Carboxyl Group-Modified Myoglobin Induces TNF-α-Mediated Apoptosis in Leukemia Cells

Previous studies have shown that chemical modification may increase the activity of proteins or confer novel activity to proteins. Some studies have indicated that myoglobin (Mb) is cytotoxic; however, the underlying mechanisms remain unclear. In this study, we investigated whether chemical modification of the carboxyl group by semicarbazide could promote the Mb cytotoxicity in human leukemia U937 cells and the underlying mechanism of semicarbazide-modified myoglobin (SEM-Mb)-induced U937 cell death. The semicarbazide-modified Mb (SEM-Mb) induced U937 cell apoptosis via the production of cleaved caspase-8 and t-Bid, while silencing of FADD abolished this effect. These findings suggest that SEM-Mb can induce U937 cell death by activating the death receptor-mediated pathway. The SEM-Mb inhibited miR-99a expression, leading to increased NOX4 mRNA and protein expression, which promoted SIRT3 degradation, and, in turn, induced ROS-mediated p38 MAPK phosphorylation. Activated p38 MAPK stimulated miR-29a-dependent tristetraprolin (TTP) mRNA decay. Downregulation of TTP slowed TNF-α mRNA turnover, thereby increasing TNF-α protein expression. The SEM-Mb-induced decrease in cell viability and TNF-α upregulation were alleviated by abrogating the NOX4/SIRT3/ROS/p38 MAPK axis or ectopic expression of TTP. Taken together, our results demonstrated that the NOX4/SIRT3/p38 MAPK/TTP axis induces TNF-α-mediated apoptosis in U937 cells following SEM-Mb treatment. A pathway regulating p38 MAPK-mediated TNF-α expression also explains the cytotoxicity of SEM-Mb in the human leukemia cell lines HL-60, THP-1, K562, Jurkat, and ABT-199-resistant U937. Furthermore, these findings suggest that the carboxyl group-modified Mb is a potential structural template for the generation of tumoricidal proteins.


Introduction
Chemical modifications have been widely used to explore the functional significance of certain amino acid residues in proteins. If modifications of proteins result in a decrease in their biological activities, it can be inferred that these modified residues are important for the manifestation of protein activity [1][2][3]. However, experimental evidence has indicated that the activity of Candida antarctica and Thermomyces lanuginose lipase increases after modification of the carboxyl group using ethylenediamine [4]. Modification of Lys and Arg residues with maleic anhydride and 3-hydroxyphthalic anhydride confers anti-HIV activity to ovalbumin, which lacks antiviral activity prior to chemical modification [5,6]. Glycated bovine serum albumin and ovalbumin with carboxyl groups modified with paminophenyl α-D-mannopyranoside show membrane-permeabilizing and antibacterial activities [7][8][9][10]. The use of semicarbazide to block the negative charge in the carboxylic acid groups of α-lactalbumin (LA) enables it to destabilize bilayer membranes and induce cancer cell death [11,12]. These results indicate that chemical modifications can increase

