Synergism and Antagonism of Two Distinct, but Confused, Nrf1 Factors in Integral Regulation of the Nuclear-to-Mitochondrial Respiratory and Antioxidant Transcription Networks

There is hitherto no literature available for explaining two distinct, but confused, Nrf1 transcription factors, because they shared the same abbreviations from nuclear factor erythroid 2-related factor 1 (also called Nfe2l1) and nuclear respiratory factor (originally designated α-Pal). Thus, we have here identified that Nfe2l1Nrf1 and α-PalNRF1 exert synergistic and antagonistic roles in integrative regulation of the nuclear-to-mitochondrial respiratory and antioxidant transcription profiles. In mouse embryonic fibroblasts (MEFs), knockout of Nfe2l1−/− leads to substantial decreases in expression levels of α-PalNRF1 and Nfe2l2, together with TFAM (mitochondrial transcription factor A) and other target genes. Similar inhibitory results were determined in Nfe2l2−/− MEFs but with an exception that both GSTa1 and Aldh1a1 were distinguishably upregulated in Nfe2l1−/− MEFs. Such synergistic contributions of Nfe2l1 and Nfe2l2 to the positive regulation of α-PalNRF1 and TFAM were validated in Keap1−/− MEFs. However, human α-PalNRF1 expression was unaltered by hNfe2l1α−/−, hNfe2l2-/-ΔTA, or even hNfe2l1α−/−+siNrf2, albeit TFAM was activated by Nfe2l1 but inhibited by Nfe2l2; such an antagonism occurred in HepG2 cells. Conversely, almost all of mouse Nfe2l1, Nfe2l2, and cotarget genes were downexpressed in α-PalNRF1+/- MEFs. On the contrary, upregulation of human Nfe2l1, Nfe2l2, and relevant reporter genes took place after silencing of α-PalNRF1, but their downregulation occurred upon ectopic expression of α-PalNRF1. Furtherly, Pitx2 (pituitary homeobox 2) was also identified as a direct upstream regulator of Nfe2l1 and TFAM, besides α-PalNRF1. Overall, these across-talks amongst Nfe2l1, Nfe2l2, and α-PalNRF1, along with Pitx2, are integrated from the endoplasmic reticulum towards the nuclear-to-mitochondrial communication for targeting TFAM, in order to finely tune the robust balance of distinct cellular oxidative respiratory and antioxidant gene transcription networks, albeit they differ between the mouse and the human. In addition, it is of crucial importance to note that, in view of such mutual interregulation of these transcription factors, much cautions should be severely taken for us to interpret those relevant experimental results obtained from knockout of Nfe2l1, Nfe2l2, α-Pal or Pitx2, or their gain-of-functional mutants.


