosis group vs poor prognosis group; ^ⅡGood prognosis group vs normal control group; ^¶Poor prognosis group vs normal control group..

Figure 4. Serum levels of anti-fumarate hydratase (FH)/Fumarate hydratase protein autoantibodies in hepatitis B virus-related liver disease. (A) The titer of anti-FH/Fumarate hydratase protein autoantibodies in serum of acute-on-chronic liver failure (ACLF) patients with different prognoses and healthy people. (B) The titer of anti-FH/Fumarate hydratase protein autoantibodies in the serum of normal humans and chronic hepatitis B (CHB), liver cirrhosis (LC), and ACLF patients. *p<0.05.

Figure 5. Predicting power of anti-FH/Fumarate hydratase protein autoantibodies on ACLF compared with traditional predicting scores. (A) Receiver operating characteristic (ROC) curves illustrating the ability of different prognostic models. MELD score=3.78×ln [TBIL (mg/dL)]+11.2×ln INR+9.57×ln [creatinine (mg/dL)]+6.43×(etiology: 0 if cholestatic or alcoholic, 1 otherwise). (B) ROC curves illustrating the ability of anti-FH/Fumarate hydratase protein antibody titers to differentiate between ACLF and non-ACLF. The area under the ROC (AUROC) was 0.741.
FH, fumarate hydratase; ACLF, acute-on-chronic liver failure; MELD, Model for End Stage Liver Disease; TBIL, total bilirubin; INR, international normalized ratio; APM, artificial liver support system-prognosis model; AFP, alpha-fetoprotein.

5. Frequency and titer of serum FH autoantibody from patients with CHB/LC/ACLF

We analyzed the frequency and titer of FH autoantibody in healthy human subjects, patients with CHB, patients with LC, and ACLF patients to compare the difference in FH expression. ACLF patients had the highest FH autoantibody frequency and titer (76.8%, 1.26±0.80, p<0.019), which was remarkably higher than that of patients with LC (10%, 0.69±0.42, p<0.001), CHB (6.7%, 0.89±0.69, p<0.022) and healthy people (0%, 0.49±0.19, p<0.001) (Table 3, Fig. 4B). By comparing the frequency of FH in different patients, we found that the expression of FH autoantibody was specifically increased in ACLF patients. The cutoff value of serum anti-FH/Fumarate hydratase protein to differentiate between ACLF and non-ACLF calculated by the Youden index was 0.745. The AUROC was 0.741 (p<0.001) (Fig. 5B).

6. Dynamic changes of FH expression in patients with ACLF

We collected peripheral blood samples of six ACLF patients with good prognosis at multiple time points to compare the different expression of FH autoantibody. The blood samples were obtained at admission (the time points of exacerbation) and the time points of recovery. It showed that ACLF patients with good prognosis maintained a high FH titer (1.5 to 2.5) during liver failure, but the FH titer decreased to a much lower level after the recovery of liver function, which was even close to that of healthy people around 0.5 (Fig. 6).

Figure 6. The dynamic changes of anti-fumarate hydratase (FH)/Fumarate hydratase protein autoantibody titers in acute-on-chronic liver failure (ACLF) patients with a good prognosis. The blood samples were obtained at admission (the time points of exacerbation) and the time points of recovery. The results showed that ACLF patients with a good prognosis maintained a high anti-FH titer (1.5–2) during liver failure, and the anti-FH titer decreased to a much lower level after liver function recovery.

Blood samples of two ACLF patients with poor prognosis were collected at admission (the time points of exacerbation) and other timepoints during hospitalization. We found that ACLF patients with poor prognosis had consistently poor liver function during hospitalization, and the FH titer remained high around 1.2, which was lower than the FH titer of ACLF patients with good prognosis (Fig. 7).

Figure 7. The dynamic changes of anti-fumarate hydratase (FH)/Fumarate hydratase protein autoantibody titers in acute-on-chronic liver failure (ACLF) patients with a poor prognosis. Blood samples were collected at admission (the time points of exacerbation) and other timepoints during hospitalization. We found that ACLF patients with a poor prognosis had consistently poor liver function during hospitalization, and the anti-FH titer remained high, at approximately 1.2.

The mortality rate of patients with ACLF which undergoing medical treatment is as high as 54.4% to 75.3%.^15,16 Accurately evaluating the patient's condition and prognosis is very important for the determination of treatment plans.¹⁷ Our study focused on exploring the effective predictive index for predicting the prognosis of liver failure. We evaluated the prognosis of ACLF from the perspective of autoantibodies produced by liver failure, which provided potential new clues to the prognostic research of ACLF.

In this study, 82 patients with ACLF were compared in laboratory examination and LFAAs expression according to different prognosis. Among the laboratory findings, AFP, MELD score and prothrombin time of ACLF patients with poor prognosis were notably higher than those of ACLF patients with good prognosis at baseline. Clinically, ACLF patients with higher AFP level are more likely to have a better prognosis.^4,18 The results of our study showed that AFP did not increase significantly in patients with good prognosis, possibly because of the baseline serums were collected at the beginning of the patient's onset, and AFP has not risen yet at that timepoint. When the regeneration of hepatocytes is more obvious in the late course of the disease, AFP will increase significantly. Application of cohort study to collect samples at multiple time points during the 3-month course of the ACLF is helpful to investigate the full picture of AFP changes.

