An Innovative Method to Identify Autoantigens Expressed on the Endothelial Cell Surface: Serological Identification System for Autoantigens Using a Retroviral Vector and Flow Cytometry (SARF)

Autoantibodies against integral membrane proteins are usually pathogenic. Although anti-endothelial cell antibodies (AECAs) are considered to be critical, especially for vascular lesions in collagen diseases, most molecules identified as autoantigens for AECAs are localized within the cell and not expressed on the cell surface. For identification of autoantigens, proteomics and expression library analyses have been performed for many years with some success. To specifically target cell-surface molecules in identification of autoantigens, we constructed a serological identification system for autoantigens using a retroviral vector and flow cytometry (SARF). Here, we present an overview of recent research in AECAs and their target molecules and discuss the principle and the application of SARF. Using SARF, we successfully identified three different membrane proteins: fibronectin leucine-rich transmembrane protein 2 (FLRT2) from patients with systemic lupus erythematosus (SLE), intercellular adhesion molecule 1 (ICAM-1) from a patient with rheumatoid arthritis, and Pk (Gb3/CD77) from an SLE patient with hemolytic anemia, as targets for AECAs. SARF is useful for specific identification of autoantigens expressed on the cell surface, and identification of such interactions of the cell-surface autoantigens and pathogenic autoantibodies may enable the development of more specific intervention strategies in autoimmune diseases.


Introduction
Inappropriate humoral and cellular immune responses mediate the tissue damage in autoimmune diseases, and the outcome of an autoimmune disease is in�uenced mainly by the tissue distribution of target self antigens [1]. e pathogenesis of most autoimmune diseases is highly complex and involves multiple cellular and humoral pathways. One part of the humoral arm of the immune assault is caused by autoantibodies, and the mechanisms of autoimmune damage mediated by many autoantibodies have been studied [2]. Clinically, speci�c autoantibodies are critical for the diagnosis, classi�cation, and monitoring of autoimmune diseases [2].
Autoantibodies cause damage through a number of mechanisms, including the formation of immune complexes, cytolysis or phagocytosis of target cells, and interference with cellular physiology [3]. e cellular localization of the target antigen is believed to play a critical role in the pathogenetic potential of autoantibodies [4]. Intracellular proteins are preferential targets of autoantibodies in autoimmune diseases, but many questions remain unanswered regarding how autoantibodies against intracellular proteins play pathogenic roles. In contrast, it is generally accepted that autoantibodies against integral membrane proteins are usually pathogenic [1]. Some autoantibodies have been clearly con�rmed to be pathogenic in several autoimmune diseases, and a model for customized and speci�c therapeutic approaches against a highly pathogenic subset of autoantibodies using small molecules have been reported [5].
In 1971, Lind�vist and �sterland �rst described autoantibodies to vascular endothelium based on indirect immuno�uorescence (IIF) experiments [6]. ese autoantibodies were called anti-endothelial cell antibodies (AECAs) and were de�ned as autoantibodies targeting antigens present on the endothelial cell (EC) membrane [7]. As target antigens of AECAs are present on the ECs, which are always in contact with these circulating antibodies, AECAs have the potential to induce vascular lesions directly. Here, we present a review of AECAs and a novel method for identi�cation of cellsurface autoantigens.
AECAs are detected even in healthy subjects [25,26]. ese natural autoantibodies interact with living ECs with lower affinity as compared to pathologic AECAs, and their antigens are highly conserved protein families. ey contribute to modulate endothelial function with protective anti-in�ammatory and anti-thrombotic functions [26].

Detection and �denti�cation of AECAs.
Methods for detection of AECAs have not been standardized, and a number of methods have been reported, including IIF, cellbased-enzyme linked immunosorbent assay (ELISA), �ow cytometry, radioimmunoassay, western blotting (WB), and immunoprecipitation [22,23]. As these each of methods have advantages and disadvantages, use of different technical approaches to obtain more robust data is recommended [7].
Human umbilical vein endothelial cells (HUVECs) are commonly used as a substrate, but antigen patterns of ECs differ among other ECs, passage numbers, and culture conditions [27]. It is also important whether ECs are �xed or not because �xation induces permeabilization of the EC membrane, and intracellular antigens become accessible to antibodies [22]. e results of AECA positivity were therefore not considered in the same light, and the prevalence of AECAs differed among studies (Table 1). Miura et al. recently reported a novel solubilized cell-surface protein capture ELISA for detection of AECAs [28], and further evaluation and standardization are needed.