The Death Receptor-Mediated Apoptotic Pathway Is Involved in SEM-Mb-Mediated Cytotoxicity
Unlike Mb, SEM-Mb was cytotoxic to the U937 cells ( Figure 1A). The 50% inhibitory concentration (IC50) was approximately 1.5 µM after 24 h of treatment ( Figure 1A). SEM-Mb-mediated cytotoxicity was further investigated using a single dose of 1.5 µM. The SEM-Mb treatment increased the number of annexin V-stained cells ( Figure 1B) and induced degradation of procaspase-3/-8 and PARP ( Figure 1C). Inhibition of caspase-3/-8 activity maintained cell viability even after SEM-Mb treatment ( Figure 1D), suggesting that SEM-Mb-treated U937 cells underwent apoptosis. There is compelling evidence that t-Bid, produced by caspase-8-mediated cleavage of Bid, triggers disruption of the mitochondrial membrane potential (∆Ψm), leading to mitochondria-mediated apoptosis [26,27]. The SEM-Mb-treated U937 cells produced t-Bid ( Figure 1E) and caused mitochondrial membrane depolarization ( Figure 1F), suggesting that SEM-Mb-induced apoptosis is related to the activation of the caspase-8/mitochondria-mediated pathway. Treatment of U937 cells with SEM-Mb did not alter the expression of MCL1, BCL1, BCL2L1, and BAK but increased the expression of BAX protein. Accumulating evidence indicates that the TNF-family death receptors recruit FADD and procaspase-8 to activate caspase-8, thereby triggering caspase-3-mediated apoptosis [27]. Knockdown of FADD attenuated the cleavage of caspases-8/-3 and Bid in the cells exposed to SEM-Mb ( Figure 1G). Furthermore, FADD silencing reduced SEM-Mb-induced cell death ( Figure 1H). These findings indicated that the death receptormediated apoptotic pathway is involved in SEM-Mb-mediated cytotoxicity.

SEM-Mb Induces Ca 2+ -Elicited ROS Generation, Which in Turn Reduces U937 Cell Viability
In a previous study, Zager and Burkhart [24] reported that Mb-induced HK-2 cell death is related to Ca 2+ and ROS signaling pathways. Thus, we analyzed whether Ca 2+ and ROS were involved in the cytotoxicity of SEM-Mb against U937 cells. The SEM-Mb treatment triggered maximal ROS production after 6 h of treatment (Supplementary Figure S1A) and continued to increase the intracellular Ca 2+ concentration ([Ca 2+ ]i) over 24 h (Supplementary Figure S1B). The BAPTA-AM (a Ca 2+ chelator) pretreatment alleviated ROS production induced by SEM-Mb (Supplementary Figure S1C), whereas NAC (an ROS scavenger) did not reduce [Ca 2+ ]i in SEM-Mb-treated cells (Supplementary Figure S1C). Intracellular ROS are mainly produced by mitochondria and NADPH oxidase (NOX) [28,29]. Therefore, we analyzed the effect of NOX inhibitors on SEM-Mb-induced ROS production. Pretreatment with BAPTA-AM, NAC, and GLX351322 (an NOX4 inhibitor), but not ML171 (an NOX1 inhibitor), reduced ROS generation in SEM-Mb-treated cells (Supplementary Figure S1D). The GLX351322 failed to inhibit the SEM-Mb-induced [Ca 2+ ]i elevation (Supplementary Figure S1C). SEM-Mb treatment also increased mitochondrial ROS levels (Supplementary Figure S1E). The BAPTA-AM, NAC, GLX351322, and Mito-TEMPO (mitochondria-targeted antioxidant) attenuated the effects of SEM-Mb on mitochondrial ROS production, ∆Ψm loss, and cell death (Supplementary Figure S1F-H). Evidently, SEM-Mb induced Ca 2+ -elicited ROS generation, which, in turn, reduced U937 cell viability.