Introduction
In all life forms, distinct types of cells are the most basic unit of their biological structures and functions, due to a fact that they are bounded by the membrane lipid bilayers and hence protected from extracellular oxidizing environments to which they have adapted. Of note, eukaryotic cells are evolutionarily compartmentalized by endomembrane systems, as a unique trait that distinguishes from prokaryotic cells [1], insofar as to give rise to distinct functionally specialized organelles, such as the mitochondria, endoplasmic reticulum (ER), and continuous membrane-surrounded nucleus in eukaryotic cells [2]. Such great benefits can allow for cellular respiration, redox metabolism, and biochemical reactions to take place properly in various temporospatial order during diverse physio-pathological life processes. This is owing to the fact that such versatile organelles participate in a large number of distinct cellular functions. Amongst them, the mitochondrion is known as a central site of cellular oxidative respiration, redox metabolism, and biosynthesis, in order to meet a host of energy and growth demands [3,4].
Of particular concern is the mitochondrial origin from aerobic α-proteobacterium, which was firstly surmised to be engulfed by a primordial anaerobic eukaryote (i.e., the endosymbiont hypothesis [5]) or originally merged with its archaeal host to give rise to the first eukaryotic cell (i.e., the standpoint of comparative genomics [6,7]). Thereby, this organelle has its own genetic system with the bacteria-like features, including a compact circular mtDNA genome (from maternal inheritance), along with a simple transcription system that yields multigenic RNA transcripts, and a translational apparatus with similar antibiotic sensitivities to prokaryotic cells. During subsequent endosymbiotic evolution, most of mitochondrial original genes have been lost and/or transferred to the nucleus in the eukaryotic cells, so that the mere reminiscing 37 genes are retained in mammalian mitochondria [3,5]. In this genome, only 13 mitochondrial genes encode its specific proteins as essential subunits of the respiratory chain, whilst the remaining 24 genes encode specific tRNA and rRNA required for its protein translation within the mitochondrial matrix [8,9]. Such a small complement of mitochondrial genes exists over the entire evolutionary process to strike a balance between the host and its endosymbionts [10,11]. However, the mitochondrial genetic system only possesses a semiautonomous nature, such that its mtDNA replication, subsequent transcription, and translation are all subjected to the predominant control by several nuclear gene-encoding factors [9,[11][12][13], e.g., mitochondrial transcription factor A (TFAM), B1 (TFB1M), and B2 (TFB2M) [14][15][16][17]. Transcriptional expression of these three factors is regulated by nuclear respiratory factor-1 (NRF-1) [18][19][20]. Here, NRF-1 is also referred to as α-Pal NRF1 , because it was originally designated α-Pal from directly binding to the palindromic consensus site (5 ′ -TGCGCATGCGCA-3 ′ ) that is essential for the transcription of eIF2α gene [21][22][23], which controls the general protein translation process and also alternative translation under ER stress. Besides, it is of importance to notice that mitochondrial structure and function, as well as its biogenesis, are also monitored dominantly by those nuclear genome-encoded proteins [17,24,25]. Notably, all the other (except the aforementioned 13) subunits of both the mitochondrial respiratory chain and oxidative phosphorylation are transcriptionally controlled by nuclear respiratory factors NRF-1 (i.e., α-Pal NRF1 ) and NRF-2 (originally called GA-binding protein, which is thus abbreviated as GABP NRF2 but with no any homology with α-Pal NRF1 ) binding to distinct consensus sites within these nuclear-encoding respiratory gene promoters [26,27].
Apart from mitochondria, the eukaryotic endomembrane system including the ER and nucleus has been posited to originate from the bacterial-like out membrane vesicles (OMVs) released by the endosymbiont mitochondrial ancestor within the cytosol of its archaeal host at eukaryote origin [1,6]. The OMV-based model accounts for (i) the functional homology of the ER and mitochondrial intermembrane space, with dynamic topological exchanges of Ca 2+ storage and the disulfide relay system required for redox signaling; (ii) the cooperative contributions of the ER and mitochondria to eukaryotic lipid synthesis, as opposed to occurring at the plasma membrane as in all prokaryotes; (iii) the distinction in lipid compositions of between the endomembrane system and plasma membrane, which originated from the fact that bacterial-like lipids replaced the archaeal lipids from the inside in the first place; and that (iv) the ER-continuous nuclear envelope is not homologous to the plasma membrane of putative archaeal host, which is just due to the formation of newly enveloped nucleus from the ER in eukaryotic cells with open mitosis [6]. Recently, functional and proteomic studies have revealed the remarkable complexity of mitochondrial protein machineries with diverse functions, such as protein translocation, oxidative respiration, metabolite transport, protein quality control, and the control of membrane architecture, that also interact with each other in dynamic networks [28]. These protein networks form distinct membrane contact sites, for example, with the ER, which are key for integrity of mitochondria with eukaryotic cellular functions in bioenergetic synthesis, oxidative metabolism, and redox signaling. The ER-mitochondria connections can also, in turn, determine mitochondrial fusionfission dynamics. Furtherly, the mitochondrial dynamics and inheritance during cell division and development are tightly regulated through evolutionary conserved mechanisms [29], such that they are properly segregated and partitioned as a functional set of their organelles to each of daughter cells during cell division, oogenesis, fertilization, and subsequent development, as well as to ensure the integrity of its mtDNA during genetic selection of functional genomes. Contrarily, defects in the processes lead to relevant cell and tissue pathologies, including cancer and degenerative diseases [29,30]. However, whether the putative ER-nuclearmitochondrial (ENUM) communication is involved in mitochondrial regulation remains elusive.
Aside from only 13 mitochondrially encoded per se proteins [9], all other thousands of the host nucleusencoded mitochondrial proteins for its biogenesis, structure, and functions in the cellular respiration, oxidative phosphorylation, energy metabolism, and redox balance are subjected to their temporospatial precision expression [10,28]. Of note, a considerable number of the nucleusencoded mitochondrial proteins are transcriptionally regulated by α-Pal NRF1 and GABP NRF2 [26,27] and also translationally monitored by α-Pal NRF1 -targeted eIF2α [21][22][23]. Particularly, another portion of those nucleus-encoded proteins, that are positioned within and around the outer and inner mitochondrial membranes, are allowed for biosynthesis in proximity to the ribosome-budded ER, because the ER is central to biosynthesis of secretory and 2 Oxidative Medicine and Cellular Longevity membrane proteins, their proper folding and processing into maturation by quality controls [31,32], before transfer to the mitochondria and anchored within double mitochondrial membranes. The intermembrane space between mitochondrial membranes is evolutionarily homologous to the most oxidizing lumen of the ER, with a lowest GSH/GSSG ratio of 1 : 1~3 : 1, than those in the relative reducing mitochondrial matrix and nuclear environments (with a higher GSH/GSSG ratio of~100 : 1) [33][34][35]. The reducing nucleus is segregated by the ER-connected envelope membranes from the relative oxidizing cytoplasm (GSH/GSSH = 30 : 1~50 : 1) [33], in which the mitochondria is a major source of reactive oxygen species (ROS) as byproducts during cellular respiration [36]. However, contribution of the ER-derived redox signaling mechanism to the nuclear-to-mitochondrial respiratory and antioxidant transcription networks remains unclear.
Amongst the putative ER-derived redox signaling machineries, the most critical antioxidant arm is mediated by the membrane-bound nuclear factor erythroid-2 p45related factor 1 (Nrf1, also called Nfe2l1) [37]. The abbreviation is almost identical with nuclear respiratory factor-1 (NRF-1, also called α-Pal NRF1 ), leading to the confused interpretation of relevant works, of which the most typical was published by L'Honore et al. (with no precision correction, as compared with its original version online in Developmental Cell [38]). As a matter of fact, Nfe2l1 (i.e., Nfe2l1 Nrf1 ) is significantly distinctive from α-Pal NRF1 , because the former Nfe2l1 Nrf1 belongs to the family of cap'n'collar (CNC) basic region-lucine zipper (bZIP) transcription factors that regulate distinct subsets of antioxidant response element (ARE, 5 ′ -TGAC/GnnnGC-3 ′ )-driven cytoprotective genes [37]. This conserved CNC-bZIP family also includes Nfe2l2 Nrf2 that is negatively regulated by Kelch-like ECH-associated protein 1 (Keap1), which is an adaptor subunit of Cullin 3based E3 ubiquitin ligase and also a key sensor for oxidative and electrophilic stresses [39]. However, hitherto no literature is available for explaining two distinct, but previously confused, Nrf1 transcription factors, albeit they shared the same abbreviations. Herein, we have identified that both Nfe2l1 Nrf1 and α-Pal NRF1 exert synergistic and antagonistic roles in integral regulation of the nuclear to mitochondrial respiratory and antioxidant transcription profiles. Such synergistic contributions of Nfe2l1 and Nfe2l2 to positive regulation of α-Pal NRF1 and TFAM were validated in Keap1 -/mouse embryonic fibroblasts (MEFs). By contrast, human α-Pal NRF1 expression was almost unaltered by hNfe2l1α -/or hNfe2l2 -/-ΔTA , albeit its target TFAM was activated by Nfe2l1, along with antagonism against Nfe2l2 in HepG2 cells. Conversely, almost all of mouse Nfe2l1, Nfe2l2, and cotarget genes were downexpressed to certain extents in α-Pal NRF1+/-MEFs. Further investigation revealed that human Nfe2l1, Nfe2l2, and relevant reporter genes were downregulated by an ectopic α-Pal NRF1 expression, but this inhibitory effect was reversed by silencing α-Pal NRF1 . Moreover, Pitx2 (pituitary homeobox 2) was identified as a direct upstream regulator of Nfe2l1 and TFAM, besides α-Pal NRF1 . Collectively, these across-talks between Nfe2l1, Nfe2l2, and α-Pal NRF1 , together with Pitx2, are integrated through multiple extranu-clear (e.g., ER-derived) signaling to the nuclear-tomitochondrial communication for their mitochondrial target TFAM, so as to finely tune distinct cellular oxidative respiratory and antioxidant gene transcription networks responsible for maintaining the robust redox homeostasis and organelle integrity, albeit they differ between the mouse and human.

Results
2.1. Knockout of Nfe2l1 Nrf1 in MEFs Leads to Substantial Decreases of Nfe2l2, α-Pal NRF1 , TFAM, and Other Target Genes. Herein, whether knockout of Nfe2l1 has an effect on the constitutive expression of Nfe2l2 and target genes in MEFs was firstly examined by real-time quantitative PCR (RT-qPCR) and Western blotting. As anticipated, loss of Nfe2l1 in mice resulted in substantial decreases in both mRNA and protein levels of Nfe2l2 expressed constitutively in Nfe2l1 -/-MEFs (Figures 1(a)-1(c)). Similar decreases in the basal expression of five of the examined ARE-driven genes HO-1 (heme oxygenase 1, HMOX1), GCLM (glutamate-cysteine ligase modifier subunit), MT-1 (metallothionein 1), GSTp (glutathione S-transferase pi 1), and SOD1 (superoxide dismutase 1) were determined in Nfe2l1 -/-MEFs (Figure 1(a)). By sharp contrast, significant increases in the basal expression of GSTa1 (glutathione S-transferase, alpha 1 (Ya)) and Aldh1a1 (aldehyde dehydrogenase family 1, A1) were also obtained from Nfe2l1 -/in MEFs (Figure 1(a)), although they were identified as Nfe2l2-target genes [40,41]. Western blotting of Nfe2l1 -/-MEFs revealed obvious decreases in protein abundances of HO-1, GCLM, and SOD1, as accompanied by a striking increase in Aldh1a1 protein levels (Figure 1(c)). In addition, it should also be noted that Nfe2l1α and its derivates were completely deleted in Nfe2l1 -/-MEFs, but the residual shorter isoforms ΔN, β, γ, and δ were detected by Western blotting with the antibodies against amino acids 291-741 of Nfe2l1 Nrf1 (Figure 1(b)) [42]. Collectively, these indicate bidirectional contributions of Nfe2l1 to positive and negative regulation of ARE-driven genes by Nfe2l2 or Nfe2l1α.