Pathogenicity of
AECAs. An experimental animal model for pathogenicity of AECAs was reported by Damianovich et al. [29]. In their experiment, BALB/c mice were actively immunized with the puri�ed AECAs from a patient with granulomatosis with polyangiitis. ree months aer a booster injection with human AECAs, mice developed endogenous AECAs, and histological examination of lungs and kidneys revealed both lymphoid cell in�ltration surrounding arterioles and venules.
AECAs have been shown to be correlated with disease activities, and have the potential to induce vascular lesions because their targets are expressed on ECs that are readily accessible to these circulating antibodies. AECAs are also considered to play roles in the development of pathological lesions by a number of methods as described below [22,23,[30][31][32].
e �rst is the cytotoxicity of ECs through complementdependent cytotoxicity (CDC) and antibody-dependent cellmediated cytotoxicity (ADCC). CDC activity of AECAs was reported in patients with SLE, Takayasu arteritis, hemolyticuremic syndrome, and Kawasaki disease [7,24,[33][34][35]. Recently, we con�rmed that �bronectin leucine-rich transmembrane protein 2 (FLRT2) is a novel target antigen of AECAs in SLE, which exerts direct cytotoxic effects through CDC [9]. e second is the induction of coagulation. AECAs may exhibit procoagulant effects by the production of tissue factor in SLE and the release of heparin sulfate in systemic sclerosis (SSc) [36,37].
e third is the induction of apoptosis. AECAs may induce EC apoptosis through CD95 or cross-reaction with anti-phospholipid antibodies [38][39][40]. Dieudé et al. reported that heat-shock protein (Hsp60) bound to ECs and induced phosphatidylserine exposure and then apoptosis [41].
Alard et al. reported that recognition of cell-surface adenosine triphosphate (ATP) synthase in the low pH microenvironment contributes to intracellular acidi�cation of ECs, which may induce cell death and trigger in�ammation [43].
As described above, there is a great deal of evidence that AECAs play pathogenic roles in collagen diseases. Identi�cation of targets of AECAs is required because (a) antigen-speci�c detection systems are important for establishing diagnostic tools and standardization of AECAs measurement, (b) identi�cation will enable thorough analysis of the pathogenicity of AECAs, and (c) AECA-autoantigen interactions may be good targets for speci�c therapeutic approaches against highly pathogenic autoantibodies.

�� �e��no�ogies for ��enti��ation of
Autoantigens for AECAs e prevalence of AECAs varies according to the type of ECs used for detection [44]. It was demonstrated that AECAs cross-react with human �broblasts [45], and partial inhibition of AECA activity was documented by absorption of the AECA-containing sera with mononuclear cells [8]. It was also reported that a structure shared by platelets and ECs was recognized by a subset of AECAs [46]. ese data suggested that the target antigens of AECAs may include not only EC-speci�c but also non-EC-speci�c molecules. Target antigens of AECAs have been investigated intensively, but they are heterogeneous, and the following classi-�cation of target antigens was proposed� membrane component, ligand-receptor complex, and molecule adhering to the plasma membrane [8]. e EC autoantigens may be either constitutively expressed or translocated from intracellular compartment to membrane by cytokines, such as IL-1 and tumor necrosis factor (TNF ), or physical effects [8,47]. e reported autoantigens and their pathogenicities are summarized in Table 2 [7, 9, 22-24, 42, 43, 47-56].
Several molecules can bind to ECs and are called "planted antigens" for AECA presumably via charge-mediated mechanisms, a DNA-histone bridge, or a speci�c receptor. Myeloperoxidase, DNA, and 2-glycoprotein I ( 2-GPI) are thought to adhere to ECs during incubation of ECs with sera from patients. Extracellular matrix components, such as vimentin, may also be target antigens for AECAs [57]. Proteinase 3 (PR3) could represent another potential cryptic target antigen [58]. PR3 has been maintained to migrate to the plasma membrane of ECs, following stimulation [8].
As methods for identi�cation of target antigens of AECAs, immunoprecipitation and WB of glycoproteins from the EC membrane with AECA-positive sera have been used [8,23]. Although numerous protein bands were reported as candidates for target antigens by this method, some of the bands were considered to be artifacts [8], and further identi�cation of given bands was also sometimes difficult.
Alternative methods have been developed, such as proteomics analysis using two-dimensional electrophoresis followed by matrix-assisted laser desorption ionization time of �ight mass spectrometry [8,23] and expression libraries [8,42,56].
Proteomics analysis identi�ed vimentin, Hsp60, voltagedependent anion-selective channel 1 (VDAC-1), peroxiredoxin 2, and ATP synthase as targets for AECAs [41,43,[48][49][50]. Expression libraries also identi�ed tropomyosin, Tplastin, and RLIP76 [42,56], and these technologies are therefore promising. e problem is that most of the molecules reported to date as targets for AECAs are intracellular proteins (Table 2) although AECAs must be directed against the cell surface. ese two methods are not speci�c for detecting cell-surface molecules rather than intracellular molecules. In addition, extraction of some membrane proteins has been reported to be difficult in proteomics analysis, and this may make it difficult to identify such proteins as AECA targets [7].
To overcome this problem, we constructed a novel expression cloning system for speci�c identi�cation of cell-surface antigens [9], which we call serological identi�cation system for autoantigens using a retroviral vector and �ow cytometry (SARF) (Figure 1), and we have con�rmed that this system is useful to identify autoantigens expressed on the EC surface [9].