SEM-Mb Induces Ca 2+ -Mediated Downregulation of miR-99a, Leading to Upregulation of NOX4 Expression
As our data indicated the involvement of NOX4 in SEM-Mb-induced ROS production, we analyzed the effect of SEM-Mb treatment on NOX4 expression. The SEM-Mb increased NOX4 protein and mRNA expression without affecting the NOX4 promoter luciferase activity (Figure 2A-C). The SEM-Mb treatment increased the half-life of NOX4 mRNA as shown by the mRNA turnover assay ( Figure 2D). The BAPTA-AM pretreatment abolished SEM-Mb-induced NOX4 upregulation ( Figure 2E,F), whereas NAC had no inhibitory effect ( Figure 2G). These results indicate that SEM-Mb induced Ca 2+ -mediated NOX4 expression. Accumulating evidence has shown that miRNAs induce translational repression and mRNA degradation [30]. Several miRNAs including miR-137, miR-25, miR-99a, and miR-146a are known to promote NOX4 mRNA decay [31][32][33]. Therefore, we analyzed the expression of these miRNAs. Among the four miRNAs, SEM-Mb reduced miR-99a expression ( Figure 2H) and BAPTA-AM pretreatment abolished this suppression ( Figure 2I). Transfection with hsa-miR-99a diminished the SEM-Mb-induced upregulation of NOX4 protein and mRNA levels ( Figure 2J,K). Consistent with this, the miR-99a mimic inhibited the SEM-Mb-induced stabilization of NOX4 mRNA ( Figure 2L). These results indicate that SEM-Mb induces Ca 2+mediated downregulation of miR-99a, leading to the upregulation of NOX4 expression. (Top panel) qRT-PCR analysis of miR-99a levels in hsa-miR-99a-transfected cells (mean ± SD, * p < 0.05). (Bottom panel) Western blot analysis of NOX4 expression in hsa-miR-99a-transfected cells after SEM-Mb treatment. (K) Transfection of hsa-miR-99a inhibited SEM-Mb-induced increase in NOX4 mRNA expression (mean ± SD, * p < 0.05). (L) Transfection of hsa-miR-99a increased NOX4 mRNA turnover in SEM-Mb-treated cells. The negative control miRNA-and hsa-miR-99a-transfected cells were treated with 1.5 µM SEM-Mb for 24 h, and then incubated with 10 µg/mL actinomycin D for the indicated time periods. The level of NOX4 mRNA was analyzed by qRT-PCR (mean ± SD, * p < 0.05). NOX4 mRNA levels in cells not treated with actinomycin D were taken as 1.

The Cytotoxicity of SEM-Mb Is Related to ROS-Mediated p38 MAPK Activation
Studies have indicated that MAPKs regulate apoptosis [34,35], and that Ca 2+ and ROS signaling modulate MAPK phosphorylation [36,37]. Therefore, we analyzed the role of phosphorylated MAPK in SEM-Mb cytotoxicity. The SEM-Mb treatment increased p-p38 MAPK levels and reduced p-ERK levels; however, the levels of p-JNK remained unchanged ( Figure 3A). Inhibition of p38 MAPK using SB202190 reversed the phosphorylation of p38 MAPK and ERK in SEM-Mb-treated cells ( Figure 3B). In agreement, studies by Liu et al. [38] have reported inverse changes in p-p38 MAPK and p-ERK levels in U937 cells due to crosstalk between p38 MAPK and ERK. Pretreatment with NOX4 inhibitor or Mito-TEMPO abolished SEM-Mb-induced p38 MAPK activation ( Figure 3C,D). Moreover, SB202190 inhibited SEM-Mb-induced cell death and ∆Ψm loss ( Figure 3E,F). These findings highlight the causal relationship between ROS-mediated p38 MAPK activation and SEM-Mb-mediated cytotoxicity.

SEM-Mb Promotes TNF-α Expression via A p38 MAPK-Dependent Post-Transcriptional Pathway
As our results suggest that SEM-Mb cytotoxicity is associated with a death receptormediated pathway, we sought to analyze the effect of SEM-Mb treatment on the expression of TNF-α family proteins. The SEM-Mb increased TNF-α protein levels, but did not alter the expression of FasL, Fas, TRAIL, DR4, DR5, TNFR1, or TNFR2 ( Figure 4A). Treatment with SEM-Mb did not affect TNF-α promoter luciferase activity but increased TNF-α transcript levels and slowed TNF-α mRNA decay ( Figure 4B-D). SB202190 abrogated the effects of SEM-Mb on TNF-α expression and mRNA stabilization ( Figure 4E-G). Collectively, our data confirmed that SEM-Mb promotes TNF-α expression via a p38 MAPK-dependent post-transcriptional pathway.