Knockout of Nfe2l2 Nrf2 in MEFs Results in Substantial
Decreases of α-Pal NRF1 , TFAM, and Other Target Genes. As shown in Figure 1(g), the basal mRNA expression of Nfe2l1 in MEFs was almost unaffected by knockout of Nfe2l2 -/-, albeit all 7 of the examined Nfe2l2-targeted genes HO-1, GCLM, MT-1, SOD1, GSTa1, GSTp, and Aldh1a1 were downregulated by loss of Nfe2l2 to different lower extents, Relative mRNA levels    Oxidative Medicine and Cellular Longevity when compared to their equivalent levels measured in wildtype (Nfe2l2 +/+ ) cells. The latter notion is also corroborated by evidence showing significant decreases in the constitutive protein expression of HO-1, GCLM, SOD1, and Aldh1a1 ( Figure 1(h)). Intriguingly, the full-length Nfe2l1α abundance was only marginally reduced in Nfe2l2 -/-MEFs, but its processed isoforms of protein-B to -D appeared to be largely abolished by Nfe2l2 -/-, which was also accompanied by a rather remarkable increased abundance of Nfe2l1β ( Figure 1(i)). These findings suggest that the putative proteolytic processing of Nfe2l1α to yield distinct isoforms may be influenced by Nfe2l2.
Comparative analysis of Nfe2l2 -/and wild-type MEFs unraveled almost no changes in mRNA expression levels of α-Pal NRF1 and TFAM (Figure 1(j)). However, both protein levels of α-Pal NRF1 and TFAM were markedly downregulated by loss of Nfe2l2 (Figure 1(k)), implying that the stabilization of α-Pal NRF1 and TFAM may be monitored in the presence of Nfe2l2 (Figure 1(l)). Intriguingly, the mRNA expression of COX5a, Ndufv1, and Ndufb6, besides SOD1, were downregulated to varying degrees in Nfe2l2 -/-MEFs (Figure 1(j)). This suggests that Nfe2l2 is likely required for controlling the basal expression of these three mitochondrial respiratory genes regulated through an α-Pal NRF1 -independent mechanism (Figure 1(l)).

6
Oxidative Medicine and Cellular Longevity in Nfe2l2 -/-MEFs (Figure 1(h)). Thereby, it is inferable that both GSTp and MT-1 genes (and possibly Nfe2l2 per se) are also likely subjected to putative dominant negative regulation by, at least, a not-yet-identified transcription factor competitively against the positive effects of Nfe2l2 and Nfe2l1 in Keap1 -/-MEFs. By contrast, further examination of Keap1 -/-MEFs uncovered remarkable increases of both α-Pal NRF1 and TFAM in their mRNA and protein expression levels ( Figure 2(d) and 2(e)). This finding demonstrates upregulation of both α-Pal NRF1 and TFAM by Keap1 -/--led coactivation of Nfe2l2 and Nfe2l1. However, Ndufv1 and Ndufb6 (both encoding two distinct subunits of the mitochondrial respiratory chain complex Ι) were only slightly promoted, whilst the expression of COX5a was roughly unaltered, in Keap1 -/-MEFs (Figure 2(d)). This result implies a putative negative transcription factor competitively against the positive regulatory effects of α-Pal NRF1 on the transcriptional expression of its three mitochondrially-targeted genes Ndufv1, Ndufb6, and COX5a encoded in the nucleus of Keap1 -/-MEFs ( Figure 2(f)).

Induction of ARE-Driven α-Pal NRF1 -Luciferase Reporter
Genes Mediated by Ectopic Nfe2l1 and/or Nfe2l2. To further determine the transcriptional regulation of α-Pal NRF1 by Nfe2l1 and Nfe2l2, we herein created 6 pairs of wild-type luciferase reporters driven by the indicated ARE consensus sites from mouse α-Pal NRF1 promoter and their corresponding mutants (as shown in Figure 2(g)). Amongst them, the basal activity of ARE2, ARE3, and ARE5 was obviously prevented by their respective mutants (Figure 2(h)). This implies putative regulation of these three reporters by endogenous ARE-binding factors Nfe2l1 and/or Nfe2l2 alone or both together. Further cotransfection of RL34 cells with an Nfe2l1 Nrf1 expression construct revealed that the transcriptional expression of all other 5 ARE-, except ARE4-, driven luciferase reporters was activated by Nfe2l1 ( Figure 2(h)). Such Nfe2l1-mediated activity of only ARE5 and ARE6 situated within the first exon was suppressed by their mutants, but other three ARE1, ARE2, and ARE3 activity mediated by Nfe2l1 was largely unaffected by their mutants (Figure 2(h)). By contrast, transcriptional activity of ARE1-, ARE2-, and ARE3-luc reporters was not only activated by Nfe2l2 but also reduced by their respective mutants to relative lower degrees. Together, these indicate that ARE1, ARE2, and ARE3 are Nfe2l2-specific regulatory sites within the mouse α-Pal NRF1 promoter region, whilst ARE5 and ARE6 are Nfe2l1-specific regulatory sites within the 5′-noncoding exon one-encoded region of α-Pal NRF1 . In addition, ARE1-, ARE2-, and ARE3-mutant reporters were also activated by Nfe2l1 but not by Nfe2l2. This suggests a possibility that the three ARE mutants from the α-Pal NRF1 promoter region could also serve as certain artificial consensus sites recruiting the putative Nfe2l1-downstream transcription factors. Thus, it is inferable that upregulation of ARE1, ARE2, and ARE3 activity to mediate the α-Pal NRF expression may also be mediated by Nfe2l1 directly and/or indirectly through its target downstream factors.

No
Alterations of α-Pal NRF1 in hNfe2l1α -/or hNfe2l2 -/-ΔTA Are Accompanied by Opposing Changes of TFAM. Next, we examined the inter-regulatory effects of between human Nfe2l1 and Nfe2l2 on α-Pal NRF1 , TFAM, and other target genes. Unexpectedly, almost no changes in protein and mRNA expression levels of α-Pal NRF1 were determined in either hNfe2l1α -/or hNfe2l2 -/-ΔTA cells (both lines were established from human HepG2 cells by Qiu et al. [44]), when compared with those equivalents obtained from wild-type cells (Figures 4 A6 and 4(b)). Accordingly, no or less alterations in the basal expression of Ndufv1, Ndufb6, and COX5a by hNfe2l1α -/or hNfe2l2 -/-ΔTA were observed (Figure 4(b)). However, it is intriguing that α-Pal NRF1 -target TFAM at its protein and mRNA levels was strikingly downregulated in hNfe2l1α -/cells (albeit with hyper-expression of Nfe2l2), but rather significantly upregulated in hNfe2l2 -/-ΔTA cells, by comparison with their wild-type controls (Figures 4 A7 and 4(b)). These observations were further substantiated by transcriptomic sequencing (Figure 4(c)). Collectively, these results indicate that both Nfe2l1 and Nfe2l2 contributes, respectively, to the putative positive and negative regulation of the human TFAM expression, possibly through a mechanism 7 Oxidative Medicine and Cellular Longevity  15 12 Relative mRNA levels

Keap1
(f) Figure 2: Continued. 8 Oxidative Medicine and Cellular Longevity independent of α-Pal NRF1 . This notion is further supported by the evidence showing that a substantial increase in protein abundances of TFAM, rather than α-Pal NRF1 , was resulted from overexpression of ectopic Nfe2l1 (i.e., Nrf1-V5, as shown in Figure 4, cf. D3 with D4). In addition, the inter-regulatory effects of between human Nfe2l1 and Nfe2l2 on the ARE-driven genes were also validated, as consistent with our previous report [44]. Of note, knockout of hNfe2l1α -/caused a marked increase in human Nfe2l2 protein abundances but not in its mRNA expression (Figures 4 A2, 4(b), and 4(c)). In turn, hNfe2l2 -/-ΔTA cells gave rise to an obvious reduction of post-synthetically processed Nfe2l1 isoforms and its mRNA levels (Figures 4 A1, 4(b), and 4(c)).
Western blotting of hNfe2l1α -/revealed that massive increases of both GCLM and HO-1 proteins were coincided with hyperactive Nfe2l2 (Figure 4, A3 and A4). RT-qPCR analysis unraveled that the basal mRNA