�� Strateg� for ��enti��ation of Ce���Surfa�e
Autoantigens: SARF 4.1. Generation of HUVEC cDNA-Expressing Cells ( Figure  1(a)). Our strategy to identify AECA target molecules involves use of a retroviral vector system and �ow cytometry [9]. As described previously, antigen patterns of ECs differ among other ECs [27]. Because we used HUVECs as a substrate for AECAs measurement, we generated a HUVEC cDNA library using HUVECs grown in the same conditions as for AECAs measurement and ligated it into the retroviral vector, pMX [59]. en, the HUVEC cDNA library in pMX was retrovirally transfected into the YB2/0 rat myeloma cell line [60]. As the localization of cellular molecules depends on their structures, only cell-surface molecules are expressed on the surface of YB2/0 cells transfected with the HUVEC cDNA library. (Figure 1(b)). AECAs can bind only to cell-surface molecules in �ow cytometry. erefore, sorting of IgG-binding cells can concentrate and isolate cells expressing target molecules for AECAs on the cell surface. Aer staining of HUVEC cDNAexpressing YB2/0 cells with AECA IgG and secondary antibody, cells with strong �uorescent signals are sorted by �ow cytometry. is step of sorting is repeated for several rounds to concentrate AECA IgG-binding cells. Aer concentration,

�.�. I������������ �� ����� C���-������� A�����������.
Aer polymerase chain reaction (PCR) ampli�cation and cloning of HUVEC cDNA inserted into the genomic DNA of cloned cells, DNA sequencing can be performed followed by BLAST analysis, which enables the identi�cation of the inserted cDNA. In this step, microarray analysis is an alternative method to identify the inserted cDNA. Next, an expression vector of the identi�ed cDNA is generated and transfected into a cell line that does not express the identi�ed protein.
Finally, it is necessary to con�rm that AECA IgG shows binding activity to 7-amino-actinomycin D-(7-AAD-) negative identi�ed protein-expressing cells. If the binding activity is con�rmed, it can be concluded that the identi�ed protein is a novel autoantigen.

FLRT2.
We reported the membrane protein FLRT2 as a novel autoantigen of AECAs in patients with SLE based on results obtained using SARF [9]. FLRT2 is type I transmembrane protein located on the plasma membrane [61]. FLRT2 was shown to be expressed in the pancreas, skeletal muscle, brain, and heart with Northern blotting [61], and we con�rmed the expression of FLRT2 on HUVECs and other ECs by �ow cytometry and IIF [9]. Anti-FLRT2 antibody activity accounted for 21.4% of AECAs in SLE, and anti-FLRT2 activity was signi�cantly correlated with low levels of complement C3, C4, and CH50 [9]. Anti-FLRT2 antibody induced CDC against FLRT2-expressing cells including ECs, indicating that anti-FLRT2 autoantibody may exhibit direct pathogenicity [9].

ICAM-1.
As AECAs can be detected in patients with collagen diseases, especially SLE, RA, and Takayasu arteritis [9], we further attempted to identify the autoantigens using SARF. One sample (X10-3) from an RA patient showed strong AECA activity (Figure 2(a)), and we selected this serum sample as the prototype of AECA for subsequent cell sorting. Using SARF, HUVEC cDNA-expressing YB2/0 cells were stained with X10-3 IgG and �uorescein isothiocyanate-(FITC-) conjugated secondary antibody, and cells with strong FITC signals were sorted by �ow cytometry (Figure 2(b)). Aer the 4th sorting, cells bound to X10-3 IgG were markedly increased (Figure 2(c), le), and the C5 clone was established from the X10-3 IgG-binding cell population by the limiting dilution method (Figure 2(c), right). Microarray analysis revealed that the signal of ICAM-1 was signi�cantly increased (2 6.16 -fold), and we con�rmed that the ICAM-1 cDNA was inserted into the genomic DNA of X10-3-C5 clone ( Figure  2(d)). �e also con�rmed the expression of ICAM-1 on the X10-3-C5 clone (Figure 2(e)). Next, we generated an expression vector for ICAM-1, which was transfected into YB2/0 cells. X10-3 IgG showed signi�cant binding activity to 7-AAD-negative ICAM-1-expressing YB2/0 cells (Figure 2(f)), indicating that X10-3 IgG has anti-ICAM-1 activity. us, the membrane protein ICAM-1 was identi�ed as a novel autoantigen of AECA in RA. ICAM-1 is an immunoglobulin-(Ig-) like cell adhesion molecule expressed by several cell types, including leukocytes and ECs. ICAM-1 plays an important role in both innate and adaptive immune responses. It is involved in the transendothelial migration of leukocytes to sites of in�ammation, as well as in interactions between antigen presenting cells (APC) and T cells (immunological synapse formation) [62].
ICAM-1 was also con�rmed to transduce signals �outside in" [63,64]. e cross-linking of ICAM-1 with monoclonal antibodies was reported to activate the mitogen-activated protein kinase (MAPK) kinases ERK-1/2 and/or JNK [65][66][67]. e activation of ERK-1 lead to AP-1 activation [66], the ERK-dependent production and secretion of IL-8 and RANTES [67], and upregulation of VCAM-1 on the cell surface [66,68]. ICAM-1 cross-linking can also upregulate tissue factor production [69] and proin�ammatory cytokines, including IL-1 [70]. Lawson et al. reported production of anti-ICAM-1 IgM aer cardiac transplantation, and the antibody induced robust activation of the ERK-2 MAPK pathway [71]. e use of anti-ICAM-1 antibody was examined for the treatment of RA, but the second course of therapy was associated with adverse effects suggestive of immune complex formation [72]. Identi�cation of anti-ICAM-1 antibody in a patient with RA suggested that this autoantibody may exhibit such pathogenic roles.

