The beginning of GPX2 and 30 years later

Abstract

We published the first paper to characterize GPX2 (aka GSHPx-GI) as a selenoenzyme with glutathione peroxidase activity in 1993. Among the four Se-GPX isozymes, GPX1-4, GPX1 and GPX2 are closely related in terms of structure, substrate specificities, and subcellular localization. What sets them apart are distinct patterns of gene regulation, tissue distribution and response to selenium. While we identified the digestive tract epithelium as the main site of GPX2 expression, later work has shown GPX2 is found more widely in epithelial tissues with concentration of expression in stem cell and proliferative compartments. GPX2 expression is regulated over a wide range of levels by many pathways, including NRF2, WNT, p53, RARE and this often results in attaching undue significance to GPX2 as GPX2 is only a part of a system of hydroperoxidase activities, including GPX1, peroxiredoxins and catalase. These other activities may play equal or greater roles, particularly in cell lines cultured without selenium supplementation and often with very low GPX2 levels. This could be assessed by examining levels of RNA and protein among these various peroxidases at the outset of studies. As an example, it was found that GPX1 responds to the absence of GPX2 in mouse ileum and colon epithelium with higher expression. As such, both Gpx1 and Gpx2 had to be knocked out in mice to produce ileocolitis. However, we note that the actual role of GPX1 and GPX2 in relation to peroxiredoxin function is unclear. There may be an interdependence that requires only low amounts of GPX1 and/or GPX2 in a supporting role to maintain proper peroxiredoxin function. GPX2 levels may be prognostic for cancer progression in colon, breast, prostate and liver, however, there is no consistent trend for higher or lower levels to be favorable.

Graphical Abstract

Discovery and characterization of GPX2.

Where did we begin?

In the mid 1980’s, Doroshow and Akman of the Department of Medical Oncology and Therapeutics Research at the City of Hope National Medical Center were interested in studying the role of oxidants in human diseases. This was motivated, in part, from the cardiotoxicity induced by an antineoplastic drug, doxorubicin (DOX) and its derivatives. The dose-limiting factor of the anti-cancer drug might hinge on the damaging redox cycling of oxygen [1–4]. An endemic Keshan disease in China had been linked to severe selenium (Se)-deficiency causing fatal cardiomyopathy [5,6]. Several animal studies had linked Se and/or vitamin E with DOX-induced cardiotoxicity [7]. These findings prompted us to explore the role of classical glutathione peroxidase (cGPX; designation used by Maiorino et al. for GPX1 [8]) possessing an antioxidant activity that could reduce hydrogen peroxide and hydroperoxides of fatty acyl groups and the only known mammalian selenoprotein at this time [8–11].

In late 1987, FF Chu joined Doroshow’s lab to help clone the cGPX/GPX1 cDNA by screening two human cDNA libraries. Before her joining the laboratory, Akman had used a 50mer oligonucleotide based on bovine GPX1 amino acid sequence to screen a bovine liver cDNA library and isolated a 69mer of bovine Gpx1 cDNA. This 69mer was then used to screen a human liver cDNA library. This resulted in the isolation of several partial GPX2 clones (lacking the ATG start codon), but not GPX1 clones, presumably due to the abundant nature of GPX2 in the human liver. Chu then used the partial human GPX2 cDNA to screen a hepatocarcinoma HepG2 cDNA library and isolated a full-length GPX2 clone. However, because the initial goal was to study the function of GPX1 in oxidative stress, the GPX2 clone was not pursued immediately. By then, human GPX1 cDNA sequence was published [12]. So, Chu used a 21mer of human GPX1 probe to screen the HepG2 library and isolated a full-length GPX1 cDNA clone. We reported the partial GPX2 cDNA sequence in an abstract in Proc. Am. Assoc. Cancer Res in 1988, referring to the GPX1 clone as Px1 and the GPX2 clones as Px2 (vol. 29, p440; #1751; not available on-line; contact authors for text copy). “Px2” was an informal short-hand designation for the GSHPx-2/GI cDNA clones. The full-length sequence of this clone was published in 1993 [13]. The GPX1 cDNA clone was transfected, and the protein product analyzed in human breast cancer MCF-7 cells, which do not express endogenous GPX1, and compared with the endogenous GPX1 in a doxorubicin-resistant MCF-7 cell line (Adr^r) [14–16]. The study used sodium selenite or Se⁷⁵-labeled selenious acid to support GPX1 synthesis above the levels found with fetal calf serum in the culture media and radiolabel GPX1 for analysis of expression using SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and titration with anti-GPX1 and anti-plasma GPX antibodies. We confirmed that Se modulated GPX1 gene expression at both mRNA and protein levels. In this paper, we also mentioned the Px2 cDNA clones isolated from a HepG2 cDNA library and noted that Px2 was not plasma GPX based on partial amino acid sequences of plasma GPX obtained by Esworthy [17].

Expression of GPX2 cDNA and characterization of the GPX2 protein

Because we were dealing with what appeared to be a new GPX, MCF-7 cells were selected to transfect GPX2 cDNA, because these cells have very low endogenous GSH-dependent hydrogen peroxide-reducing activity, lacked GPX1 or GPX2 mRNA or any native cytosolic selenoprotein in the 75-100 KDa size range with a 22-25-KDa subunit [14–16]. One stable GPX2 transfectant resistant to neomycin selection, named neo-D1 clone, was selected for detailed analysis [13]. Neo-D1 expressed a 22-KDa peptide labeled with [⁷⁵Se]-selenious acid, resolved by SDS-PAGE, indicating GPX2 is a selenoprotein. GPX2 was a tetramer, the same size as GPX1, analyzed by gel filtration chromatography with Sephadex G-200. GPX2 was predominantly located in cytosol, similarly to GPX1. GPX2 was not recognized by polyclonal anti-GPX1 or anti-GPX3/plasma antibodies. GPX2 shared the same substrate specificities as GPX1: the neo-D1 clone had elevated cytosolic and native-sized 88K Da GSH-dependent activity to reduce hydrogen peroxide, tert-butyl hydroperoxide, cumene hydroperoxide and linoleic acid hydroperoxide compared to the untransfected MCF-7 cells and as with GPX1, did not have activity with the GPX4 substrate, phosphatidylcholine hydroperoxide. This work established that GPX2 is an active GPX sharing biochemical properties with GPX1.

GPX2 is expressed predominantly in the gastrointestinal tract

To profile the tissues expressing the GPX2 gene, we used Northern blots to analyze human and rodent tissues [13]. In human tissues, GPX2 mRNA was detected in liver, stomach, small intestine, colon and one of six breast tissues samples analyzed. GPX2 mRNA was not detected in human uterus, placenta, and lung. Although GPX2 is highly expressed in rodent digestive tract, including esophagus, stomach, small and large intestine, it is not expressed in the rodent liver. This suggests that rodents cannot be blindly used as surrogates for studies of GPX2 function in human tissues. Other rodent tissues analyzed that did not have GPX2 mRNA were skeletal muscle, heart, lung, kidney, spleen, brain, testis, epididymis, uterus, and mammary gland. This predominantly gastrointestinal tract (GI) tissue expression pattern of GPX2 prompted us to name it GSHPx-GI or GPX-GI.

Mapping of human GPX2 gene to further establish its distinction from GPX1, 3 and 4

The Human Genome Project did not begin until October 1990 and was provisionally completed in April 2003 (official completion 2022) [18, 19]. Prior to this, genes with strong homology to GPX1 were mapped to chromosome 3, 21 and X using somatic cell hybrids [20]. The authors of that paper indicated that the chromosome 21 and X loci were probably pseudogenes (later documented) [21]. It was important to clarify that the GPX2 gene was distinct from other known GPX genes, i.e. GPX1, GPX3 and GPX4. Probing Southern blots containing genomic DNA from human x hamster somatic cell hybrids with cDNAs of GPX1-4, we mapped GPX2 to chromosome 14, GPX3 to chromosome 5 and GPX4 to chromosome 19 [22]. As expected, we confirmed the three loci containing GPX1-like sequences. Later, using in situ hybridization, we found human GPX2 resides on chromosome 14q23.1 [23]. Additionally, the mouse GPX2 gene was mapped to mouse chromosome 12, a homologous region to human 14q24.1 and a mouse pseudogene was found on chromosome 7 [24,25].