Oxidative Medicine and Cellular Longevity
Nfe2l2 Nfe2l2    Oxidative Medicine and Cellular Longevity 2.8. Another Contribution of α-Pal NRF1 to Transrepression of Human Nfe2l1, Nfe2l2, and ARE-Driven Genes. To clarify putative contributions of α-Pal NRF1 to the transcriptional expression of Nfe2l1, we herein made two luciferase reporters driven by mouse and human Nfe2l1 gene promoter regions, called mNfe2l1-luc and hNfe2l1-luc, respectively. As unexpected, ectopic α-Pal NRF1 -V5 overexpression only caused a slight reduction in activity of mNfe2l1-luc reporter ( Figure 5(a)), albeit this reporter gene was significantly induced by tBHQ ( Figure 5(b), b1). Similarly, activity of hNfe2l1-luc reporter was substantially activated by tBHQ ( Figure 5 k)). In addition, GCLM protein and mRNA levels were almost unaffected by silencing of α-Pal NRF1 (Figure 5(l) and M7). Together, these results indicate that α-Pal NRF1 contributes to negative regulation of human Nfe2l1, Nfe2l2, and HO-1 ( Figure 5(n)). However, no dose responses of Nfe2l1, Nfe2l2, and HO-1 to different concentrations of α-Pal NRF1 -silenced siRNAs indicate a possible involvement of other transcription factors (e.g., Pitx2 and GABP NRF2 ). Moreover, α-Pal NRF1 also exerts disparate effects on GCLM, possibly depending on whether Nfe2l1, Nfe2l2, and other unidentified transcription factors were also stimulated by α-Pal NRF1 overexpression or its RNA-silencing. 13 Oxidative Medicine and Cellular Longevity [38]. Here, we determine whether Pitx2 also acts as a direct upstream factor to mediate transactivation of human Nfe2l1. As shown in Figure 6(a), activity of mNfe2l1-luc reporter was significantly transactivated by the ectopic Pitx2 expression, when compared with the basal activity measured from its cotransfection with an empty pGL3-Basic vector. Similar results were also obtained from those hNfe2l1-luc reporters (Figures 6(b) and 6(c)). Of note, overexpression of ectopic Pitx2 caused gradual increments in the transactivity of hNfe2l1-luc, hNfe2l1-luc1, and hNfe2l1-luc2 reporters ( Figure 6(c), C2-C4). The hNfe2l1-luc activity was also strikingly induced by tBHQ to the extent of hNfe2l1-luc1 transactivation mediated by Pitx2 (Figure 6(c), cf. C1 with C3). In addition, four distinct consensus Pitx-response elements (PitxREs), which are situated at the upstream of the indicated promoter region to have yielded hNfe2l1-luc1 (Figure 6(b)), were also identified herein. The results revealed that only PitxRE2-luc activity was modestly activated by Pitx2, but the basal activity of PitxRE3-luc and PitxRE4-luc was partially suppressed by Pitx2 (Figure 6(d)). Together, these indicate that the transcriptional expression of human Nfe2l1 gene is likely coregulated by Pitx2 and its target factors (e.g., α-Pal NRF1 ).
RT-qPCR analysis revealed that the endogenous mRNA expression of human Nfe2l1 and HO-1 was increased by Pitx2 overexpression (Figure 6(e)). Western blotting of Pitx2-expressing cells unraveled significant increases in abundances of all 7 examined proteins Nfe2l1, Nfe2l2, HO-1, GCLM, α-Pal NRF1 , TFAM, and SOD1 ( Figure 6(f), F2-F8). These demonstrate that Pitx2 may also serve as an upstream regulator to mediate transcriptional expression of Nfe2l1 (and Nfe2l2), besides α-Pal NRF1 . This notion is further supported by the evidence obtained from silencing of Pitx2 by its three distinct siRNA sequences (as listed in Table S1). Consequently, silencing of Pitx2 led to varying decreases in basal mRNA expression levels of α-Pal NRF1 , TFAM, Nfe2l1, HO-1, GCLM, COX5a, and SOD1 to less extents than their equivalent controls (Figures 6(g) to 6(n)). Thereby, these results unveil that Pitx2 is required for mediating transcriptional expression of Nfe2l1, α-Pal NRF1 , and their target genes (Figure 6(o)).