SEM-Mb-Induced p38 MAPK Activation Increases miR-29a Expression, Leading to the Inhibition of TTP Expression
Liu et al. [39] discovered that hydroquinone-induced upregulation of protein phosphatase 2A catalytic subunit α (PP2Acα) promoted the degradation of tristetraprolin (TTP) in U937 cells, thereby slowing TTP-mediated TNF-α mRNA decay. We analyzed the expression of TTP and PP2Acα and observed that SEM-Mb treatment reduced TTP but not PP2Acα protein expression ( Figure 5A). Additionally, SEM-Mb did not affect TTP promoter luciferase activity, but reduced TTP transcript levels and mRNA stability ( Figure 5B-D). These findings suggest that SEM-Mb downregulates TTP expression through a post-transcriptional mechanism, rather than through proteasomal degradation. The SEM-Mb failed to induce TNF-α upregulation in cells ectopically expressing TTP ( Figure 5E,F). Overexpression of TTP deprived SEM-Mb of the ability to stabilize TNF-α mRNA ( Figure 5G) and induce cell death ( Figure 5H). The SB202190 abrogated the ability of SEM-Mb to downregulate TTP expression and destabilize TTP mRNA ( Figure 5I-K). A previous study revealed that miR-29a targets the 3 -UTR of TTP mRNA, thus suppressing TTP expression [40]. We examined whether miR-29a was involved in SEM-Mb-induced TTP suppression. The SEM-Mb treatment increased miR-29a expression, whereas SB202190 pretreatment inhibited SEM-Mb-induced events ( Figure 5L). Transfection with anti-miR-29a impaired the ability of SEM-Mb to reduce TTP protein and mRNA expression ( Figure 5M,N). Compared to the negative anti-miR control-transfected cells, anti-miR-29a-transfeced cells decreased the decline of TTP mRNA after SEM-Mb treatment ( Figure 5O). These findings confirm that SEM-Mb-induced p38 MAPK activation increases miR-29a expression, leading to the inhibition of TTP expression.  The negative control anti-miR-and anti-miR-29a-transfected cells were treated with 1.5 µM SEM-Mb for 24 h, and then incubated with 10 µg/mL actinomycin D for the indicated time periods. The levels of TTP mRNA were analyzed by qRT-PCR (mean ± SD, * p < 0.05).

SEM-Mb Triggers p38 MAPK/TTP Axis-Dependent TNF-α Expression by Promoting SIRT3 Degradation
Previous studies have indicated that SIRT3 inhibition promotes mitochondrial ROS production in U937 cells [41]. Therefore, we analyzed the effect of SEM-Mb treatment on SIRT3 expression. The SEM-Mb treatment reduced SIRT3 protein levels but not mRNA expression ( Figure 6A,B). Pretreatment with MG132 abrogated SEM-Mb-induced SIRT3 downregulation ( Figure 6C), indicating that SEM-Mb treatment promotes SIRT3 degradation. Inhibition or depletion of NOX4 diminished the SEM-Mb-induced downregulation of SIRT3 ( Figure 6D,E), indicating that NOX4 suppresses SIRT3 expression. The enforced expression of SIRT3 alleviated SEM-Mb-induced mitochondrial ROS production and cell death ( Figure 6F,G). Compared to cells transfected with the empty vector, SEM-Mb failed to alter the expression of p-p38 MAPK, TNF-α, and TTP in pCMV-N-His-SIRT3-transfected cells ( Figure 6H). These findings confirmed that SEM-Mb elicited mitochondrial ROS generation by promoting SIRT3 degradation, thereby triggering p38 MAPK/TTP axis-dependent TNF-α expression.