Name change from GPX-GI (GSHPx-GI) to GPX2

GPX-GI was the fourth Se-dependent GPX enzyme characterized, after GPX1, GPX4/PHGPX and GPX3/GPX-P, in that order [9,26,27]. Why was GPX-GI not named GPX4? When we submitted the manuscript of mapping human GPX2, GPX3 and GPX4 genes to chromosome 14, 5 and 19, respectively, the newly established HUGO (Human Gene Organization) gene nomenclature committee asked us to name the genes based on the published record of human cDNA sequences [22, 28]. The mouse GPX1 genomic sequence was first reported in 1987, which was the first report showing selenocysteine was encoded by the UGA codon in mammalian cells [29]. While the human GPX1 cDNA sequence was published in 1987 and the gene mapped in 1988, it was referred to as human glutathione peroxidase and selenium-dependent glutathione peroxidase [12, 20]. Later, Chada, et al. reported on mapping the gene and formally applied to HUGO to establish the gene family nomenclature using their in-laboratory designation for the cDNA clone used in mapping, GPX1 [30]. Akasaka, Mizoguchi and Takahashi published the human GPX2 cDNA sequence in a one-page sequence report, “A human cDNA sequence of a novel glutathione peroxidase-related protein” in 1990 [31]. They also referred this clone as a GSHPx-related sequence since it was obtained by using a mouse GPX1 probe and isolated from a human liver cDNA library. Plasma GPX was the second recognized, since the plasma activity was being discussed as early as 1975 along with its dependence on Se [32, 33]. The problem was, being a selenoenzyme with similar hydroperoxide substrates as GPX1, there was no compelling reason to think it was not a leaked or an actively secreted version of GPX1. It was not until Takahashi and Cohen reported that an anti-GPX1 antibody did not precipitate plasma GPX in 1986 that the possibility was seriously pursued that plasma GPX was a distinct protein and might derive from a different gene [34]. The plasma GPX gene was named GPX3 since the human plasma GPX cDNA sequence was published in 1990 [35]. They used oligonucleotide probes generated from a partial amino acid sequence of plasma GPX to screen a human placenta cDNA library. Recognition of GPX2 probably suffered from the context that GPX1 seemed utterly unique as a selenoperoxidase activity and might have been perceived as one of a kind in mammals, although variation in physical properties of GPX purified from different rat organs and other mammals was noted [11]. That was until 1985, when PHGPX was found to be a selenoenzyme [36]. That and the report on 3 GPX1-like loci in the human genome in 1988 prompted active searches for GPX isoenzymes [31, 35]. The results from screening human liver and HepG2 cell line cDNA libraries suggested GPX1 and GPX2 might have comparable abundance in liver, pointed out in our 1988 abstract. There were several efforts to biochemically and immunologically characterize GPX activity from different tissue sources [11]. Miwa et al. purified human liver GPX in 1983 and found what appeared to be a single protein entity throughout the process and produced antibodies in rabbits that titrated red blood cell (GPX1) and placental GPX activities with some small difference from liver activity, lending no real clue for multiple isoenzymes [37]. Based on our experience, the purification methods used were unlikely to resolve GPX1 and GPX2 [15, 38]. Had they used an anti-erythrocyte GPX1 antibody on the liver samples, GPX2 might have been discovered much earlier (see our comments on ⁷⁵Se-labeled HepG2 samples and anti-GPX1 antibody for Fig. 3 in [14]).

Although phospholipid hydroperoxide glutathione peroxidase (PHGPX; peroxidation-inhibiting protein in first reports) was first described by Ursini et al. in 1982, and it is the 2^nd GPX activity detected as clearly distinct from GPX1 by its monomeric native structure and unique substrates, its human cDNA sequence was first published in 1994 [27, 39]. Thus, it is named GPX4. In fact, because of its unique substrate specificities, it is considered the most important GPX since deletion of the Gpx4 gene in mice results in embryonic lethality [40, 41].

Other members of GPX superfamily GPX5-8 were identified later. Among the eight members, only GPX1-4 and 6 have selenocysteine in the active site. GPX5 and GPX6 are mainly expressed in the male reproductive tract including epididymis, sperm and seminal plasma [42, 43]. GPX3-6 present in the epididymis play a role in sperm maturation and sperm membrane and DNA integrity by regulation of redox balance in the tissue. The last two GPX members, GPX7 and GPX8, are endoplasmic reticulum (ER)-resident protein disulfide isomerase (PDI) peroxidases, and use Cys, instead of selenocysteine, as a redox center [44]. Their main function appears to be H2O2-dependent PDI oxidase in cells and play a role in oxidative protein folding [45].

The different gene expression pattern of Gpx1 and Gpx2 in the GI tract

Since GPX1 and GPX2 have similar substrate specificities, both reduce hydrogen peroxide and fatty acid hydroperoxides but not phospholipid hydroperoxides, they would have similar functions when present in the same cells. To distinguish their possible sub-tissue localization, we used an EDTA and dithiothreitol-containing buffer to elute epithelial cells from rat and mouse small intestine and analyzed GPX1 and GPX2 activity in the fractions representing villus, crypt eluted through time and the submucosa/muscle remnant (after removing the epithelial cells) [38, 46, 47]. Using this crude cell fractionation technique, we found that GPX1 and GPX2 contributed roughly an equal share of GPX activities in the crypt epithelium as determined by anti-GPX1 or anti-GPX2 antibody titration and activity assays as well as a comparison of GPX activity in GPX1-KO (Gpx1−/−Gpx2+/+) and WT mice. GPX1, but not GPX2, was detected in the remnant. Later, using in situ hybridization, we found GPX2 mRNA is highly expressed in the crypt of human small intestine [48]. Florian et al. and Komotsu et al. used immunohistochemistry (IHC) to demonstrate that GPX2 protein is highly enriched in the crypt of mouse ileum and colon glands, while GPX1 protein is expressed evenly between the crypt/gland and luminal compartments [49–51]. Lack of Gpx2 expression in GPX2-KO mice (Gpx1+/+Gpx2−/−) did not reveal pathology to our untrained eyes in the initial studies [52]. With the guidance of Scott Binder (pathologist) and the experience of working with Gpx1 and Gpx2 double-knockout (GPX1/2-DKO; Gpx1−/−Gpx2−/−) mice, we were eventually able to identify elevated colon gland apoptosis in Gpx1+/−Gpx2−/− versus Gpx1+/−Gpx2+/− mice [53]. Part of the problem in assessing mild apoptosis related pathology may have involved the use of mixed C57BL/6 (B6) and 129Sv1 (129) strain mice, which appeared in later studies to be an unfortunate combination for our work; the B6 background restrained pathology observed in Gpx1+/−Gpx1−/− and Gpx1−/−Gpx2+/− mice; the 129 background aggravated pathology in Gpx1+/−Gpx1−/− and Gpx1−/−Gpx2+/− mice [54, 55]. Studies using GPX2-KO mice on a B6 background showed that there was 2-2.5-fold more colon gland apoptosis in GPX2-KO mice versus WT mice and Gpx1+/−Gpx2−/− mice versus Gpx1−/−Gpx2+/−mice [51, 53]. Interestingly, loss of Gpx2 gene expression resulted in overexpression of GPX1 protein and activity, but not Gpx1 mRNA levels [56]. But GPX2 does not appear to compensate the loss of GPX1 activity analyzed in mouse jejunum, ileum and colon epithelium [56]. In the study of GPX1/2-DKO mice, we showed that these two isoenzymes are the major GPX activity to reduce hydrogen peroxide since the DKO mice had no GPX activity in the epithelium of the small and large intestines [56].

Apparently, the high expression of Gpx2 in the crypt epithelial cells helps protect those cells from apoptotic death. In fact, lack of GPX activity, i.e., both GPX1 and GPX2, promoted cell death in the crypt epithelium in GPX1/2-DKO mice [56]. This apoptosis is induced by gut microflora, since germ-fee mice do not have this pathology [57]. The excessive cell death may also trigger cell proliferation, since the GPX1/2-DKO mice also have high levels of mitotic cells in the crypt.

GPX1/GPX2-DKO mice have spontaneous and early-onset ileocolitis

Because of the redundancy of GPX1 and GPX2 activities, generally only elevated crypt/gland apoptosis with resulting distortion of crypt/gland architecture and fewer goblet cells (moderate pathology) was observed in the single KO mice without exposure to environmental stresses [51]. On the contrary, the homozygous GPX1/2-DKO mice had gut inflammation triggered by commensal bacteria colonization [56, 57]. The original colony of GPX1/2-DKO mice tested positive for Helicobacter spp., a known pathogen in mice. Fifteen of 16 GPX1/2-DKO mice on a mixed B6 and 129Sv1 background and on a regular chow diet had moderate (5 mice) to severe (10 mice; colitis) colon pathology before weaning (21 days of age). For mice with a single WT GPx1 allele (GPX1+/−GPX2−/−), 9 appeared normal while 10 showed moderate pathology and 2 exhibited colitis [56]. None of the 12 Gpx1−/−Gpx2+/− mice had colitis. Nine appeared normal while 3 had moderate pathology. Ileum pathology was also observed in Gpx1/2-DKO mice at and after weaning (20-27 days of age:7 moderate, 4 ileitis). Chronic inflammation is a risk factor for cancer, and this was demonstrated in the GPX1/2-DKO mice after they reached adulthood [57].

