Reliability of ROS and RNS detection in hematopoietic stem cells − potential issues with probes and target cell population

Recent studies have revealed that a population of hematopoietic stem cells (HSCs) that resides in the hypoxic niche and mostly relies on anaerobic glycolysis is highly sensitive to oxidative stress (Simsek et al., 2010; Suda et al., 2011; Unwin et al., 2006). HSCs, and stem cells in general (Han et al., 2008), are highly vulnerable to increased levels of reactive oxygen species (ROS), whereas committed hematopoietic progenitor cells (HPCs) are not (Chen et al., 2008b; Ito et al., 2004; Jang and Sharkis, 2007). Nevertheless, an appropriate production of ROS is required for HSC functions, such as mobilization, survival, differentiation and proliferation (Guzy and Schumacker, 2006; Hole et al., 2010; Hosokawa et al., 2007; Jang and Sharkis, 2007; Sattler et al., 1999; Tesio et al., 2011; Wang et al., 2013). Furthermore, ROS modify gene expression and the activity of key metabolic regulators, thereby determining specific metabolic patterns and cell-fate decision (Bigarella et al., 2014). However, an excess of ROS is associated with DNA damage, lipid peroxidation, senescence and apoptosis (Naka et al., 2008; Navarro-Yepes et al., 2014; Shao et al., 2011). Furthermore, elevated ROS levels and accumulation of damaged DNA in human cord blood Lin⁻CD34⁺38⁻ cells (enriched in HSCs) result in a reduction of their capacity to reconstitute hematopoiesis in immunosuppressed mice (Yahata et al., 2011). In addition to ROS, reactive nitrogen species (RNS) are integral to the redox status of the cell; ROS and RNS have overlapping functions and also affect each other's functions (Nathan and Cunningham-Bussel, 2013). Considering the role of RNS, nitric oxide (NO•) inhibits proliferation of HSCs and HPCs, and blocks their progression through the cell cycle, as well as induces apoptosis of CD34⁺ bone marrow (BM) cells in vitro (Reykdal et al., 1999). Furthermore, inhibition of nitric oxide synthase (NOS) increases the number of stem cells in the BM and the longevity of hematopoiesis in continuous BM cultures (Epperly et al., 2007; Michurina et al., 2004). However, compared with the work on ROS, there are considerably fewer studies that address RNS-mediated effects on HSCs and the signaling networks they participate in (Aleksinskaya et al., 2013; Bonafè et al., 2015; Nogueira-Pedro et al., 2014).

It has proven difficult to identify the specific ROS and RNS species that affect the functions of HSCs. This is partly due to a lack of suitable methods and chemical probes that allow the unequivocal distinction between different ROS and RNS, and to detect the complex intracellular chemistry of these diffusible and short-lived species. Furthermore, with the exception of a few studies that have been performed by using cell populations that were highly enriched in functional HSCs (Chuikov et al., 2010; Jung et al., 2013; Liu et al., 2009; Taniguchi Ishikawa et al., 2012), most studies that describe the role or consequence of ROS levels in redox measurements of HSCs were performed in heterogeneous cell populations that consist mainly of committed progenitor cells and only a small portion of functional HSCs (Table 1). In this Opinion article, we discuss how these limitations can lead to incorrect interpretation of data. Moreover, on the basis of recent advances in research, we present strategies that might enable to accurately detect ROS and RNS in HSCs and, thus, yield a better interpretation of the obtained results. We believe that these methods might result in important progress in our understanding of the redox regulation of stem cell physiology.