SEM-Mb Induces Apoptosis in HL-60 Cells via NOX4/SIRT3/p38 MAPK/TTP Axis-Dependent TNF-α Expression
To further examine whether SEM-Mb exhibited cytotoxicity in other leukemia cells through a similar pathway, we investigated its cytotoxicity in other leukemia cells. Daunorubicin and cytarabine are standard drugs for the treatment of acute myeloid leukemia [42,43]. Some studies have shown that, compared to U937 cells, HL-60 cells are more resistant to daunorubicin and cytarabine, owing to which HL-60 cells highly expressed myeloperoxidase [44,45]. Therefore, we investigated the cytotoxicity of SEM-Mb in leukemia HL-60 cells. Unlike Mb, SEM-Mb was cytotoxic to HL-60 cells with an IC50 value of approximately 2 µM ( Figure 7A). The SEM-Mb treatment induced apoptosis in HL-60 cells ( Figure 7B), whereas caspase inhibitors maintained cell viability, regardless of SEM-Mb treatment ( Figure 7C). Use of BAPTA-AM, but not GLX351322, inhibited the SEM-Mb-induced [Ca 2+ ]i elevation ( Figure 7D). Both BAPTA-AM and GLX351322 abrogated cellular and mitochondrial ROS production in SEM-Mb-treated cells and protected HL-60 cells from SEM-Mb cytotoxicity ( Figure 7E-G). Pretreatment with Mito-TEMPO attenuated SEM-Mb-induced mitochondrial ROS production and cell death ( Figure 7F,G). The SEM-Mb-triggered NOX4 upregulation was alleviated by pretreatment with BAPTA-AM ( Figure 7H). Collectively, our data indicate that SEM-Mb stimulated mitochondrial ROS production in HL-60 cells through Ca 2+ -mediated NOX4 expression. The SEM-Mb decreased SIRT3 and TTP expression but increased p38 MAPK and TNF-α levels, whereas GLX351322 inhibited SEM-Mb-induced effects ( Figure 7I). Transfection with pCMV3-N-His-SIRT3 eliminated the SEM-Mb-induced p38 MAPK phosphorylation and cell death ( Figure 7J,K). Pretreatment with SB202190 increased TTP expression and decreased TNF-α expression in HL-60 cells treated with SEM-Mb ( Figure 7L). Overexpression of TTP attenuated the induction of TNF-α expression and cell death induced by SEM-Mb ( Figure 7M,N). These results confirm that SEM-Mb induces apoptosis in HL-60 cells via NOX4/SIRT3/p38 MAPK/TTP axis-dependent TNF-α expression.    Figure S2B). Furthermore, SEM- Figure S2C). Pretreatment with SB202190 inhibited SEM-Mb induced apoptosis and TNF-α upregulation, indicating that SEM-Mb-induced death of THP-1, K562, and Jurkat cells occurs through p38 MAPK-dependent TNF-α expression. The BCL2 inhibitor ABT-199 has been used to treat chronic lymphocytic leukemia and AML [46]. However, ABT-199-induced upregulation of MCL1 expression confers resistance to this drug [46]. We further examined whether SEM-Mb potentially induced the death of ABT-199resistant U937 cells and observed that the cytotoxicity of SEM-Mb to ABT-199-resistant U937 cells was related to p38 MAPK-dependent TNF-α expression (Supplementary Figure S2A-C). Collectively, these findings indicate that SEM-Mb is cytotoxic to other leukemia cells.