Redox imbalances, either too much or too little oxidant production, contribute to gut inflammation. Imbalance of redox physiology in intestinal epithelium may cause very-early-onset inflammatory bowel disease (VEOIBD), based on diagnosis before 6 years of age [58]. A major source of reactive oxygen species (ROS) is the NADPH oxidase (NOX) family- NOX1-5 and dual oxidases (DUOX) 1 and 2, which are conserved transmembrane enzymes expressed in a variety of human tissues [59]. NOX1 and DUOX2 are expressed in the GI epithelium and loss of activities was associated with VEOIBD in patients without deficiency in the antioxidant defense system [58]. We found both NOX1 and DUOX2 contribute to ileocolitis in GPX1/2-DKO mice. In triple-KO (TKO) mice, deficient in GPX1/2 and NOX1, all the pathology exhibited in the DKO mice was abolished [60]. However, TKO mice deficient in GPX1/2 and Duoxa (functionally Duox2 in GI tract) had attenuated gut pathology, except crypt apoptosis [61]. GPX2 along with GPX1 protects gut epithelial cells from bacteria-associated inflammatory responses involving NOX1 and DUOX2. Two of the more consistently and highly up-regulated genes in IBD are DUOX2 and DUOXA2 [62, 63]. Our IHC study of human colitis samples suggested that the up-regulation of DUOX2 involved expansion of the range of expression from the surface epithelium into the gland compartment [61].

Regulation of GPX2 gene expression

Selenium (Se) and impact on studies using cell lines

Here and in the next 4 sections, we will discuss some regulatory pathways involved in GPX2 expression and protein levels. GPXs 1-4 and 6 are Se-dependent glutathione peroxidases, which have an UGA codon for selenocysteine, which is at the enzyme active site. However, selenium is not only involved in GPX protein expression, but also involved in mRNA stability. The GPXs mRNAs have a stem-loop structure in the 3’-untranslated region (UTR) known as selenocysteine-inserting sequence (SECIS). The SECIS is recognized by a SECIS-binding protein 2 (SECISBP2/SBP2), and the SECIS:SECISBP2 (RNA:protein) interaction is essential for selenocysteine incorporation or recoding of UGA along with a specific elongation factor (EFSec) and a Sec-tRNA [64]. In addition to SECIS, the SECISBP2 protein also binds to ribosomes and facilitates selenocysteine insertion. A selenium-hierarchy in the order of GPX2 ≥ GPX4 > GPX3 = GPX1, in terms of resistance of mRNAs to Se-deprivation was reported based on studies in cell lines [65]. The stronger binding affinity of SECISBP2 to GPX4 than GPX1 mRNAs may explain the resistance to faster GPX1 RNA degradation than GPX4 and likely GPX2 [64]. However, resistance of the mRNAs to degradation does not imply that GPX2 and GPX4 activity is supported in low selenium conditions, only that the impact may be somewhat less than for GPX1 activity [15, 66, 67]. This was noted in studies in mouse GI tract samples for GPX2 activity as well as in cell lines for protein levels (Western blots) [53,67].

The important message here is that while it is rare to find Se-deficiency in humans and animals on regular diets, studying GPX gene expression in cell lines without Se-supplementation (5-10% fetal calf/bovine serum; FCS/FBS) may result in misinterpretation in studies. Selenium was identified as an essential component of defined cell culture media by 1976 and the link to supporting GPX activity in cultured cells was being examined by 1977 [68–71]. The standard practice of using 5-10% FCS/FBS may not support GPX activity in culture, given variation in selenium levels among different sources of serum [66]. Sandstrom, Carlsson and Markland published a list of 9 cell lines cultured in the same batches of serum (10%) with and without selenite supplementation (100 nM) [72]. The GPX activity increased with supplementation from a high of 1.9-fold to a low representing no increase. It was found that adding 60 nM selenite to culture media with 10% FBS increased the GPX activity of mouse neuroblastoma cells by 30-fold [71]. There are a couple examples of primary culture of tissues where extended culture in 10-25% serum containing media resulted in diminished GPX activity; one case was for adrenocortical cells (6-fold decrease reversed by adding 20 nM selenite; 10% FBS) and second involving monocytes (3.5-fold decrease, 25% human serum, no attempt to reverse) [73, 74]. GPX and PHGPX activity in the HepG2 cell line plus 100nM sodium selenite (5% FBS) is fairly close in magnitude to human liver GPX and PHGPX activity, suggesting that HepG2 might be a reasonable choice for in vitro studies of the role of GPXs in liver or liver tumor-related redox phenomena (surrogate) when the culture media is supplemented with selenite [15, 48].

Not using selenium supplementation may impact the outcomes of studies using cell lines. One study using hepatitis C virus (HCV) replicon-expressing human hepatoma cells alternated between using 10% FBS with and without 50 nM sodium selenite [75]. The GPX2 mRNA results are not shown for all conditions used. The authors stated that GPX2 mRNA was detectable in the absence of selenium supplementation upon induction by retinoic acid (RA) or IFNα (IFNα; Fig 3a) while Western blots of cells treated with RA showed GPX2 protein only with selenium supplementation. Thus, a biological effect (suppression of HCV replicon expression) of endogenous GPX2 gene expression induced by RA occurred only with supplemental selenium. This was substantiated by the study of Wingler et al. who showed that supplemental selenium approaching 50 nM, was required for HepG2 cells to support substantial GPX2 protein production [68]. In the HCV study, the presence of GPX1 can be inferred from the GPX activity analysis in cells with HCV expression that suppresses GPX2 mRNA and protein compared to a control non-HCV expressing cell line. The GPX1 activity seemed to be at parity with GPX2 and was ignored after no impact of HCV was found. One alternative explanation of that study is that after selenium supplementation, the biological effect might have involved both GPX1 and GPX2 [65]. Furthermore, the higher selenium conditions may have enhanced peroxiredoxin (PRDX; thioredoxin peroxidase activities) function by increasing the pool of reduced thioredoxin via increased levels of the selenoprotein family of thioredoxin reductases [76].

Figure 3.

Comparison of GPX2 (left panel) and GPX1 (right panel) transcripts expressed between tumor and normal tissues in GEPIA2 database (gepia2.cancer-pku.cn). Nine set of tissues compared are: BRCA (breast invasive carcinoma), COAD (colon adenocarcinoma), ESCA (esophageal carcinoma), LUAD (lung adenocarcinoma), LUSC (lung squamous cell carcinoma), PAAD (pancreatic adenocarcinoma), READ (rectum adenocarcinoma), SKCM (skin cutaneous melanoma), and STAD (stomach adenocarcinoma). The X axis is the number of samples in each group. The Y axis is the number of transcripts per million (TPM) in a log scale. Red dots represent tumor samples, and green dots represent normal samples. Those having higher GPX2 TPM in the normal than tumor tissues are shown in green, while those having higher GPX2 TPM in the tumor tissues are shown in red. Those tissues that do not have different GPX gene TPM are shown in black.

Table 1 shows a partial list of regulatory pathways indicated to affect GPX2 expression levels along with chemicals and other agents or methods that were used in the studies. We selected four pathways for comment that were presented as part of the discussion on GPX2 above (RA-RARE) or will be raised later in relationship to GPX2 compartmentalization or involve key points, such as NOX1 interaction with GPX2 (NRF2, p53 family-specifically DeltaNp63-gamma and WNT). The remainder are listed in alphabetical order, and we will not be commenting on these.

Table 1.