The biochemical pathways in aerobic organisms that utilize O₂ continuously generate the metabolic by-products ROS and RNS. Thus, aerobes have developed physiological mechanisms that enable ROS and RNS to be either detoxified or used as intracellular or intercellular signaling messengers to maintain cellular homeostasis. However, ROS and RNS are not single entities but a broad range of chemically distinct reactive species with diverse biological reactivities (Nathan and Cunningham-Bussel, 2013). There are several intracellular sources of ROS, including the mitochondrial electron transport chain (ETC), the membrane-bound NADPH oxidase (NOX) complex, the endoplasmic reticulum, oxidoreductase enzymes and metal-catalyzed oxidation reactions (Nathan and Cunningham-Bussel, 2013; Wang et al., 2013) (Fig. 1). Leaked electrons from the ETC can be captured by O₂ to form the anion superoxide (O₂•⁻), which – together with (NO•) – initiates the production of ROS and RNS. O₂•⁻ can be rapidly converted into hydrogen peroxide (H₂O₂) by superoxide dismutases (SODs) or can react with NO• to form peroxynitrite (ONOO⁻). H₂O₂ can be converted into the hydroxyl radical (•HO) in the presence of metal ions (Fe²⁺ or Cu ²⁺) in the Fenton reaction. In the presence of CO₂, which is formed during multiple intracellular enzymatic reactions, ONOO⁻ gives rise to the RNS nitrosoperoxocarboxylate (ONO₂CO₂⁻), which decomposes rapidly into the nitrogen dioxide (NO₂•) and carbonate (CO₃•⁻) radicals. The vast majority of biological reactions of ONO₂CO₂ ⁻, including those with oxidant-sensitive probes, are mediated by NO₂• and CO₃•⁻ (Ferrer-Sueta and Radi, 2009). In the presence of chloride or thiocyanate, myeloperoxidase catalyzes the conversion of H₂O₂ into the oxidants hypochlorous acid (HOCl) or hypothiocyanous acid (HOSCN), respectively (Wardman, 2007; Winterbourn, 2014). The cellular levels of these oxidants are determined by the balance between their generation and capture by a number of antioxidant enzymes (e.g. catalases, peroxiredoxins, thioredoxins, thioredoxin reductases, and methionine sulphoxide reductases) (Nathan and Cunningham-Bussel, 2013). In addition, important redox cellular regulators are enzymes of the glutathione redox cycle, that detoxify H₂O₂ by using the reduced glutathione (GSH) (i.e. glutathione peroxidase) and regenerate oxidized glutathione (GSSG) to reduced glutathione (glutathione reductase) (Nathan and Cunningham-Bussel, 2013).

ROS and RNS have distinct chemical properties that result from differences in their reactivity, half-live and interaction with lipid moieties (D'Autréaux and Toledano, 2007). A selective reactivity with biological targets is the basis for the specific signaling pathways that are induced by individual ROS and RNS (D'Autréaux and Toledano, 2007). For instance, O₂•⁻ preferentially reacts with proteins that contain iron–sulphur [Fe–S] clusters, whereas the uncharged H₂O₂ does not react with these proteins. Instead, H₂O₂ preferentially reacts with cysteine thiol groups that are present in the catalytic centers of enzymes. Moreover, not all cysteine residues are equally able to be oxidized by H₂O₂; this provides specificity of signaling induced by this ROS. NO• interacts with proteins that contain Fe–S clusters, heme (Cooper, 1999) or specific cysteine thiol groups (Marozkina and Gaston, 2012). The reactions mentioned above induce allosteric or conformational changes in target proteins (Bigarella et al., 2014; Finkel, 2011). These changes affect the function of proteins and their interaction with other proteins, and so can trigger distinct signaling events (D'Autréaux and Toledano, 2007; Stamler et al., 1992).

Several metabolic regulators and signaling pathways have been implicated in the maintenance of the redox homeostasis of HSCs. Activities of these regulators predispose HSCs to certain cell fates as summarized in Table 1.

It must be noted that the definition of an HSC is functional, based on the self-renewal capacity and differentiation potential of the cell. Unlike committed progenitors, HSCs are rare (1 in 10,000 within adult bone marrow cells), multipotent, self-renewing cells with a remarkable regenerative capacity that enables these cells to sustain hematopoiesis over a lifetime. As stated by John Dick: “The only conclusive method to assay stem cells is to follow their ability to repopulate conditioned recipients…” (Dick et al., 1997). HSCs can be identified either by limiting dilution or by single-cell transplantation assays that are based on the capacity of HSCs to hematopoietically reconstitute. For detection of human HSCs, immunosuppressed mice are typically used as recipients and their fate is followed up over 12 to 16 weeks (Ivanovic et al., 2011; Szilvassy et al., 1990; Till et al., 1964). A single HSC is sufficient for successful long-term hematopoietic reconstitution of an immunosuppressed mouse (Osawa et al., 1996).