Identification of TFAM as another Direct Cotarget of
Human Nfe2l1, Nfe2l2, and Pitx2. Although TFAM is known as a direct target of α-Pal NRF1 [18,45,46], we further determined whether it also serves as another direct cotarget of Nfe2l1, Nfe2l2, and/or Pitx2. As illustrated in Figure 7(a) (upper panel), two similar α-helical structures were wheeled, respectively, by the successive mitochondria-targeting sequences (MTS1 and MTS2) in each nuclearly encoded TFAM. Alignment of three amino acid sequences from human and mouse showed that MTS1 (aa 1-18) and MTS2 (aa 21-38) are located within the N-terminal region of this MHG-(high mobility group-) box-binding transcription factor TFAM (Figure 7(a), lower panel). Further bioinformatic analysis revealed several putative consensus cis-regulatory elements, such as AREs, PitxRE, and α-Pal-binding site within the promoter region of human TFAM (Figure 7(b)). Amongst them, ARE-and PitxRE-driven luciferase reporters were established herein (Figure 7(c)). As a result, the luciferase assays of both PitxRE-Luc activity (Figure 7(d)) and ARE3-Luc activity (Figure 7(e)) demonstrated that the transcriptional expression of TFAM is, de facto, coregulated by human Pitx2, Nfe2l1, and Nfe2l2 (Figures 7(d)-7(f)). Together with the previous work by L'Honore et al. [38], these demonstrate that TFAM is a bona fide cotarget gene of Nfe2l1, Nfe2l2, and α-Pal NRF1 , as well as their upstream regulator Pitx2 (Figure 7(f)).
2.11. Distinct Contributions of Human Nfe2l1 and Nfe2l2 to Differential Expression of TFAM, PGC-1α/β, and Associated Genes That Are Involved in the Nuclear-to-Mitochondrial Communication. Since TFAM is of crucial importance in the nuclear control of the mitochondrial function, it is also postulated to be involved in redox regulation by Nfe2l1 and Nfe2l2. As anticipated, Figure 8(a) showed that TFAM was upregulated by Nfe2l1, but downregulated by Nfe2l2, in their respective tetracycline-inducible cell systems (as described by Wang et al. [47]), albeit both CNC-bZIP factors positively contributed to upregulation of TFB1M and TFB2M. These were corroborated by further evidence obtained from Nfe2l1α -/-, Nfe2l1α -/-+siNfe2l2, and Nfe2l2 -/-ΔTA cell lines (Figure 8(b)).
An overview of multilayer interaction networks with the aforementioned key molecular nodes is illustrated in Figures 9(a)-9(j), to provide an explicit understanding of the nuclear-to-mitochondrial communication by multiple extranuclear signaling directed towards the nuclear controls of the mitochondrial respiration, energy metabolism, and redox regulation. Such hierarchical network consists of several subnetworks, with each of important transcription factors serving a hub in the center. Of note, the mitochondrial genome is transcriptionally controlled by TFAM, TFB1M, and TFB2M (Figures 9(h) and 9(j)), although they are nucleus-encoded, and also controlled by α-Pal NRF1 , GAB-Pα NRF2 , Nfe2l1 Nrf1 , and/or Nfe2l1 Nrf2 (c-g). Those differentially expressed genes (DEGs) as key players in this network were further analyzed by RNA-sequencing of Nfe2l1αinduced cells (Figures 9(k) to9(m)) and Nfe2l2-induced cells (n-p), when compared with their WT controls. 14 Oxidative Medicine and Cellular Longevity
Since PGC-1α/β and PRC1 are accepted as key transcription cofactors involved in the nuclear control of mitochondrial biogenesis and function, here we established an interaction network, with distinct families of 21 transcriptional factors and 6 cofactors, by using the STRING tool ( Figure 10(a)). Differential expression levels of these genes possibly affected by Nfe2l1 or Nfe2l2 were determined by comparison of the RNA-sequencing data of either Nfe2l1αor Nfe2l2-induced cells versus WT controls (Figure 10(b)) and also of Nfe2l1α -/-, Nfe2l1α -/-+siNfe2l2, or Nfe2l2 -/-ΔTA cell lines versus another WT controls (Figure 10(c) Figure 5: A negative effect of α-Pal NRF1 on Nfe2l1, Nfe2l2, and other antioxidant genes in HepG2 cells. (a) HepG2 cells were cotransfected for 8 h with mNfe2l1-luc reporter and the pRL-TK control, along with an expression construct for mouse α-pal NRF1 or an empty pcDNA3.1, and then allowed for a 24 h recovery before the luciferase activity was measured (A1). Total cell lysates were also subjected to identification by Western blotting with distinct antibodies against α-Pal NRF1 (A2) or V5 tag (A3). (b) HepG2 cells were cotransfected for 8 h with mNfe2l1luc (B1) or hNfe2l1-luc (B2), plus pRL-TK, and then treated with 50 μmol/L tBHQ or the DMSO vehicle for 24 h, before being allowed for additional 24 h recovery. Subsequently, these samples were subjected to dual luciferase assays. The results are shown as mean ± SEM (n = 3 × 3) with significant increases ($, p < 0:01). (c) HepG2 cells were cotransfected for 8 h with hNfe2l1-luc (C1) or ARE×6-luc (C2) together with pRL-TK, plus a human α-Pal NRF1 expression construct or an empty pcDNA3.1 and then allowed for 24 h recovery from cotransfection, before the reporter activity was measured. The data are shown as mean ± SEM (n = 3 × 3) with significant decreases ( * p < 0:01). Total cell lysates were also subjected to characterization by Western blotting with distinct antibodies against α-Pal NRF1 (C3) or V5 tag (C4). (d) Distinct protein levels of α-Pal NRF1 , TFAM, SOD1, Nfe2l1, Nfe2l2, HO-1, and GCLM were determined by Western blotting of HepG2 cells that had been transfected with α-Pal NRF1 expression plasmid or an empty pcDNA3 vector. (e-l) Distinct changes in mRNA levels of α-Pal NRF1 (e), TFAM (f), SOD1 (g), COX5a (h), Nfe2l1 (i), Nfe2l2 (j), HO-1 (k), and GCLM (l) were analyzed by realtime qPCR of HepG2 cells that had been transfected with 0, 25, 50, and 100 nM of siRNA against α-Pal NRF1 . The resulting data are shown as mean ± SEM (n = 3 × 3) with significant decreases ( * p < 0:01, * * p < 0:001) or increases ($, p < 0:01). (m) Such siRNA-transfected cell lysates were subjected to Western blotting analysis of distinct protein abundances as indicated. (n) A model is proposed to present disparate effects of α-pal NRF1 overexpression or its knockdown on the nuclear-to-mitochondrial respiratory and antioxidant genes in HepG2 cells. 17 Oxidative Medicine and Cellular Longevity

Discussion
In the present study, we have, for the first time, provided a better interpretation of two totally distinctive Nrf1 transcription factors (i.e., Nfe2l1 Nrf1 and α-Pal NRF1 , Figures 11(a) and 11(b)). This is an attempt to rectify the previously confusing explanations of both Nrf1 factors, from their commonly shared abbreviations in the literature as far as we know, in which one of the most typical representative is the publication confused by L'Honore et al. [38]. Similarly, another two similar Nrf2 transcription factors [48,49] are also referred to as Nfe2l2 Nrf2 and GABPα NRF2 , respectively. Of note, the antioxidant master Nfe2l2 Nrf2 is highly conserved with Nfe2l1 Nrf1 , but neither have evolutionary conservation with those nuclear respiratory factors α-Pal NRF1 and GAB-Pα NRF2 , of which the latter two factors have also no any homology with each other (Figure 11(c)). Collectively, our present study has clearly elucidated both synergistic and antagonistic roles of Nfe2l1 Nrf1 , Nfe2l2 Nrf2 , and α-Pal NRF1 in integrative regulation of multiple signaling to the nuclear-to-mitochondrial respiratory and antioxidant transcription networks, which differ between the mouse and human, though (also see Figure S2).
3.1. Synergism of Mouse Nfe2l1 Nrf1 , Nfe2l2 Nrf2 , and α-Pal NRF1 to Coordinate the Nuclear-to-Mitochondrial Respiratory and Antioxidant Gene Responses in MEFs. Clearly, Nfe2l1 and Nfe2l2 are two highly conserved members of the CNC-bZIP transcription factors, which predominantly regulate distinct subsets of ARE-driven cytoprotective genes against cellular oxidative stress [37,39]. These target genes are critically involved in maintaining cellular homeostasis and organ integrity during normal development and growth, as well as healthy life process, in which Nfe2l1 and Nfe2l2 cannot only exert their combinational and/or overlapping roles, but also manifest their distinctive functions from each other. Importantly, Nfe2l1 elicits its unique indispensable physiopathological functions, because its loss of function in the mouse leads to distinct pathological phenotypes, even in the presence of Nfe2l2 [37]. This implies a fact that the loss of Nfe2l1 cannot be compensated by Nfe2l2, albeit Nfe2l2 has been accepted as a master regulator of antioxidant, detoxification, and cytoprotective responses [39,50]. Such significant distinctions in the functioning of between Nfe2l1 and Nfe2l2 are dictated by their respective intrinsic features.
Interestingly, Nfe2l1, but not Nrf2, with several domains being more highly conserved with its ancestral CNC, Skn-1, and Nach proteins [51], is located in the ER and its connected nuclear envelope membranes. Of note, there exist two extra functional domains (i.e., NTD and NST) within Nfe2l1, rather than Nfe2l2. The ER-targeting NTD of Nfe2l1 enables it to be topologically anchored within and around the membranes [52,53], whilst its NST glycodomain facilitates its proper protein folding and subsequent processing [42,54]. The membrane-topology of Nfe2l1 determines the vectorial selective proteolytic processing of this CNC-bZIP protein by proteasomes and/or other cytosolic proteases in close proximity to the ER, in order to yield distinct lengths of its