Pathways and conditions that impact GPX2 Expression

Pathway	agent(s)	tissue/cell line	fold-change GPX2	side notes	Ref.
p62/KEAP1/NRF2	t-butyl hydroquinone/sulforaphane	CaCo2	2-3.5-fold (up)	mRNA	[77]
	15-deoxy-D12, 14-prostaglandin J2	CaCo2	1.6-fold (up)	mRNA	[107]
	NRF-KO mice	colon	2.5-fold (down) from wild type	mRNA	[107]
	NRF-KO mice	lung	2.5-fold (down) from wild type	NQO1; HO-1; GST-Ya/Yc co-regulated	[77; 82]
	NRF-KO mice/hyperoxia	lung	3-4-fold (increase) wild type	2-6-fold less increase in NRF-2-KO	[80]
	fenofibtrate	Rat liver	9-fold up	2 of 19 papers on liver carcinogens	[141]
NRF2/PTEN/AKT	diethylnitrosamine/indole-3-carbinol	Rat liver	11.4-fold up		[142]
RARE	all trans-retinoic acid	MCF-7; HuH7	13-fold (up)	activity/protein detected	[48;75; 89; 93]
p53	p63-gamma; DeltaN p63-gamma	MCF-7	13-fold (up) rough estimate	protein detected; basal cell	[93; 117]
Wnt	Wnt3a/APC (through β-catenin)	NIH-3T3; NIH-3T3 Wnt3a: HT-29; HT-29 APC	3-fold (up; Wnt); 3-fold (down; APC)	TrxR2; TrxR3 co-regulated	[96]
Estrogen receptor	estrodiol	MCF-7	3.5-fold (up)	GPX1; SOD1;CAT co-regulated	[103]
	fulvestrant	MCF-7	3-fold (down)	CAT co-regulated	[103]
Ets/CD13/Nrf1	5-fluorouracil; CD13; ubenimex	PLC/PRF/5	2-fold (up) 5-FU; 2-fold (up) CD13	CD13 induction by overexpression negated by ubenimix	[110]
FOXO1	FOXO1 si	NHEK	2-fold (down)	CAT; SOD1; PRDX 3 co-regulated	[105]
FOXM1	FOXM1 si	NHEK	5-fold (down)	CAT; PRDX; SOD2 co-regulated;	[106]
Hepatitis C	viral replicons	HuH7	10-20-fold (down)	Se issues; impacts protein detection	[75]
HIPPO/YAP	YAP-S127A	CRL-5889; HTB-182; L78	2-10-fold (down)	ROS/RNS 2.5-fold (up-CRL-5889)	[104]
	digitoxin	L78; HTB-182	2.5-4-fold (down)	enhances YAP nuclear retention	[104]
	DeltaN p63-gamma	HTB-182; L78	1.9-30-fold (up)	overexpression counters YAP-S127A	[104]
NLUCAT1/hypoxia	Crispr-CAS9 NLUCAT1-KO	A549	10-fold (down) rough estimate	hypoxia and NRF2 involvement;	[108]
Nkx3.1	Nkx3.1-KO mice	anterior prostate	2.8-5.5-fold (down) KO/wild type	age dependant; PRDX6 co-regulated	[143]
OXR1	OXR1 si	HeLa	4-fold (down)	NRF2-p21 mediated and independent mechanisms; HeLa barely express GPX2 flavenol that kills colon	[119]
PI3K/AKT	cudraflavone C	CaCo2	5-fold (down)	cancer cells	[109]
STAT-3	IL-22	CaCo2	1.75-fold (up)	mRNA	[107]

Regulation of GPX2 gene expression in p62/KEAP1/NRF2 pathway

How a gene is regulated provides clues to its function. The antioxidant role of GPX2 is evident not only by its enzyme activity, but also by its induction by oxidative stress. A well-known pathway of GPX2 gene expression is via p62/KEAP1/NRF2 complex [77]. NRF2 is a nuclear receptor and a transcription factor, which is activated under oxidative stress and subsequently up-regulates many antioxidative genes by binding to the antioxidant response element (ARE) in the promoters (Fig. 1) [78]. Under homeostatic conditions, NRF2 binds to KEAP1, a component of E3 ubiquitin ligase and is targeted for proteosome-dependent degradation. KEAP1 has multiple stress sensors to allow diverse cellular inputs to regulate NRF2 activity and dysregulate autophagy [78]. Essentially, the diverse chemicals which elicit the oxidative stress response share a common property, an electrophilic center that allows them to react with cysteine residues in proteins including KEAP1. The p62 encoded by sequestosome 1 (SQSTM1) is a multifunctional protein which interacts with the NRF2-binding site of KEAP1 and competitively inhibits the KEAP1/NRF2 interaction [79]. Under oxidative stress, p62 is phosphorylated and upregulated by NRF2. Phosphorylated p62 increases its binding affinity for KEAP1 to release NRF2 to activate cytoprotective target genes. The p62 protein functions as a receptor for ubiquitinated proteins, organelles, and microbes, which it sequesters into the autophagosome as a p62/KEAP1 complex.

Figure 1.

NRF2/NFE2L2 pathway. It is downloaded from https://pubchem.ncbi.nlm.nih.gov/pathway/WikiPathways:WP2884. NRF2/NFE2L2 (NFE2-like bZIP transcription factor 2) a basic leucine sipper (bZIP) protein is a transcription factor that regulates genes which contain antioxidant response element (ARE) in the promoters. Under homeostatic conditions, NRF2 forms a complex with KEAP1, a component of E3 ubiquitin ligase to be targeted for proteosome-dependent degradation. However, upon exposure to diverse chemicals with an electrophilic center, KEAP1 will be oxidized and release NRF2 into nucleus to activate ARE-containing genes. Among the NRF2 responsive genes are antioxidant proteins including GPX2, GPX3, SOD and other proteins such as P62/SQSTM1

A role for GPX2 as a target of the NRF2 pathway was identified by Kleeburger’s group in a series of studies on hyperoxic lung injury using Nrf2-KO mice beginning in 2002 [80]. They reported induction of Gpx2 by hyperoxia resulted in a 3-4-fold increase in mRNA levels (48 and 72 hours) and was significantly lower in Nrf2−/− mice compared with Nrf2+/+ mice. In 2012, they reported that Gpx2-KO neonates had more neutrophil infiltration under hyperoxia (100% oxygen, 3 days exposure) than wild-type mice and by bioinformatics found putative Nrf-2 activation sites in the gene [81]. In the intervening years, Banning et al. showed that the GPX2 gene harbored an antioxidant response element, which would make it a down-stream target of Nrf2 [77]. This was demonstrated in reporter construct assays and by chromatin immunoprecipitation (ChIP). Later, it was shown that Gpx2 was induced in mouse lung by cigarette smoke via the Nrf2 pathway (2-fold) resulting in a 50% increase in GPX2 protein [82]. While the GPX2 gene does not have a high frequency of alleles recognized to affect function, the NRF2 gene has promotor and coding regions polymorphisms in mice and humans which affects the levels of expression and protein function with associations to susceptibility for lung injury in hyperoxia and other stresses [83–86].

Induction of GPX2 gene expression by retinoic acid in epithelial cells

We identified retinoic acid response elements (RAREs) in the promoter region of GPX2 gene and found that GPX2 gene expression was induced 13-fold by supplementation of retinoic acid into the culture media of MCF-7 human breast cancer cells, which have virtually no GPX1 or GPX2 gene expression [48]. Although we did not observe retinoid acid induced GPX2 gene expression in HT29 human colon cancer cells, it was likely due to the high levels of basal GPX2 gene expression. Retinoic acid signaling drives differentiation toward the absorptive lineage in intestine organoids [87]. All-trans-retinoic acid (ATRA), a major active metabolite of vitamin A, promotes barrier functions in human-induced-pluripotent-stem cells (iPSCs) by inducing iPSCs differentiation into intestinal organoids [88]. A recent study showed that ATRA, attenuated transmissible gastroenteritis virus (TGEV)-induced apoptosis in IPEG-J2 cells, an intestinal epithelial cell line isolated from the jejunal epithelium of a neonatal unsuckled piglet [89]. TGEV is a member of the coronaviruses; it can infect pigs of all ages and cause diarrhea, dehydration, and high mortality in piglets. TGEV treatment suppressed expression of several antioxidant enzyme including GPX1, GPX2, SOD1 and catalase. Addition of ATRA can reverse the suppression of the expression of these genes. This is a similar circumstance to HCV infection of hepatoma cells mentioned earlier, where ROS are indicated to elicit some of the pathology [75]. Heme oxygenase 1 overexpression in hepatoma cells also reversed the HVC impact further linking oxidative stress to the viral pathology [90]. Since retinoic acid can maintain homeostasis at the epithelial barrier and immunity, it’s induction of the GPX2 gene may also play a role in the anti-inflammatory function [91].

Regulation of GPX2 gene expression by p53 family of transcription factors (TFs) in squamous epithelial cells

The p53 family consists of p53, p63 and p73, each of which has multiple isoforms due to transcription at two separate promoters and alternative or exon-skipping splicing. While p53 functions as tumor suppressor, two isoforms of p63, TAp63 and ΔNp63, regulate distinct target genes and often have opposing regulatory functions [92]. TAp53 is transcribed from a distal promoter and the proteins contain the canonical transactivating domain (TA1), while ΔNp63 is transcribed from the proximal promoter thus is truncated at the NH2 terminus. There are five variants identified for both isoforms denoted with Greek alphabets, α, β, γ, δ, and ε. The ΔNp63, but not the TAp63, isoform is highly expressed in the basal epithelial cells and often amplified in tumor cells. Yan and Chen found GPX2 gene is highly induced by ΔNp63γ and weakly by ΔNp63α, but not by p53 in MCF-7 human breast cancer cells [93]. Activation of GPX2 gene can alleviate hydrogen peroxide-induced cell death via p53 activation and it also indicates that GPX2 activation can promote cancer cell growth.