Discussion
Our data indicate that SEM-Mb induces Ca 2+ -mediated miR-99a downregulation, resulting in increased NOX4 expression in U937 cells. The NOX4-triggered SIRT3 degradation promotes mitochondrial ROS/p38 MAPK-mediated miR-29a expression, thereby downregulating TTP expression. As TTP destabilizes TNF-α mRNA, SEM-Mb-induced TTP downregulation increases TNF-α expression ( Figure 8). Consequently, SEM-Mb activates the TNF-α-mediated death pathway in U937 cells. Abolishment of TNF-α upregulation protected U937 cells from SEM-Mb-induced cell death, indicating a causal role of TNF-α in SEM-Mb-mediated cytotoxicity. A similar mechanism also explains SEM-Mb-induced death of HL-60 cells. In addition, p38 MAPK-mediated TNF-α expression was associated with cytotoxicity of SEM-Mb against THP-1, K562, Jurkat, and ABT-199-resistant U937 cells. Our data showed that SEM-Mb effectively inhibited the viability of chemo-resistant leukemia cells. Recent studies have revealed that docetaxel induces SIDT2-mediated lysosomal degradation of miR-25, resulting in decreased NOX4 mRNA decay [47]. However, the depletion of SIDT2 using siRNA did not increase miR-99a expression in SEM-Mb-treated U937 cells (data not shown). Therefore, the mechanism underlying SEM-Mb-induced miR-99a downregulation remains unclear. Compelling evidence indicates that SIRT3-mediated deacetylation is essential for the activity of SOD2 to remove mitochondrial ROS [48,49]. Shanmugasundaram et al. [50] demonstrated that NOX4 modulates mitochondrial ROS production. Consistent with these findings, our data showed that SEM-Mb-induced NOX4 expression promotes SIRT3 degradation, thereby increasing mitochondrial ROS production. A previous study indicated that Skp2, an E3 ubiquitin protein ligase, modulates SIRT3 degradation [51]. Pretreatment with SZL P1-41 (an Skp2 inhibitor) did not affect the SEM-Mb-induced SIRT3 downregulation (data not shown). Therefore, the mechanism underlying NOX4-elicited SIRT3 degradation remains elusive. Wang et al. [41] demonstrated that albendazole induces TTP degradation in U937 cells by triggering p38 MAPKdependent TTP phosphorylation. Other studies have indicated that hydroquinone induces p38 MAPK/PP2A axis-dependent TTP degradation in U937 cells [47]. The data in this study revealed that p38 MAPK-mediated miR-29a expression destabilizes TTP mRNA in U937 cells after SEM-Mb treatment. Taken together, these results suggest that the regulatory mechanism of p38 MAPK in suppressing TTP expression is stimulus-dependent. Notably, previous studies have shown that, in addition to apoptosis, increased ROS production triggers necroptosis or ferroptosis [52]. Although necroptosis can be activated by the ligand binding to TNF family receptors, activated caspase-8 is not involved in the necroptosis pathway [53]. On the other hand, the mechanism of ferroptosis is independent of the caspase-mediated pathway [54]. As the cytotoxicity of SEM-Mb is related to the activation of caspase-8/-3 ( Figure 1D), it is conceivable that SEM-Mb-induced cell death is not mediated through necroptosis or ferroptosis. The heme oxygenase activity of Mb catalyzes the production of cellular ROS [55,56]. Because SEM-Mb lacks a heme group as shown by the loss of the Soret peak, we hypothesized that SEM-Mb-induced ROS generation is independent of heme oxygenase activity. Braganza et al. [57] argued that the heme group of Mb is essential for Mb-induced inhibition of breast cancer cell proliferation and that the growth