The first indication that ROS are important in stem cells came from the work of Ito and co-workers (Ito et al., 2004). They demonstrated increased ROS levels in HSC-enriched populations after deletion of the ataxia telangiectasia mutated (ATM) gene in mice (Atm^−/− mice) (Ito et al., 2004). These results were obtained in a phenotypically defined cell population that was Lineage (Lin⁻, ‘L’) negative, stem cell antigen (Sca-1, ‘S’) positive, and c-kit (CD117, ‘K’) positive. This so-called LSK cell population contains at best ∼30% of ‘true’ HSCs or multipotent progenitors (MPPs) (Oguro et al., 2013). However, often <10% of LSK cells have real HSC potential based on in vivo long term-reconstitution capacity (Osawa et al., 1996); instead, the vast majority of these cells are multipotent and committed progenitors (Bryder et al., 2006). Furthermore, the LSK population can be subdivided into fractions based on the expression of SLAM family markers, CD150 (Slamf1) and CD48 (Slamf2): HSCs are particularly enriched in the CD150⁺CD48⁻LSK population, MPPs in the CD150⁻CD48⁻LSK population, whereas CD150⁻CD48⁺LSK and CD150⁺CD48⁺LSK cells contain heterogeneous committed progenitors (Kiel et al., 2005, 2008). The HSCs in the CD150⁺48⁻ population can be further divided into cells that do or do not express the CD34 marker. The most quiescent and potent HSCs were found in the CD34⁻CD150⁺CD48⁻ subpopulation (Wilson et al., 2008). Despite the availability of these markers to determine a high enrichment in HSCs and MPPs, the resulting populations remain functionally heterogeneous (Oguro et al., 2013).

In Atm^−/− mice, an increase in the levels of ROS coincides with the loss of HSC activity despite the maintenance of a phenotypically-defined LSK population (Ito et al., 2004). The underlying mechanisms were shown to involve ROS-mediated activation of p38 mitogen-activated protein kinases (MAPK11 and MAPK14), which upregulates the cyclin-dependent kinase inhibitors p16^Ink4a and p19^Arf. Upregulation of these inhibitors results in a consecutive block of cell-cycle-induced senescence and premature exhaustion of HSCs (Ito et al., 2006). Furthermore, Forkhead box class O (FoxO) transcription factors are crucial regulators of oxidative stress defense and ROS production (Storz, 2011). FoxO-deficient mice showed elevated levels of ROS associated with increased cycling and impaired long-term repopulating potential of LSK cells (Miyamoto et al., 2007; Tothova et al., 2007). FoxO-mediated redox regulation in HSCs is achieved partly through the regulation of ATM expression (Yalcin et al., 2008). Treatment of Atm^−/− or FoxO-deficient mice with N-acetyl cysteine (NAC), a scavenger of HOCl, •HO, and H₂O₂ (Aruoma et al., 1989), reverses the ROS-induced defects, suggesting these oxidants have roles in HSCs and HPCs. Recently, it has been shown that ATM is activated directly by oxidative stress, which – in turn – stimulates an anti-oxidative cell response (Ditch and Paull, 2012). One such anti-oxidative protective mechanism is ATM-mediated suppression of the pro-apoptotic protein BID, which is known to act as an inducer of mitochondrial ROS (Maryanovich et al., 2012). The tuberous sclerosis complex (TSC)-mammalian target of rapamycin (mTOR) pathway is a crucial regulator of cell metabolism (Wullschleger et al., 2006). In response to oxidative stress, ATM-dependent inhibition of mTOR maintains quiescence of HSCs by repressing ROS production and mitogenesis (Chen et al., 2008a; Ditch and Paull, 2012). In addition, ROS induce activation of p53, which triggers apoptosis or the expression of the anti-oxidative defense enzymes in LSK cells (Abbas et al., 2010). Hypoxia-inducible factor 1-alpha (HIF1α) has also been implicated in a negative feedback loop of ROS regulation (Honma et al., 2013; Simsek et al., 2010; Takubo et al., 2010); ROS-stabilized HIFα (Piccoli et al., 2007a) induces adaptive metabolic responses that ensure redox homeostasis (Cam et al., 2010), such as inhibition of mTOR signaling and mitochondrial activity (Cam et al., 2010), as well as activation of glycolytic metabolism (Takubo et al., 2013).

Several factors are known to coordinate the balance between stem cell self-renewal and commitment with mitochondrial function. These include members of the polycomb family of proteins, such as B lymphoma Mo-MLV insertion region 1 homolog (Bmi1) and PR-domain-containing 16 (Prdm-16), as well as regulators of the DNA damage response, for example apurinic/apyrimidinic (AP) endonuclease1/redox factor-1 and redox regulator nuclear factor erythroid-2-related factor 2 (Nrf2) (Alfadda and Sallam, 2012; Chuikov et al., 2010; Liu et al., 2009; Wang et al., 2013).