Not all tissues express p63 [94]. Among the digestive tract sections, esophagus and salivary gland have high and low expression, and p63 is undetectable in the stomach, small intestine and colon [95]. The ΔNp63 isoform is predominately expressed in all benign and neoplastic esophageal squamous cells. In fact, ΔNp63α is the prominent isoform of TP63 expressed in all squamous cells including the skin, head and neck, lung, and some pancreatic, urothelial, and prostate cancers. Expression in the basal cell compartment of breast tissues will be discussed in the stem/progenitor cell section. ΔNp63α coordinates chromatin-remodeling enzymes to orchestrate the tissue-specific enhancer landscape and chromatin remodeling to drive oncogenic target expression during squamous cell carcinoma development [92].

Regulation of GPX2 in the NOX1-associated WNT pathway

GPX2 expression in the mid-lower GI tract is driven, in part, by the WNT proliferative pathway providing one mechanism for elevated GPX2 levels in the stem/proliferative cell crypt/gland compartments of the small and large intestines in the absence of ΔNp63. The GPX2 promoter contains five putative β-catenin/TCF binding sites [96]. Only one of these sites was required to perceive an impact of β-catanin/TCF activation of the GPX2 promoter in cell lines such as HepG2. The results were extended to mouse colon, where induced knockout of β-catenin in isolated colonic gland cells decreased basal GPX2 protein levels by almost 4-fold.

Coant et al found that NOX1 influenced the balance of WNT and Notch pathways in mouse colon progenitor cells, with its expression favoring the WNT proliferation pathway [97]. RAC1, a GTP-dependent activator of NADPH oxidases, was found to be upregulated in adenomatous polyposis coli (APC) negative colon cancer and supported proliferation in a ROS dependent manner [98]. APC normally suppresses WNT signaling. The APC gene is mutated in the bulk of colorectal cancers leading to constitutively active WNT signaling [98]. These findings were further supported in a 2021 study of mouse colon epithelium where NOX1 was localized in stem/proliferative cells and drove proliferation [99]. However, the authors connected cell proliferation to EGFR signaling. NOX1 was also a source of oxidants that activated the WNT pathway in a feed-forward mechanism through oxidation of nucleoredoxin (NRX) and disassociation from dishevelled (DVL) [100]. In this model, WNT initially induces RAC1 GTPase, a component of NADPH oxidases, which activates NOX1 to de-repress the WNT pathway at the DVL step. Reduced NRX sequesters DVL from β-catenin allowing degradation. Oxidized NRX disassociates from DVL, which stabilizes β-catenin so that it can function as a transcriptional activator [100]. Therefore, NOX1 oxidant production has a regular function in the proliferative compartments of the intestine and impact in cancer that seems to require selenium-dependent GPXs to prevent damaging effects, at least in the crypt/gland compartments. In pancreatic and cervical cancer, it is proposed that high GPX2 expression maintains higher levels of mRNA for WNT2 and WNT pathway components, TCF-1 and β-catenin [101, 102].

GPX2 is part of a system of hydroperoxidases that include GPX1, catalase and peroxiredoxins (PRDXs)

Because GPX2 is one of many hydroperoxidase activities in cells besides PRDXs, GPX1 and catalase, detection of GPX2 may not mean that it has any significant effect even after induction of expression [111, 112]. While Winterborn’s estimate that PRDXs would account for ~97% of hydroperoxide consumption in cells may be an overstatement, most researchers agree that PRDXs are a major component of hydroperoxide catabolism [113–116]. With the advent of RT-PCR methods and derivative technologies enabling detection of very low levels of mRNA, GPX2 could then be detected in tissues and cell lines where it is barely expressed (negative by Northern blotting and probing) and/or in tissues confined to stem cell/proliferative compartments as discussed below [101,111, 12, 117–119]. This is coupled with the practice of showing relative changes in levels of GPX2 mRNA without the context of where the initial levels stacked up to other antioxidant genes, and absence of selenium supplementation in culture media leaves doubts in concluding that GPX2 was involved in the described biological effects, based on the assumption it has a direct and significant impact on hydroperoxide levels.

Even large fold-changes in mRNA levels is not a valid standard for assessing GPX2 impact on cellular hydroperoxide metabolism. Human dermal fibroblasts induced to acquire stem cells properties by retroviral transduction with pMXs-hNANOG, pMXs-hOCT4, pMXshSOX2, pMXs-hKLF4, and pMXs-hcMYC, expressed 10,000-fold higher levels of GPX2 (20% serum replacement media; GPX2 was detectable by Western blotting in the induced cells) [118]. However, the final mRNA levels were only ~1% HepG2 cell GPX2 mRNA levels and the cells showed significant up-regulation of PRDXs (up to 10-fold) relative to expression in untransduced dermal fibroblasts [48, 65, 118]. The comparison to HepG2 cell expression suggests GPX2 expression was next to nothing in the dermal fibroblasts and may still be a minor component of the total antioxidant arsenal even after up-regulation. Mouse and rat liver express Gpx2 in the range of 0.07-1.3% of Gpx1 levels [111, 112]. Better standards for analysis of GPX2 are Western blotting or IHC [49, 50, 75, 112]. GPX2 protein was not detected in wild-type mouse liver as anticipated by the low mRNA expression levels. At roughly 7-fold higher mRNA levels the protein was just detectable by Western blotting (~2% GPX1) and at ~35-fold higher, the protein was now easily detected at about 14% of GPX1 levels [112]. For commonly used cell lines there is the example of Hela, which barely expresses GPX2 (see Hela cells in references 101 and 109-sup.Fig. 2) [101, 117]. Another paper using Hela cells mentioned GPX2 to be downregulated by depletion of OXR1 in a p21/NRF2 dependent pathway [119]. They pointed out that since this involved NRF2, many genes might be responsible for the sensitization of the cells to hydrogen peroxide-induced stress after down-regulation of OXR1, and demonstrated this for the NRF2 target gene, HMOX1(heme oxygenase 1). However, the authors state that neither GPX2 nor HMOX1 were detected by Western blotting. A second study also used GPX2 as one marker to suggest the involvement of NRF2 in an antioxidant response downstream of p62 in bladder cancer [120]. Rather than homing in on GPX2, researchers could examine other co-regulated genes in the pathways.

The digestive tract (esophagus to rectum), prostate, bladder, lung, human liver, pancreas, cervix, skin, breast and the trophoblast of mouse embryos express GPX2 at some level (Table 1 and Table 2; mixed methods of detection) [81, 101,102, 120–145]. Outside of the digestive tract, prostate and human liver, normal tissues seem to have low levels of GPX2 expression, as we noted above. In the digestive tract and prostate high expression is generally limited to basal cell (stem cells and proliferative cells) compartments and this seems to be the case for mammary tissues where at the whole tissue level GPX2 expression seems to be low [49–51, 117, 140]. There is evidence of confinement of GPX2 expression to tissue sub-compartments in mouse skin and lung (Fig. 2) [146].

Table 2.