inhibitory mechanism of intracellularly expressed Mb is achieved through the inhibition of mitochondrial fission. Our results showed that extracellular treatment with SEM-Mb, but not Mb, activated the TNF-α-mediated death pathway in leukemia cells. Zager and Burkhart [24] found that Mb treatment induced the death of the human proximal tubular cell line HK-2. Other studies have suggested that megalin-and cubilin-mediated endocytosis of Mb in the renal proximal tubule leads to nephrotoxicity [58]. Therefore, the possibility that internalization of SEM-Mb modulates its cytotoxicity warrants further investigation. Notably, treatment with 10 mg/mL (~588 µM) Mb for 24 h resulted in an approximately 65-85% reduction in the viability of HK2 cells [24].
Our results indicate that SEM-Mb at a concentration of ≤2.0 µM considerably reduced the viability of U937 and HL-60 cells. It is evident that the modification of the carboxyl groups renders Mb more cytotoxic. Some studies have indicated that free heme released from hemoproteins induces TNF-α-dependent death in macrophages [59]. Given that SEM-Mb does not contain a heme group, we believe that the mechanism by which SEM-Mb induces TNF-α upregulation in leukemia cells differs from that of heme in macrophages.
Compelling evidence indicates that oleic acid stabilizes the molten globular conformation of apo-LA, which is essential for the cytotoxicity of apo-LA-oleic acid complexes [14,60]. The structure of apo-Mb has a molten globule-like character [61], and lipid-protein complexes formed by apo-Mb and oleic acid also show antitumor activity [62]. Notably, carboxyl group-modified LA with a molten globule conformation exhibits cytotoxicity in U937 and breast cancer MCF-7 cells [11,12]. Together with the finding that modification of the carboxyl groups causes dramatic distortion of the Mb structure [25], our results underscore the idea that conformational changes are related to SEM-Mb cytotoxicity. Previous studies have indicated that the apo-Mb 56-131 peptide exhibits membrane-perturbing activity in physiological solutions, whereas apo-Mb and Mb failed to show this activity [17]. More importantly, the apo-Mb 56-131 peptide loses the characteristic α-helical structure of apo-Mb and Mb [17]. These results indicate that the inherent activity of the peptide fragment at positions 56-131 was hindered by the intact protein structure of Mb and apo-Mb. Although the data from this study revealed that SEM-Mb is cytotoxic to leukemia cells, the application of SEM-Mb as a tumoricidal agent requires further study to evaluate its feasibility. Importantly, the potential immunogenicity of SEM-Mb for inducing antibodies against SEM-Mb in vivo should be considered. Several studies have shown that the repeated use of protein pharmaceuticals induces antibodies against these agents, thereby reducing their efficacy [63,64]. To eliminate this possibility, a viable idea is to decipher the active region in SEM-Mb responsible for anti-leukemia activity and then use this active region to develop peptidomimetics. Notably, El-Maksound et al. [65] reported that milk protein-mucilage complexes formed by milk proteins with Isabgol husk mucilage or Ziziplus spina-christi mucilage enhanced the anti-cancer activity of milk proteins. As reducing the dose and frequency of using SEM-Mb can delay its potential antibody induction, complexation of SEM-Mb with natural components may be a feasible strategy to increase its anti-leukemia activity at a reduced dose.