22
Oxidative Medicine and Cellular Longevity isoforms, one of which is mature CNC-bZIP factor translocated into the nucleus [42,53]. Such the membrane-bound Nfe2l1 factor can also be activated in the unfolded protein response to ER stress stimulated by tunicamycin (to inhibit its N-linked glycosylation) [55], in addition to its proteasomal "bounce-back" response [42,56]. By sharp contrast, the water-soluble Nfe2l2 is segregated by Keap1 within the cytoplamic compartments, where it is targeted for its ubiquitination-mediated proteasomal degradation [39]. Rather, upon stimulation of Nfe2l2 by oxidative stress, this CNC-bZIP factor is dissociated from the cytoplamic Keap1, so that it is subsequently translocated into the nucleus and transactivates ARE-driven target genes. Collectively, these distinctions of Nfe2l1 from Nfe2l2 demonstrate disparate selection of their tempospatial activation or inactivation insomuch as to finely tune distinct subsets of target genes. As such, this also presages synergism and antagonism of between Nfe2l1 and Nfe2l2 in regulating distinct cognate gene expression. In fact, mouse Nfe2l1 -/embryonic lethality occurred at middle to late gestation starting at E13.5 [57], whereas double knockout of Nfe2l1 -/-:Nfe2l2 -/caused an earlier death of the resulting mouse embryos at E10.5 due to extensively increased apoptosis and growth retardation induced by  TFAM  TFB1M  TFB2M  TIMM17A  TP53  ALAS1 ATP5B C10orf2 Fold changes (log severe endogenous oxidative stress [58]. The elevated ROS levels were resulted from severely impaired expression of antioxidant defense genes (e.g., Mt-1, Gclm, Gclc, Ferritin H, Ho-1, and Nqo1) in Nfe2l1 -/-:Nfe2l2 -/-MEFs, when compared with their individual knockout of Nfe2l1 -/or Nfe2l2 -/-. Hence, it is inferred that Nrf2 is also allowed for a partial  Nfe2l2-induced

25
Oxidative Medicine and Cellular Longevity compensation for the loss of Nrf1's function in synergistically regulating critical genes for the intracellular redox homeostatic setting at the robust threshold state during embryogenesis. Such partially overlapping functions of both Nfe2l1 and Nfe2l2 are determined by coexpression patterns of the two CNC-bZIP factors [59][60][61] and their similarity of conserved sequences [37], which are also, though, driven by their respective ARE-containing gene promoters (Table S2).
It is, to our surprise, that the evidence presented herein reveals that knockout of Nfe2l1 in mice leads to significant decreases of both mRNA and protein levels of Nfe2l2 expressed in Nfe2l1 -/-MEFs. By contrast, almost no changes    (a) A model is based on our experimental evidence, to give a better understanding of regulatory cross-talk among Nfe2l1 Nrf1 , Nfe2l2 Nrf2 , α-Pal NRF1 , and Pitx2, all targeting TFAM. During the nuclear-to-mitochondrial communication, Nfe2l1 Nrf1 makes opposing contributions to bidirectional regulation of Nfe2l2 Nrf2 and α-Pal NRF1 by itself and target proteasome (PSM) at two distinct layers. Nfe2l2 Nrf2 can determine posttranscriptional regulation of Nfe2l1 Nrf1 and α-Pal NRF1 , but the detailed mechanism remains unclear. In turn, mouse α-Pal NRF1 contributes to positive regulation of Nfe2l1 Nrf1 and Nfe2l2 Nrf2 , although no canonic GC-rich α-Pal-binding sites exist in these mouse CNC-bZIP gene promoter regions (Table S2). Contrarily, human α-Pal NRF1 makes a negative contribution to transcriptional expression of Nfe2l1 Nrf1 and Nfe2l2 Nrf2 (the latter Nfe2l2 Nrf2 is dominantly negatively regulated by human Nfe2l1 Nrf1 ), albeit all three factors are activated by redox inducer tBHQ. In addition to negative regulation of Nfe2l2 Nrf2 by Keap1, the adaptor subunit of Cullin 3-based E3 ubiquitin ligase can also make a positive contribution to transcriptional regulation of mouse Nfe2l2 Nrf2 , rather than Nfe2l1 Nrf1 , as found in MEFs. However, human Nfe2l1 Nrf1 is essential for stabilization of Keap1, but it is unknown whether this adaptor protein is involved in the proteolytic processing of Nfe2l1 Nrf1 . Notably, the nucleus-controlled mitochondrial respiratory and oxidative phosphorylation are also a primary source of ROS in cells, which triggers activation of Nfe2l1 Nrf1 , Nfe2l2 Nrf2 , and α-Pal NRF1 to certain extents so that cellular redox homeostasis is maintained at a steady state. Besides, GABP NRF2 is also required for this process, but possible cross-talks of this ETS family factor with Nfe2l1 Nrf1 and Nfe2l2 Nrf2 are not yet identified here. The "M"-marked arrowheads indicate those activity in the mouse but not in the human; such distinction was also shown (in Figure S2) 27 Oxidative Medicine and Cellular Longevity in mRNA expression levels of Nfe2l1 in Nfe2l2 -/-MEFs, but the loss of Nfe2l2 results in obvious decreases in abundances of the full-length Nfe2l1α and its processed proteins (i.e., isoform-B, C, and D), although this was accompanied by an increased Nfe2l1β. Collectively, these results demonstrate that Nfe2l1 acts as a dominant positive regulator to monitor the basal constitutive expression of Nfe2l2, at least in MEFs ( Figure S2A). In turn, the latter Nfe2l2 does not control transcription of mouse Nfe2l1 gene, but it is required for stabilization of Nfe2l1α and its proteolytic processing to yield a mature CNC-bZIP factor. This notion is further substantiated by the evidence obtained from Keap1 -/-MEFs, showing that accumulation of Nfe2l2 is accompanied by significant increases in abundances of Nfe2l1α and its derivates, but basal mRNA expression of Nfe2l1 remains to be unaffected, whilst Nfe2l2 mRNA levels were decreased significantly.
Recently, emerging evidence showed that the regulatory cross-talks of Nfe2l2 (but Nfe2l1 was not shown) with either α-Pal NRF1 or GABPα NRF2 are involved in the nuclear control of mitochondrial biogenesis and relevant biological functions [48,49,62]. In this study, we have unravelled that knockout of Nfe2l1 causes significant decreases in mRNA and/or protein levels of α-Pal NRF1 and its target genes TFAM, Ndufv1, Ndub6, COX5a, and SOD1 in Nfe2l1 -/-MEFs, whereas knockout of Nfe2l2 -/also leads to marked decreases in protein abundances of α-Pal NRF1 and TFAM, but almost unaltered mRNA levels of both α-Pal NRF1 and TFAM were observed in Nfe2l2 -/-MEFs. These results demonstrate that Nfe2l1 acts as a dominant regulator to monitor transcriptional expression of α-Pal NRF1 , TFAM, and other target genes in MEFs, whilst stability of both α-Pal NRF1 and TFAM proteins may also be monitored by a Nfe2l2-mediated mechanism. This notion is further supported by another evidence showing that Keap1 -/-MEFs gave rise to evident increases in basal abundances of Nfe2l2, Nfe2l1α, and its processed isoforms, which were accompanied by obvious increased proteins of both α-Pal NRF1 and TFAM. However, only modest increases in their mRNA expression levels of α-Pal NRF1 and TFAM were examined in Keap1 -/-MEFs, with an exception of significantly reduced mRNA levels of Nfe2l2, but not of Nfe2l1. This exception implies that the transcriptional expression of Nfe2l2 requires for another transcription factor involving a putative positive feedback mechanism regulated by Keap1, albeit this Nfe2l2 inhibitor negatively regulates stability of this CNC-bZIP protein. Moreover, our further luciferase assays also revealed that both Nfe2l1 and Nfe2l2 contribute to transcriptional expression of distinct ARE-driven α-Pal NRF1 reporters. Conversely, marked decreases in mRNA and protein levels of Nfe2l1 and Nfe2l2 resulted from heterogeneous deletion of mouse α-Pal NRF1+/-. Overall, these indicate bidirectional cross-talks of Nfe2l1 and/or Nfe2l2 with α-Pal NRF1 at distinct layers to coordinate the extranuclear signaling directed towards the nuclear-to-mitochondrial respiratory and antioxidant transcription networks.
Intriguingly, we found almost no effects of hNfe2l1 -/or hNfe2l2 -/on mRNA and protein levels of human α-Pal NRF1 in this experimental setting, albeit both CNC-ZIP genes contain at least one consensus α-Pal binding site within their promoters (Table S3). This is also further confirmed by transcriptomics sequencing of hNfe2l1 -/-+siNfe2l2 cells, as compared with those obtained from either hNfe2l1 -/or hNfe2l2 -/cells. Such results are really contrary to those obtained from mouse Nfe2l1 -/and Nfe2l2 -/as described above. Conversely, significantly decreases of human Nfe2l1 and Nfe2l2 protein levels were caused by forced expression of ectopic α-Pal NRF1 . Similarly, transcriptional activity of hNfe2l1-luc, rather than mNfe2l1-luc, reporters was also markedly diminished by ectopic α-Pal NRF1 . By contrast, a recovery of human Nfe2l1 and Nfe2l2 from inhibition of α-Pal NRF1 was also acquired after this respiratory factor was silenced. Notably, silencing of α-Pal NRF1 can only cause a modest decrease in its target TFAM expression, implying that this mitochondrial transcription factor is also monitored by other factors (e.g., GABPα NRF2 and/or Pitx2) beyond α-Pal NRF1 . In this study, we also found that TFAM was downregulated in hNfe2l1 -/cells (albeit with accumulation of Nfe2l2) and hence upregulated in hNfe2l2 -/cells. Our further evidence reveals that TFAM along with α-Pal NRF1 is significantly induced by the redox inducer tBHQ, but such inducible expression levels of both factors are almost completely abolished by hNfe2l1 -/-. Collectively, these indicate that Nfe2l1 and Nfe2l2 contribute, respectively, to the putative positive and negative regulation of TFAM, even though both CNC-ZIP factors are inhibited by α-Pal NRF1 through an as-yet-unidentified mechanism (Figures 11(a) and S2B). Consistently, the positive regulation of TFAM by Nfe2l1 is further corroborated by its consensus ARE3driven report assays. However, it should also be noted that the opposing effects of Nfe2l2 on the endogenous TFAM and its consensus ARE3-luc reporter could depend on distinct ARE3-adjoining promoter contexts of the genome backgrounds.
Importantly, we have presented the experimental evidence that activity of hNfe2l1-luc reporter is transactivated by Pitx2 (albeit it serves as a direct upstream regulator of α-Pal NRF1 [38]). Further evidence also reveals that endogenous expression of Nfe2l1 (and Nfe2l2) was significantly upregulated by overexpression of Pitx2, but rather downregulated by silencing of this homeobox factor. Such Pitx2-directed alternations of Nfe2l1 and Nfe2l2 are accompanied by corresponding changes of HO-1 and GCLM, as well as α-Pal NRF1 , TFAM, and COX5a. In addition to human Nfe2l1, TFAM is another potential target of Pitx2, because its consensus PitxRE-luc reporter was transactivated by this homeobox factor. These, together with our previously reported evidence [44,55], demonstrate multiple regulatory across-talks between Nfe2l1, Nfe2l2, α-Pal NRF1 and Pitx2, and their cotarget TFAM, are integrated by multiple extranuclear (e.g., ERdriven) signaling to the nuclear-to-mitochondrial controls of distinct cellular respiratory and antioxidant transcription networks, albeit they differ between the mouse and human ( Figure S2B). 28 Oxidative Medicine and Cellular Longevity However, the nucleus-controlled mitochondrial respiratory and oxidative phosphorylation are also a primary source of endogenous ROS byproducts in eukaryotic cells. This facilitates to trigger physiological activation of Nfe2l1, Nfe2l2, and α-Pal NRF1 to certain extents in so much as that the intracellular redox homeostasis is maintained at a robust steady-state. Altogether, interaction between these transcription factors comprises a complex feedback circuit with hierarchical regulatory networks served to maintain robust redox homeostasis by balancing a distinct cellular oxidative respiratory system and an antioxidant cytoprotective response.