Primary tumors or pathology and chemically-induced tumor models and GPX2 expression

Primary tumor source/or pathology	Chemical initiator	promotor	Fold changes vs. proximal normal	vs. distal normal/normal	side notes/prognosis indicator	Ref.
Esophagus squamous cell carcinoma	NA	NA	2.5-fold IHC (up)	normal barely detectable	low GPX2 bad; HR 5.7, CI 2.34-13.9, p<0.001)	[127]
Barrett’s esophagus	NA	NA		4-fold (up)	mRNA	[121]
Gastric cancer	NA	NA		3.25-fold	mRNA; IHC high GPX2 bad; HR = 1.635; p = 0.021	[126]
Gastric cancer	NA	NA			detected 76% of CD44v9 positive tumors; 60% of CD44v9 negative tumors; IHC	[113]
Colorectal adenomas	NA	NA		4.7-fold (up)	mRNA	[137]
Colorectal cancer	NA	NA		>2-fold (up)	mRNA	[128]
Colorectal carcinoma	NA	NA		0.57-fold (down)	mRNA	[124]
colorectal adenocarcinomas	NA	NA			detected 96%; in situ hybridization	[145]
colorectal adenocarcinomas	NA	NA		increased expression	mRNA; metric unclear; GPX2 not prognostic	[130]
stage II colon cancer	NA	NA			detected 100%; IHC; GPX2 not prognostic	[131]
Prostate cancer	NA	NA		increased expression 29%	highest expression in basal cells; high GPX2 bad; HR 5.63, CI 2.3-13.8, p<0.001	[140]
Bladder and upper urinary tract cancers	NA	NA			GPX2 down with stage (2.5-fold) or node + (3.5-fold)	[120]
Lung NSCLC (adenocarcinoma and SCC)	NA	NA			high GPX2 bad; HR=1.63, CI 1.38–1.93) p=5.1e–09	[132]
Lung NSCLC (adenocarcinoma)	NA	NA			GPX2 not prognostic	[132]
Lung NSCLC (SCC)	NA	NA			GPX2 not prognostic	[132]
Lung adenocarcinoma	NA	NA			GPX2 not prognostic	[133]
Lung adenocarcinoma	NA	NA			GPX2 not prognostic	[133]
Lung adenocarcinoma; cis-platin resistant	NA	NA	2.25-fold (up) vs. sensitive		cell lines cultured from primary tumors	[134]
Lung adenocarcinoma; cis-platin senstive	NA	NA	reference group		cell lines cultured from primary tumors	[134]
Mouse liver	diethylnitrosamine phenobarbital			3.5-fold higher	tumor vs.wild type normal	[112]
Mouse liver thioredoxin reductase 1-KO	diethylnitrosamine phenobarbital			9-fold higher	tumor vs TR-KO normal	[112]
Mouse liver thioredoxin reductase 1-KO	NA	NA		5-fold higher vs. wild type	GPX2 levels 0.03% vs. GPX1 in wild-type; mRNA;	[112]
Rat liver	diethylnitrosamine			10-35-fold higher	derived cell lines in 10% FBS vs. normal liver; GPX2 protein not detected in normal	[135]
Rat liver	diethylnitrosamine thioacetamide			9.2-fold	mRNA: normal rat liver GPX2 levels are less than 1% of GPX1	[136; 112]
Hepatocarcinoma TNM stage 1	NA	NA		1.7-fold (up)	high GPX2 bad; HR 2.430, CI 1.405–4.204, p<0.001)	[144]
Pancreatic cancer	NA	NA	20% to 2-fold (up) mRNA; 4-fold IHC (up)	4-fold IHC (up)	TNM stages III/IV (86%-positvity rate); TNM stages I/II (50%); p=0.006); not prognostic	[102]
Cervical cancer (lymph node −)	NA	NA		2-fold (up) IHC	IHC; GPX2 low but detactable in normal	[101]
Cervical cancer (lymph node +)	NA	NA		3-fold (up)	IHC	[101]
Cervical cancer (metastatic)	NA	NA		10-fold (up)	IHC	[101]
Hras128 rat mamary tissues	NA	NA		3-fold (up) mRNA vs wild type	GPX2 barely detectable in Hras128 rats on Western blots	[122]
Hras128 rat mamary tumors	MNU; DMBA; PhIP	NA		7-fold: 5.9-fold; 5.9-fold (up)	MNU; DMBA; PHIP, respectively; vs. Hras128; GPX2 abundant on Western blots	[122]
Breast Cancer (heterogenous samples)	NA	NA		normal not detectable (IHC)	detected 100%; IHC; reduced by grade; proximal normal occasionally elevated levels	[122]
mouse mammary tissues (PyMT model)	NA	NA			Metastatic sublines, low GPX2 vs. weakly metastatic, high GPX2 on Western blots	[139]
Breast Cancer (heterogenous human samples)				3-fold lower (FPKM)	Basal-like tumors show lowest levels	[139]

Figure 2.

Epidermal sheets were isolated from back skin of 3 each Wt, Gpx1/2-DKO, Gpx2-KO and Gpx1-KO mice [147]. The tissue was processed to assay GPX activity using hydrogen peroxide as substrate [38]. Belly skin yielded similar results. Previously unpublished preliminary work (RSE and FFC). The low level of GPX2 activity (Gpx1-KO) suggests the significant effect of GPX2 on UV-induced skin cancer, as found in Walshe et al., probably derived from expression in a minor compartment of skin, likely the basal cells [147]. GPX1 activity (Gpx2-KO) is elevated (p≤ 0.04) over WT suggesting compensation for loss of GPX2 as observed in GI tract by Florian et al. and for Gpx1 mRNA levels in mouse PyMT1/GPx2 KD tumors compared to control PyMT1 tumors by Ren et al. [51, 139].

We want to emphasize that unless a high expressing-tissue compartment can be identified, it may be meaningless to evaluate GPX2 when it is expressed at very low levels in tissues except for its value as marker of pathways. For rat and mouse liver, no Gpx2 expressing sub-population of cells has been described in the organ. Thus, it is doubtful that large increases in levels of Gpx2 would have physiological significance in many tissues and cell lines under the assumption it has a direct, large impact on hydroperoxide levels.

GPX2 is expressed in the stem/progenitor compartments of many tissues

It is becoming clear that the stem/progenitor (basal cell) compartments of many tissues are the sites of real GPX2 expression. A few studies with stem cells of various origins and more recently breast epithelium suggests that in stem/progenitor cell compartments GPX2 is consistently expressed. Furthermore, one study indicated that GPX2 and glutathione are the major hydroperoxide reducing components. In 2004, the first report of Gpx2 expression being higher in undifferentiated than differentiated mouse embryonic stem cells (cell line derived from the inner cell mass of strain 129 blastocyst; culture media with 10% FCS, no selenium supplementation) was proposed and the outcome effectively replicated in follow up studies and extended to include human embryonic stem cells and induced stem cells [148–151]. The first study was designed to examine genes involved in the apparent resistance of the undifferentiated line to DNA damage. Control of ROS levels was thought to be a feature of the resistance. A short list of antioxidant genes that showed variation in levels during differentiation included Gpx2, Gpx3, Gpx4, Prdx2, Sod2 but not Gpx1. The 2004 paper showed that in the transition from cultured, undifferentiated line to differentiated embryoid bodies, Gpx2 expression declined 5-fold while Prdx2 expression increased 3-fold [148]. ROS/RNS levels (2’, 7’-dichlorodihydrofluorescein diacetate; DCFH-DA) increased 4-fold over the 6-day time span of differentiation [152]. However, it is not possible to infer any dominance in antioxidant function by GPX2 from that paper. The results were reproduced in a follow-up paper where the embryonic cell line was modified to over express telomerase [149]. The authors extended the results to two human embryonic stem cell lines, where the same pattern of GPX2 expression level decline was noted upon differentiation [150]. ROS levels (DCFH-DA) increased a little over 2-fold with differentiation as did mitochondria mass (2.5-fold) and, mitochondria superoxide production (2-fold) [150]. The increased number or performance of mitochondria with differentiation suggests a link to the early findings on stem cells from the 1980s using rhodamine 123 to measure mitochondrial potential [153–156]. The same group showed similar findings in induced pluripotent stem cells derived from human adult dermal fibroblasts [151]. The basis for detection of GPX2 mRNA in induced human dermal pluripotent stem cells was discussed for the paper where GPX2 levels increased by 10,000-fold [118].

Kannan et al. showed that in primary basal mammary epithelial cells GPX2 and glutathione were the major antioxidant combination, declining in luminal cells in favor of GPX1 and PRDXs [117]. There is a minor error when this paper cites references to GPX2 as mitochondrial. The strength of the paper is the use of cells freshly isolated from human mammary tissues with short incubation times for in vitro steps, providing a glimpse into authentic antioxidant protein and GSH levels from tissues and sub-tissue distribution. The results show that luminal cells have greater ROS/RNS levels (dihydroethidium and DCFH-DA) attributed to greater numbers of mitochondria. GSH content (monochlorobimane assay) increased 2.5-fold from basal to luminal cells. Both NRF2 and ΔNp63 were elevated in basal cells over luminal cells, 40% and >1000-fold, respectively, and ΔNp63 protein levels showed a correlation with GPX2 levels (Fig. S2) [117]. GPX1 and PRDX 1-5 protein increased in levels from basal to luminal cells, with PRDX6 and catalase increasing only a small amount. Basal cells levels of GPX2 mRNA were ~70-fold higher (167-fold in another data set) than in luminal cells and GPX2 protein levels appeared greater than GPX1 and to varying degrees PRDX1-6 (GPX2>PRDX3,5,6>PRDX1, 4 and GPX1>PRDX2). Over all GPX activity levels increased from basal to luminal cells; however, the assay used cumene hydroperoxide so there may be some contribution from GSTs. Despite the lower GSH levels, basal cells showed 6-fold greater vulnerability to GSH depletion (BSO) in culture and an 8-9-fold greater vulnerability than luminal cells in culture after shRNA was used to suppress GPX2 expression. Both outcomes suggest a real dependence on GPX2 and GSH in the cultured basal cells but not PRDXs. However, basal cells had a greater vulnerability to hydrogen peroxide and x-rays than luminal cells. This group rated the GPX2/GSH combination as weak compared to luminal cell antioxidant arsenal based on the results from hydrogen peroxide and x-rays. This fits in with the original characterization of stem cells as having a low ROS loads due to fewer and less active mitochondria [153–156].