Pk (Gb3/CD77).
Using serum from an SLE patient who showed hemolytic anemia, SARF revealed that cDNA inserted into the cloned cells that were sorted with this AECA-IgG was alpha 1,4-galactosyltransferase (A4GALT).   is AECA showed signi�cant binding activity to 7-AADnegative A4GALT-overexpressing YB2/0 cells. e A4GALT locus encodes a glycosyltransferase that synthesizes the terminal Gal 1-4Gal of Pk (Gb3/CD77) glycosphingolipid [73,74]. is means that synthesis of the terminal Gal 1-4Gal is needed for the binding of this AECA-IgG. Gb3 is the Pk blood group antigen and has been designated CD77 [74]. Monoclonal antibodies against Pk (Gb3/CD77) are used as markers for Burkitt's B-cell lymphoma and are able to initiate apoptosis [75]. Pk (Gb3/CD77) plays a direct role in the entry of Shiga toxin into the cell [76], and the presence of Pk (Gb3/CD77) in the ECs of the kidney accounts for the development of hemolytic uremic syndrome during bacterial infection with Shigella species that produce verotoxin [77]. e anti-Pk (Gb3/CD77) antibody was reported to cause acute intravascular hemolytic transfusion reactions and recurrent spontaneous abortions due to damage to the placenta [73,78]. ese data suggested that Pk (Gb3/CD77) is one of the target antigens of AECAs in SLE patients manifesting hemolytic anemia, and that anti-Pk (Gb3/CD77) antibody may exhibit some pathogenic roles.
Identi�cation of A4GALT indicated the usefulness of SARF, which can be used to identify genes that encode not only the membrane protein itself, but also the transferase(s) responsible for modifying the membrane protein.
As described above, this system is very useful for iden-ti�cation of cell-surface autoantigens. Although this system seems to present difficulties in sorting cells at very low frequency, we could isolate and clone autoantigen-expressing cells by repeated sorting.
As AECAs are a heterogeneous group of autoantibodies that target ECs, it is predicted that there are different autoantigens. us, it is important to determine the clinical signi�cance and potential pathogenicity of identi�ed autoantibodies. If an autoantibody is speci�c for a disease or pathophysiology, it could be used as a marker for diagnosis or classi�cation according to the underlying pathophysiology. At the same time, the pathogenic potential of the autoantibody should also be examined. Along with in vitro studies mentioned previously, experimental animal models of iden-ti�ed autoantibody should be constructed to determine the pathogenetic reactions in vivo.

Summary
AECAs are considered to be critical, especially for vascular lesions in collagen diseases, but most are directed against molecules localized within the cell and not expressed on the cell surface. In addition to conventional immunoprecipitation and WB, proteomics and expression library analyses have been performed to identify the targets for AECAs with some success. SARF was developed to identify autoantigens expressed on the EC surface with greater sensitivity. Using SARF, we successfully identi�ed three different membrane proteins as targets for AECAs: FLRT2 from patients with SLE, ICAM-1 from a patient with RA, and Pk (Gb3/CD77) from an SLE patient with hemolytic anemia. Using this technology, it may be possible to determine cell-surface autoantigens of AECAs and achieve a comprehensive understanding of AECA-mediated vascular injury. Furthermore, SARF can be used when autoantibodies against cell-surface molecules are considered to take part in autoimmune diseases. e identi�cation of such pathogenic autoantibodies may enable the development of more speci�c intervention strategies in autoimmune diseases.
�on��ct of �nterests e authors declare that they have no con�ict of interests.