Conclusion
Generally, most biological functions and physio(patho)logical responses at different levels are determined by a range of regulatory mechanisms from various signaling pathways toward multiple transcription factor-mediated gene networks. Such being the case, these transcription factors and other regulatory molecules could be simplified as key modules which cross-talk amongst them, in order to form a hierarchical network. Nfe2l1 and Nfe2l2 are embedded in such a complex molecular interaction network, responsible for antioxidant, detoxification, and cytoprotective adaptation to distinct physio-pathological stresses during life process. Of note, Nfe2l1 is endowed with unique indispensable biological functions. This fact confers Nfe2l1 to be distinguished from Nfe2l2 at regulating distinct subsets of ARE-driven cognate genes (e.g., Aldh1a1 and GSTa1). But, the inter-regulatory synergism of Nfe2l1 and Nfe2l2 confers both factors to exert their overlapping roles for cotarget genes (e.g., MT-1, GSTp, and SOD1) in MEFs. By sharp contrast, apparent synergistic and antagonistic relationships of between human Nfe2l1 and Nfe2l2 are also demonstrated. Nfe2l1 is a dominant repressor, whilst Nfe2l2 is thus negatively regulated by Nfe2l1 in HepG2 cells. Such function of Nfe2l1 is considered as a brake to avoid Nfe2l2 overshoot. For this simplification to be valid, there should always be a correlation between Nfe2l1 and Nfe2l2 so that both CNC-bZIP factors along with cognate target genes are finely tuned in order to meet the changing needs of cells under all conditions. In fact, Nfe2l1 and Nfe2l2 cannot be reduced as a simple module factor alone ( Figure 9). Now, the question becomes what is the minimal modular network to reproduce the observed data and why it is so evolved. At least three of them α-Pal NRF1 , TFAM, and Pitx2 should also be included, in addition to both Nfe2l1 and Nfe2l2. Overall, this study also redefines that the nuclear controls of mitochondrial respiration and biogenesis, as well as certain protein synthesis and degradation, are physiologically integrated with multiple extranuclear redox signaling to the antioxidant cytoprotective responses mediated by Nfe2l1 and/or Nfe2l2 for maintaining a robust steady-state of redox homeostasis. Lastly, in view of such mutual interregulation of between Nfe2l1 and Nfe2l2 in different experimental cell lines, we should have to take severe cautions to interpret the relevant experimental results obtained from loss or gain of Nfe2l1, Nfe2l2 alone, or both.

Materials and Methods
5.1. Chemicals, Antibodies, and Other Reagents. All chemicals were of the highest quality commercially available. The tertbutylhydroquinone (tBHQ) was from Sangon Biotech (Shanghai, China). Specific antibodies against Nfe2l1 were made in our own laboratory [42], whilst other antibodies against α-Pal NRF1 or V5 ectope were from Abcam and Invitrogen, respectively. Besides, both β-actin and secondary antibodies were from ZSGB-BIO (Beijing, China). Of note, the detailed information about other antibodies, as well as all other key reagents and resources used in this study, was all shown in Table S1.