The studies discussed above suggest that there are tissue compartments where GPX2 expression is elevated enough that GPX2 could be participating in a significant way toward hydroperoxide reduction. This assertion would involve at least partial separation of GPX1 and PRDX expressing compartments from GPX2 expressing compartments. Results like those found in mammary basal and luminal cells should also be found in esophagus, stomach, intestine, colon, prostate and skin. The results we obtained with GPX1/2-DKO mice may be in line with this claim in the sense that while GPX1 has a supporting role, PRDXs and catalase did not offer protection from crypt/gland pathology in the absence of GPX1 and GPX2 [56].

However, low Se diets do not induce GPX1/2-DKO mouse-type pathology in rodents [157]. We and Brigelius-Flohe found that it was only for Gpx1+/−Gpx2−/− and Gpx1+/+Gpx2−/− mice that low Se diets elevated levels of crypt apoptosis with occasional inflammation in Gpx1+/−Gpx2−/− mice [51, 53]. GPX2 activity in Se-deprived Gpx1−/−Gpx2+/− mice was reduced by roughly 90% in our study, putting into question that a major direct impact on hydroperoxide catabolism by GPX2 affected phenotypes [53]. This opens the possibilities that GPX1 and GPX2 buffer over-oxidation and partial inactivation of PRDXs and the partial collapse of the PRDX network precipitates the pathology or NOX1 and other oxidant generating activities are much less under GPX2 restriction [158]. Under the high oxygen and potentially Se-deficient conditions of conventional cell culture, where Se-dependent thioredoxin reductase activity could also be impacted, there is a strong possibility that PRDXs are impacted [76, 159]. This might explain how studies on GPX2 consistently demonstrate a major impact on cellular ROS/RNS levels (DCFH-DA assay) upon knock-down or over-expression of GPX2, although in some studies GPX2 levels were almost vanishing low [93, 118, 119]. Many current studies on GPX2 fail to assess the levels of hydroperoxide catabolism by GPX2 in cell lines relative to other activities (or assess relative antioxidant protein levels as done in Kannan et al.), independent of the DCFH-DA assay, and tend to focus narrowly on GPX2 based on large changes in expression levels [117]. We overlooked the role that PRDX 1-5 have in hydroperoxide metabolism when initiating studies on GPX1/2-DKO mice since they were just being described in mammals; PRDX6 (Aop2) was included in the study of GPX2-KO mice [52, 158]. The roles of PRDXs should not be ignored in future studies of GPXs. We did explore and find overlap in function between GPX1 and GPX2 in the GI tract, took the time to appreciate selenium requirements in cell culture before commencing our initial studies on GPX1-4 in cell lines and acknowledged a role for PRDXs in the pathology of Gpx1/2-DKO mice [158].

GPX2 gene expression levels as a prognosis indicator for tumorigenesis

Using GEPIA2 (Gene Expression Profiling Interactive Analysis; gepia2.cancer-pku.cn), we compared GPX2 gene expression levels in various human normal and cancer tissues and found that elevated GPX2 gene expression is associated with malignant transformation in multiple GI tissues including colon adenocarcinoma (COAD), esophageal carcinoma (ESCA), rectum adenocarcinoma (READ), stomach adenocarcinoma (STAD), as well as in lung adenocarcinoma (LUAD), lung squamous cell carcinoma (LUSC), and pancreatic adenocarcinoma (PAAD) (Fig. 3, left panel). However, decreased GPX2 gene expression is associated with breast invasive carcinoma (BRCA) and skin cutaneous melanoma (SKCM). GPX1 gene expression is only associated with PAAD and SKCM and is elevated in both types of cancer (Fig. 3, right panel).

Banning et al. studied the tumorigenic effect of GPX2 in HT29 human colon adenocarcinoma cells and found the GPX2-knockdown cells had diminished ability to grow anchorage independently in soft agar compared to the control cells (Se-supplemented, 50 nM,10% FCS) [160]. By our estimate using Northern blots and probing with P³² radiolabeled cDNA, GPX2 expression was 4.5-fold greater than GPX1 in HT29 cells, making it possible that GPX2 might be partially competitive with PRDXs (5% FBS, 100 nM selenite) [48]. As in the HVC expressing hepatoma cells ±Se, a selenium effect was observed in some of the in vitro assays conducted for this study in parallel with the work on GPX2-knockdown clones (Se-supplementation, 50nM and 10% FCS vs.10% FCS) [75, 160]. Kranenburg and his associates studied patient-derived colonosphere cultures (30 nM Se supplementation) and found that GPX2 silencing cause accumulation of ROS, sensitizing to H₂O₂-induced apoptosis, and strongly reduced metastasis-forming capacity. These GPX2-suppressed cells also lacked differentiation potential and formed slow-growing undifferentiated tumors [161]. However, overexpression of GPX2 stimulated multilineage differentiation, proliferation, and tumor growth. They found high GPX2 expression was associated with early tumor recurrence in the colon. These studies support the carcinogenic property of GPX2 overexpression in the colon.

GPX2 may have an opposite effect as a tumor suppressor in breast cancer. Ren et al. reported that loss of GPX2 expression stimulates malignant progression of breast cancer [139]. They have analyzed GPX2 mRNA expression from The Cancer Genome Atlas (TCGA) breast cancer datasets (BRCA) and found GPX2 gene expression is decreased about 3-fold (median) compared to matched normal breast tissue (compare to Fig. 3 and Table 2). Analysis of the Gene Expression Omnibus BC dataset (containing 1809 BC patients) showed an association between GPX2 mRNA in three subtypes of BC (luminal B, HER2-enriched, Basal-like) tumors and poor patient survival. All three subtypes showed effects of GPX2 loss on epithelial to mesenchymal transition (EMT) signaling, including Notch and TGF-β. In general, GPX2-loss has broad effects on the tumor phenotype that are consistent with ROS oncogenic signaling. They also found the expression of GPX1, GPX3 mRNA was not correlated with patient survival, while high GPX4 expression in Basal-like BCs was associated with lower survival.

This group also uncovered the role of GPX2 in the mouse polyoma-middle-T (PyMT) mammary tumor cell model and identified that GPX2 loss stimulates malignant progression due to ROS/hypoxia-inducible factor-α (HIF1α)/vesicular endothelial growth factor A (VEGFA) signaling, causing poor perfusion and hypoxia, which were reversed by GPX2 re-expression or HIF1α inhibition [139]. Ingenuity Pathway Analysis revealed a link between GPX2 loss, tumorigenesis, metabolic modulation and HIF1α signaling. Single-cell RNA-sequencing (scRNA-seq) data analysis and bioenergetic profiling revealed that GPX2 loss stimulated the Warburg effect (shifting from mitochondrial oxidation/phosphorylation to aerobic glycolysis for ATP production) in most tumor cell subpopulations. GPX2-KD PyMT tumor cells formed larger tumors inoculated in the mammary fat pad of female athymic nude mice as compared to PyMT1 control cells. The GPX2-KD PyMT tumors were notoriously reddish, implying angiogenesis and high metastasis to the lungs. Overexpression of GPX2 in PyMT cells drastically inhibited the tumor formation in the fat pad.

It is intriguing why GPX2 expression is associated with tumor promotion in GI tissues, but its expression in breast tissue is associated with tumor suppression. GPX1 generally does not have the same range of variation between normal and tumor as GPX2 (Fig. 3). However, a 7.4% lower expression in breast tumor vs. normal prompted examination of GPX1 levels and they were recommended as a prognostic marker for overall survival (high GPX1, better; OS; hazard ratio [HR] = 0.89, 95% CI 0.79–1.00, p = 0.0469) [162]. From Fig. 3, pancreatic adenocarcinomas show significantly up-regulated GPX1. The one study on pancreatic adenocarcinomas (ductal) and GPX1 found an opposite result using IHC (comparing to adjacent tissue) [163]. High GPX1 expression was associated with better overall survival (HR-0.707, 95% CI-0.525-0.951, p=0.022). While melanoma is a second cancer type showing a large difference between normal and tumor, there appear to be no papers reporting on prognosis for melanomas and GPX1 (Fig 3).