Previous evidence has shown that infiltration of mononuclear macrophages and inflammatory factors storm are accompanied with the development of liver failure.^1,19 In our research, we screened out 15 specific proteins (LFAAs) related to different prognoses of the ACLF, such as FH, heat shock protein (HSP), actin and cytokeratin, which play a key role in inflammation and cell necrosis. Previous studies have shown that HSP, actin and cytokeratin are closely related to liver failure.^20,21 But little is known about whether FH participates in the pathogenesis of liver failure or not. Thus, we chose five anti-LFAAs protein autoantibodies included HSP27, HSP70, actin, cytokeratin, and FH to verify the seroprevalence of them in ACLF patients by ELISA technology. The expression of anti-FH/Fumarate hydratase protein autoantibodies showed the highest positive rate among them.

In our study, we found that ACLF patients have a much higher frequency and the titer of anti-FH/Fumarate hydratase protein autoantibodies than healthy people. It is well known that FH is a tricarboxylic acid cycle enzyme localized in the mitochondrial matrix. Recently, the hot area of FH is its metabolic activity in gene transcription linking to tumor cell growth. The activities of various enzymatic processes in cells closely related to the prevention and development of cancer can be regulated by the metabolites produced by FH.^22,23 We speculate that the significantly high expression of FH in ACLF patients is because of its affection on the cell metabolism and cellular signaling.²⁴

ACLF is an acute liver failure syndrome that appears on the basis of chronic hepatitis or cirrhosis. Therefore, we clarified the expression of autoantibodies of FH in patients with CHB and HBV-related cirrhosis. ACLF patients had much higher FH autoantibody frequency (76.8%) than that of patients with LC (10%), CHB (6.7%), and normal human (0%). The cutoff value of serum anti-FH/Fumarate hydratase protein to differentiate between ACLF and non-ACLF calculated by Youden index was 0.745. The AUROC was 0.741. The result indicates that the high expression of FH autoantibody is a specific biomarker of ACLF, which is clearly different from CHB and LC.

Meanwhile, we collected peripheral blood samples of six ACLF patients with good prognosis and two ACLF patients with poor prognosis at multiple time points to analyze the dynamic changes of FH expression during the following clinical course. We found that the titer of FH autoantibody was very high when the patient was critically ill, but FH expression decreased significantly when the patients recovered. It is in line with the result of our study that ACLF patients have much higher of frequency and the titer of anti-FH/Fumarate hydratase protein autoantibodies than healthy people.

More importantly, the frequency and the titer of anti-FH/Fumarate hydratase autoantibodies in the serum of ACLF patients with good prognosis were significantly higher than that of patients with poor prognosis. The cutoff value of serum anti-FH/Fumarate hydratase protein to differentiate between ACLF patients with good or poor prognosis calculated by Youden index was 1.28. Predicting power should be compared with traditional predicting scores. Receiver operating characteristic curve showed that anti-FH/Fumarate hydratase protein had the largest AUROC, superior to MELD scores and APM model. The result indicates that higher serum level of FH autoantibodies may predict a good outcome of ACLF. It is reported that decreases of FH can lead to adenosine triphosphate depletion by crippling tricarboxylic acid cycle and oxidative phosphorylation.²⁵ Therefore, we speculate that because of the involvement of higher frequency and titer of FH, ACLF patients with good prognosis have an advantage in hepatocyte regeneration compared with ACLF patients with poor prognosis.

In particular, the FH-specific T cell response was related to the levels of the target organ inflammation.²⁶ Excessive inflammatory response has been confirmed to play an important role in the pathogenesis of ACLF. FH may also be involved in the pathogenesis of liver failure due to its role in inducing T cells to release chemokines and cytokines, and to produce autoantibodies.^27,28 Fan et al.²⁹ found that FH deficiency leads to a higher level of reactive oxygen species production. ACLF patients with poor prognosis may have more severe inflammatory response and oxidative stress because of relatively less FH. The high expression of autoantibody of FH is an important difference in immune response between ACLF patients with different prognoses. This is our original research and there is no relevant report in the previous literature. We will conduct in-depth research on the role of FH in the specific pathogenesis in ACLF in the next step.

It is necessary to declare that the number of cases in our study is limited, which lead to the limitations of the findings based on the study. In our research, viral hepatitis is the most predominant pathogeny of chronic liver disease in ACLF, followed by alcoholic liver disease and drug-induced liver injury. Among ACLF patients, there was no statistically significant difference in the titer of anti-FH/Fumarate hydratase autoantibodies between HBV-ACLF and non-HBV ACLF patients.

In conclusion, the anti-FH antibody in serum may be a potential marker for predicting the prognosis of ACLF. Our study may provide new clues to the prognostic research of ACLF. Further studies should be conducted to investigate correlations between LFAAs levels and outcomes of ACLF patients.

This work was supported by the Beijing Municipal Administration of Hospitals Clinical Medicine Development of Special Funding Support (XMLX201830, ZYLX202125; M.L.); National Natural Science Foundation of China (81770611; F.R.); 2018 Beijing Youan Hospital Scientific Research Project for Young & Middle Aged Talent's Cultivation (YNKTTS20180119; L.W.), Beijing Municipal Administration of Hospitals Incubating Program (PX2021065; L.W.), and National Natural Science Foundation of China (82100653; L.W.)

Authors would like to thank all the ACLF patients included in our study, and the medical and nursing team of the Second Department of Liver Disease Center, Department of Oncology, and Beijing Institute of Hepatology, Beijing Youan Hospital, Capital Medical University, who are caring about the ACLF patients.

No potential conflict of interest relevant to this article was reported.

Study concept and design: M.L., F.R., L.W. Data acquisition: M.L., B.X., S.Z. Data analysis and interpretation: T.W., S.C., Y.L., X.H. Drafting of the manuscript: L.W., T.W. Critical revision of the manuscript for important intellectual content: M.L., F.R. Statistical analysis: S.C., Y.L. Obtained funding: M.L., F.R., L.W. Administrative, technical, or material support; study supervision: M.L., F.R. Approval of final manuscript: all authors.