Expression Constructs, Reporter Plasmids, and Other
Oligos Used for sgRNA or siRNA. Besides four expression constructs for human and mouse Nfe2l1 or Nfe2l2 saved by our group [42,53], another four expression constructs for human and mouse Pitx2, and α-pal NRF1 , were here made by cloning each of those full-length cDNA sequences into the pcDNA3.1 vector. Particularly, the CRISPR/Cas9 plasmid constructs containing guide RNAs specifically targeting mouse α-pal NRF1 were created to generate a heterogeneous (α-pal NRF1+/-) knockout cell line from MEFs. Such sgRNAs (listed in Table S1) were designed with highly specific targets for precision genomic positions (as shown in Figure S1), before being employed in the sgRNA-directed gene-editing of α-pal NRF1 .
Furthermore, several specific cis-regulatory luciferase reporter plasmids were prepared from cloning the indicated gene promoter regions. The mouse Nfe2l1 promoter region (−2047 to +1210) was amplified by PCR from its genomic loci and then inserted into the pGL3-Basic vector. Another group of the cis-regulatory consensus, e.g., ARE and PitxRE-(Pitx2 response elements-) adjoining sequences were cloned from the indicated gene promoter regions of, such as Nfe2l1, TFAM, and α-pal NRF1 and then inserted into the pGL3-Promoter vector. In addition to those intact reporter genes, such as PitxRE-Luc and ARE-Luc, their relevant pointmutant reporters were also engineered herein. The fidelity of the above-described constructs was all confirmed to be true by sequencing. All these primers in the above gene manufacture, and other oligos for siRNA-mediated knockdown of the indicated genes, were listed in Table S1, which all were synthesized by Tsingke (Chengdu, China).

Cell
Lines, Culture, and Transfection. Wild-type mouse embryonic fibroblasts (MEFs) were given as a gift from Akira Kobayashi. Their relevant knockout MEF lines, e.g., Nfe2l1 -/-, Nfe2l2 -/-, and Keap1 -/-, were also obtained from the groups of Profs. Kobayashi and Hayes. Of note, all of these cell lines were originally prepared from Prof. Yamamoto's laboratory. But, another heterogeneous knockout line of α-pal NRF1+/in the mouse was established by its gene-editing in our own laboratory as described above. Besides, the human wild-type (i.e., hNfe2l1/2 +/+ ) hepatocellular carcinoma (HepG2) cells were originally from the American Type Culture Collection (ATCC, Manassas, VA, USA). The fidelity was conformed to be true by its authentication profiling and STR (short tandem repeat) typing map (by Shanghai Biowing Applied Biotechnology Co., Ltd) [55]. On this base, both hNfe2l1α -/and hNfe2l2 -/-ΔTA were previously established and characterized in our own laboratory [44]. These experimental cell lines were maintained for growth in Dulbecco's Modified Eagle's Medium (DMEM) supplemented with 5 mM glutamine, 10% (v/v) fetal bovine serum (FBS), and 100 units/mL penicillin-streptomycin, in the 37°C incubator with 5% CO 2 . Subsequently, the indicated cell lines were subjected to transfection for 8 h, which was performed by using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) containing different combinations of indicated plasmids, and then allowed for 24 h recovery from transfection in a fresh medium before relevant experimentation. In addition, two Nfe2l1α-inducible and Nfe2l2-inducible cell lines, along with their control cells, were described by Wang et al. [47].

Real-Time qPCR Analysis.
Equal amounts of experimental cells were subjected to isolation of total RNAs by using the RNA simple Kit (Tiangen Biotech Co., Beijing, China). Then, 500 ng of total RNAs was added in a reverse-transcriptase reaction to generate the first strand of cDNA (with the Revert Aid First Strand Synthesis Kit from Thermo, Waltham, MA, USA). The synthesized cDNA fragments were served as the template for qPCR, in the GoTaq®qPCR Master Mix (from Promega), before being deactivated at 95°C for 10 min, and then amplified by 40 reaction cycles of the annealing at 95°C for 15 s and then extending at 60°C for 30 s. The final melting curve was validated to examine the amplification quality, whereas the mRNA expression level of β-actin served here as an optimal internal standard control. All the primers used for qPCR (Table S1) were synthesized by Tsingke (Chengdu, China).

Western Blotting Analysis.
Experimental cells were harvested in a lysis buffer (0.5% SDS, 0.04 mol/L DTT, pH 7.5), which was supplemented with the protease inhibitor cOmplete Tablets EASYpack. The lysates were then denatured immediately at 100°C for 10 min, sonicated sufficiently, and diluted in 3 × loading buffer (187.5 mmol/L Tris-HCl, pH 6.8, 6% SDS, 30% Glycerol, 150 mmol/L DTT, and 0.3% Bromphenol Blue) at 100°C for 5 min. Thereafter, equal amounts of protein extracts were subjected to separation by SDS-PAGE containing 5-12% polyacrylamide and subsequent visualization by immunoblotting with distinct primary antibodies as indicated (Table S1). On some occasions, the blotted membranes were stripped for 30 min and then were reprobed with additional primary antibodies. Here, β-actin served as an internal control to verify equal loading of proteins in each of electrophoretic wells. 5.6. Distinct Cis-Regulatory Reporter Gene Assays. Equal numbers (1:4 × 10 5 ) of indicated experimental cells were plated into 12-well plates. When allowed for growth to the density reaching 70-80% cell confluence, they were then transfected with Lipofectamine 3000 reagent (Invitrogen). For distinct lengths of the gene promoter studies, the pGL3-Basic vector containing 5 kb (−2248 to +2777), 2.7 kb (−2248 to +483), or 3 kb (−259 to +2777) of human Nfe2l1 promoter, or 3.2 kb (−2047 to +1210) of mouse Nfe2l1 promoter, or an empty pGL3-Basic vector as a blank control was cotransfected for 8 h with an indicated expression construct or an empty pcDNA3.1 as a control, plus another internal control pRL-TK reporter. Subsequently, the cells were allowed for a recovery from transfection in a fresh complete medium for 24 h, before they were lysed in a passive lysis buffer (E1910, Promega, Madison, WI, USA) for dualluciferase assays. About 20 μL of the supernatant of cells was assayed for the luciferase activity. Both firefly and renilla luciferase activities in each of the same samples were measured by the dual-luciferase reporter assay system (Promega). For the short consensus sequences of the indicated transcription factor-binding sites, the specific driven luciferase plasmids or corresponding paired point-mutants were cotransfected with the internal control pRL-TK, together with an expression construct for Nrf1, Nrf2, Pitx2, or an empty pcDNA3.1. All the resultant data were normalized and calculated as a fold change (mean ± S:D) relative to the activity of the control group (at a given value of 1.0).

Statistical
Analysis. The statistical significance of changes was determined using the Student's t-test and/or Multiple Analysis of Variations (MANOVA). All the data presented in this study are shown as a fold change (mean ± S:D), each of which represents at least 3 independent experiments undertaken on separate occasions that were each performed in triplicate. In addition, statistical analysis of RNA-sequencing data was carried out as described by Wang et al. [47].

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

Disclosure
It should also be noted that the preprinted version of this paper had been initially posted in bioRxiv at https://doi.org/ 10.1101/2020.02. 12.945568. A commentary letter on this preprinted paper had been published in Developmental Cell [63].

Conflicts of Interest
The authors declare no conflict of interest.

Authors' Contributions
S.Z. performed the most experiments with the help of Y.X., S.H., and L.Q., except that Y.D. did the remaining experiments and repeated part of the work by S.Z. Both S.Z. and Y.D. collected the relative data and also made draft of this manuscript with most figures and supplemental tables. L.Q. also established some cell lines used in this study. Lastly, Y.Z. designed and supervised this study, analyzed all the data, helped to prepare all figures with cartoons, and wrote and revised the paper. Of note, Shuwei Zhang and Yangxu Deng contributed equally to this work. 30 Oxidative Medicine and Cellular Longevity