The overlapping roles of GPX2 and PRDXs

Since PRDXs can reduce H₂O₂ and other hydroperoxides, what distinguishes them from GPX2? This is far from clear since similar claims of variation in expression levels and impact have been made for one or several PRDXs in almost every cancer and cell type where GPX2 has been implicated to have an effect (partial listing of references that can be found (163–186); can be compared to Table 2) [164–187]. This often involves very similar methods, generally proceeding from differences in expression levels between tumor and normal (or pathological; e.g. Barrett’s esophagus) to validation of effects and examination of mechanism in cell line surrogates. Invariably, an impact of overexpression or knockdown in cell lines reveals differences in ROS/RNS levels, generally determined by using DCFH-DA (common starting point), and tending to reveal significant alterations. Oddly, no matter whether GPX2, PRDXs, HMOX1 or even GST-π (GSTP1) has altered expression in cell lines (knock-down or over-expression), the magnitude of the differences in fluorescence in the DCFH-DA assay are remarkably comparable (see GPX2 vs. HMOX1, sup. Fig. 7 [119]; GPX2 vs. GSTP1, [140] vs. [188]) [119, 140, 188]. This encourages the pursuit of individual antioxidants enzymes as having the major role in the cellular phenomena under study.

It is seldom clear that GPX2 could have such a major impact, particularly in cell lines cultured under potentially sub-optimal conditions for selenoprotein production and, in some cases, lacking in significant levels of GPX2, that is, undetectable by Western blotting [119]. The only recent paper to provide strong evidence that GPX2 could account for a major portion of hydroperoxide reduction was Kannan et al. for the basal cell compartment of breast from human subjects, almost invariably representing selenium-sufficient conditions [117]. That determination required measurement of protein levels of GPX1, catalase and PRDXs. In the absence of clear indications that GPX2 is a major feature in hydroperoxide catabolism in many cases, the answer may be that GPX2 is merely a marker of pathways, where co-regulated genes are responsible for tumor promotion or suppression.

It is possible that we have not identified all the functions of GPX2 and by extension, GPX1. This could represent interactions that impact levels of other gene products, effects on mitochondrial performance or sparing the PRDX network from over oxidation [118, 119 139, 158, 189–191] . GPX2 was detected in the nucleus of cells in the stomach, ileum and colon [49]. While GPX1 seems to interact with selenium binding protein 1 (SBP-1) and this does have an effect on SBP-1 mRNA levels (attributed to ROS levels and ARE sites in the SBP-1 gene promotor), GPX1 and GPX2 do not have protein-protein interactions like PRDXs [192, 193]. However, co-precipitation and ChIP assays using GPX2 antibodies might show associations of GPX2 with nuclear proteins and DNA binding proteins and possibly reveal other roles for GPX2.

Are there essential GPX2 functions?

Whether the expression of GPX2 and GSH in mammary basal cells represents some fundamental feature of antioxidant processes that has so far eluded us or is just a happenstance of evolution that manages to work is unclear. Outside of the mid to lower GI tract, it is doubtful that Gpx2 has a major impact on the normal development and function of the tissues where it can be detected. GPX2-KO and Gpx1+/−Gpx2−/− dams don’t show any defects that compromise fostering of pups, indicating nominally functional mammary tissues. This might be a consequence of GPX1 compensating in experimental GPX2 deficiency since they have an overlapping hydroperoxide substrate range (Fig. 2) [56, 139]. Even with possible compensation by GPX1, the stresses of hyperoxia or OVA-aluminum-induced allergy resulted in more inflammation of lung in the GPX2-KO mouse sets and ultraviolet A exposure of skin resulted in more cutaneous squamous cell carcinomas in the GPX2-KO than in wild-type or Gpx1-KO mice based on both incidence and multiplicity (p<0.05) [81, 146, 147]. Within the GI tract, GPX2 may subtlety modulate stem cell fates after weaning and this had an impact on weight gain; GPX2-KO mice showed ~18% less weight gain than wild-type mice from 4-12 weeks of age [194]. Colon expression of the mRNA for the stem cell marker, Lgr5, was significantly lower in the GPX2-KO mice. That may be consistent with GPX2 and GPX1 cooperating to protect from the damage of WNT-related NOX1 oxidant generation and adjustments to levels of WNT receptor components (LGR5) made in the absence of GPX2. In some sense this puts us back where we started between 1993 and 2001 with the finding of high GPX2 expression in the basal compartment of the GI tract epithelium and the collapse of the crypt/gland epithelial barrier in the intestine and colon in GPX1/2-DKO mice, later found to be related to the normal stresses of growth and microflora colonization [ 13, 38, 48–50, 56, 57, 158].

Concluding remarks

In this review, we presented the history of GPX2 discovery and characterization. There has been tremendous progress made towards the characterization of tissue expression and roles of GPX2 in several tissues, mostly in the digestive tract, breast, lung, skin, and stem cells. Although GPX2 has similar substrate specificities as GPX1, the regulation of GPX2 gene expression and compartmentalization of GPX2 protein in tissues suggests that GPX2 is involved in redox homeostasis in a complementary manner to GPX1. On the other hand, when studying GPX2 function, one should make sure that it is the predominant form in the cells or tissue of interest. In cells co-expressing both GPX1 and GPX2, the damage is evident in the absence of both isoenzymes (such as the crypt epithelium of intestine). It is unclear whether the antioxidant effect of GPX2 (and GPX1) is based directly on reducing the bulk of ROS or indirectly on protection of PRDXs, another family of peroxidases, from their inactivation by ROS. Recently, with more systemic approaches, combining RNA-seq pathway analysis, GPX2 appears to play a role in tumorigenesis and cancer progression. Relative GPX2 gene expression levels appear to be a prognostic marker for cancers in the GI and breast tissues. Conceivably, the changes in GPX2 expression levels could be used as a guide to treatment selection or an indicator for the effectiveness of cancer treatments.

Highlights.

• GPX2 was discovered at least twice by investigators screening human liver cDNA libraries with bovine or mouse GPX1 probes, one group trying to clone human GPX1. At this phase in the work no one had any idea what the GPX2 cDNA represented.

• GPX2 was eventually characterized as a cytoplasmic selenoprotein with GPX activity very similar to GPX1 and high expression levels in the gastrointestinal tract (GI). A role for GPX2 as a protective factor in the crypt/gland regions of the mid-lower GI was indicated by work with GPX1-knockout mice, GPX2-knockout mice and a combination of the two lines.

• Generally independent regulation and compartmentalization define semi-independent roles for the GPX2 and GPX1 isoenzymes. There is interaction between the two in the GI tract and likely other tissues.

• More recent work has broadened the range of tissues with GPX2 expression with suggestion of roles for GPX2 with impact on tumororigenesis and metastasis in several tissue sites. However, much work is reported without adequate context for GPX2 as a single component in an array of hydroperoxidases, including peroxiredoxins, other GPXs and catalase and performed under conditions where GPX2 protein levels may be miniscule due to cell line source and/or inadequate selenium in culture media.

• Possible dominance of hydroperoxide reduction by GPX2 in basal cell compartments (stem and proliferative cells) is indicated by a few studies and represents a possible future direction for GPX2 research, including work on cancer.

Abbreviations:

129

129Sv1 (mouse strain)

BSO

buthionine sulfoximine

B6

C57BI/6 (mouse strain)

CAT

catalase

DCFH-DA

2′, 7′-dichlorodihydrofluorescein diacetate

DKO

double knockout

GPX

glutathione peroxidase

GPX1

classical GPX, cGPX, aka GPX-1

GSHPx-2,Px2

GPX2 aliases in successive eras

GPX4

GSHPx-GI, GPX-GI, giGPX, GPX-2

PHGPX

phospholipid hydroperoxidase

GST

glutathione S-transferase

HCV

Hepatitis C virus

IHC

immunohistochemistry

KEAP1

Kelch-like ECH-associated protein 1

KO

knockout

NRF2

nuclear factor-erythroid 2-related factor 2

NSCLC

non-small cell lung cancer

PRDX

peroxiredoxin

RNS

reactive nitrogen species

ROS –superoxide and hydrogen peroxide

reactive oxygen species

RT-PCR

Reverse transcription polymerase chain reaction

SDS-Page; Se-selenium

selenium, sodium dodecyl sulphate–polyacrylamide gel electrophoresis

Sec

selenocysteine

Secis

selenocysteine insertion sequence

SOD

superoxide dismutase

TFs

transcription factors