Modification of Mb with Semicarbazide
Chemical modification of the carboxyl groups in Mb with semicarbazide (Supplementary Figure S3) was conducted according to the procedure described in [25]. A MALDI-TOF analysis showed that there were approximately 19 semicarbazide-conjugated carboxyl groups in the SEM-Mb. Measurement of circular dichroism spectra indicated that the carboxyl group modification profoundly distorted the secondary structure of Mb [25].

Cell Culture
Human AML U937, AML HL-60, AML THP-1, CML K562, and ALL Jurkat cells from BCRC (Hsinchu, Taiwan) were grown in RPMI-1640 medium supplemented with 10% FCS, 2 mM L-glutamine, 1% sodium pyruvate, and 1% streptomycin/penicillin. Cells were maintained in a humidified incubator with a 95% air/5% CO 2 atmosphere at 37 • C. Cell viability was detected using a MTT assay [66]. Cell apoptosis was detected by staining cells with annexin V-FITC kit (Thermo Fisher Scientific) followed by flow cytometric analysis. The ABT-199-resistant U937 cells were prepared according to the procedure described in our previous studies [67]. Cell culture supplements and media were the products of GIBCO/Life Technologies Inc.

Detection of ROS, [Ca 2+ ]i, and Mitochondrial Depolarization
Intracellular and mitochondria ROS levels were detected using H 2 DCFDA and Mi-toSOX Red, respectively. Rhodamine 123 and Fluo-4 AM were employed to measure mitochondrial membrane potential (∆Ψm) and [Ca 2+ ]i, respectively. After treatment in the indicated conditions, cells were incubated with 10 µM H 2 DCFDA, 5 µM MitoSOX, 1 mM Fluo-4 AM, or 20 nM rhodamine 123 for 15 min. Afterwards, cells were washed in phosphate-buffered saline, and the fluorescent signal was measured using flow cytometry.

qRT-PCR
The RNeasy mini kit (QIAGEN, Leiden, The Netherlands) was used to extract total cellular RNA. The M-MLV reverse transcriptase (Promega, Madison, WI, USA) was used to synthesize cDNAs. The qPCR was conducted using GoTaq qPCR Master mix (Promega). Gene expression was calculated using 2 −∆∆Ct method relative to GAPDH mRNA levels. The PCR primer sequences are shown in Supplementary Table S1. To determine TNF-α, TTP, and NOX4 mRNA half-lives, transcription was inhibited by incubation with 10 µg/mL actinomycin D for the indicated time periods.

Transfection with Plasmid and siRNA
The plasmids pCMV3-N-His-SIRT3 and pCMV-TTP-HA were described in our previous studies [41]. Gene silencing experiments were performed using FADD siRNA, NOX4 siRNA, and negative control siRNA (Santa Cruz Biotechnology). TurboFect™ Transfection Reagent (Thermo Fisher Scientific) was used to transfect plasmids and siRNAs into cells.

Stem-Loop qRT-PCR for miRNAs
Reverse transcription of miRNAs was performed using ImProm-II Reverse Transcription System (Promega) and stem-loop reverse transcription (RT) primer. The expression levels of miRNAs were analyzed using GoTaq qPCR Master mix (Promega). Stem-loop RT primers and qPCR primers for miRNAs are listed in Supplementary Table S2.

Transient Transfection of miRNA Mimic and Inhibitor
Ambion TM anti-miR-29a, hsa-miR-99a, negative control miR, and negative anti-miR control from Life Technologies Corporation (Carlsbad, CA, USA) were transfected into cells using TurboFect™ Transfection Reagent (Thermo Fisher Scientific). After 24 h posttransfection, the cells were treated with SEM-Mb for 24 h.

Luciferase Assay
The luciferase construct pGL3-TNF-α and pGL3-NOX4 were described elsewhere [12,69]. The human TTP promoter region containing nucleotides −1345 to +66 was subcloned into pGL3-basic vector (Promega) between SacI and XhoI sites. TurboFect™ Transfection Reagent (Thermo Fisher Scientific) was used to transfect the promoter constructs into cells. Luciferase activity was measured using dual-luciferase reporter assay system kit (Promega). To normalize the transfection efficiency, the luciferase signal was calculated using the Renilla luciferase signal as an internal control.

Statistical Analysis
Results are expressed as the mean ± SD of at least three independent experiments. The statistical significance of differences between two groups was evaluated using Student's t-test. Comparison of multiple groups was performed using one-way ANOVA with Tukey's test. p value of < 0.05 was defined statistically significant.

Conclusions
Our data reveal that chemical modification of the carboxyl groups with semicarbazide enables Mb to show cytotoxicity and demonstrate that the chemically modified Mb activates TNF-α-mediated apoptosis in U937 cells through a NOX4/SIRT3/p38 MAPK/TTP axismediated pathway. A similar pathway also describes the cytotoxicity of semicarbazidemodified Mb in HL-60, THP-1, K562, Jurkat, and ABT-199-resistant U937 cells. Notably, Mb does not show cytotoxic activity prior to modification of the carboxyl group. Together with the finding that carboxyl group modification leads to conformational distortion in Mb [25], the data from this study highlight that chemical modification is a viable strategy for exploring intrinsic activities hindered by the conformation of protein. Furthermore, our data suggest that the carboxyl group-modified Mb is a potential structural template for the generation of tumoricidal proteins.