An increase in ROS levels in an LSK population is paralleled by a decline in HSC activity (as measured by reduction of short- and long-term BM-reconstitution capacity) (Ito et al., 2004; Maryanovich et al., 2012; Miyamoto et al., 2007). However, in view of the fact that the functional HSCs represent often <10% of the LSK population, it is not clear whether ROS levels increase in HSCs themselves. In fact, the exhaustion of HSCs that have been detected in these experiments might originate from a release of ROS by non-HSCs, which represent the vast majority of LSK cells. These released ROS could induce HSC differentiation, senescence and apoptosis and/or necrosis.

Despite progress in understanding their effects on regulation of HSC and HPC functions, the exact ROS- and RNS-mediated mechanisms and their downstream signaling components have not been elucidated. For instance, it is unknown how ROS and RNS mediate cell fate at the molecular level. To fully understand how ROS and RNS contribute to the HSC fate, it is essential to detect these species accurately with regard to their identity, localization, kinetics and quantity because different reactive oxygen and nitrogen species might have very different roles in cellular signaling. For instance, compared with other ROS species, H₂O₂ is relatively stable (with a cellular half-life of ∼1 ms and steady-state levels of 10⁻⁷ M), diffusible and has selective reactivity (D'Autréaux and Toledano, 2007). These characteristics make it a likely candidate for a signaling molecule. Therefore, it is important to accurately detect this specific molecule without any erroneous contributions of other ROS and RNS species. However, most currently utilized techniques for the detection of ROS and RNS in HSCs and HPCs are not specific for a particular species, and any obtained measurements can be affected by the presence of other ROS or RNS species. Below, we will discuss the limitations and caveats of the commonly used techniques for the analysis of the redox status of stem cells.

The probes most commonly used to detect ROS and RNS are oxidant-sensitive compounds; these probes are often cell-permeable lipophilic esters that undergo intracellular hydrolysis and are retained in the cell, where – upon oxidation – they become fluorescent. Dihydrodichlorofluorescein (DCFH₂), hydroethidine (HE), diaminofluorescein (DFA-2), Amplex Red as well as the chemiluminescence detectors, belong in this category (Wardman, 2007; Winterbourn, 2014) (Table 2).

These probes can be oxidized directly by various cellular radical species or by metal-dependent processes, which result in a radical-probe intermediate that is subsequently oxidized into the fluorescent product (Wardman, 2007). The radical-probe intermediate can react with various oxidants or can be captured and scavenged by cellular antioxidants. This means that oxidant-sensitive probes are generally non-specific for a particular oxidant, and their complex chemical behavior makes them unsuitable for mechanistic studies of cellular events that are induced by specific ROS and RNS species (Wardman, 2007; Winterbourn, 2014). Owing to the fact that a positive signal can be generated from non-specific reactions, the reactivity of oxidant-sensitive probes with particular ROS or RNS species should be confirmed by other techniques or by use of appropriate inhibitors (Zielonka et al., 2012b). However, their main advantage is that they are easy to use.

A common ROS probe is DCFH₂ or its widely used esterified cell-permeable diacetate (DCFH₂-DA); neither react directly with H₂O₂ to form the fluorescent product dichlorofluorescein (DCF) but require a peroxidase or redox-active metal catalyst (i.e. Fe²⁺). Indeed, it has been shown that the exposure to H₂O₂ alone is unable to oxidize DCFH₂ (Karlsson et al., 2010). This requirement might lead to misinterpretations of data obtained with this probe and this issue has been discussed in detail in a number of reviews (Kalyanaraman et al., 2012; Karlsson et al., 2010; Wardman, 2007; Winterbourn, 2014). For example, a negative signal could be obtained if a metal catalyst is absent, whereas increased oxidation of DCFH₂ in the presence of cytochrom c – which is released during apoptosis or under conditions of increased iron availability from the lysozyme – could be mistakenly associated with an increased oxidant production, as critically analyzed elsewhere (Burkitt and Wardman, 2001; Karlsson et al., 2010). Furthermore, oxidation of DCFH₂ is entirely non-specific and can be achieved by a number of different free radicals – such as derivates of ONOO⁻, tyrosine, thiols and HOCl as well as by light (Marchesi et al., 1999) – independently of H₂O₂ (Wardman, 2007; Wrona et al., 2005).

Thus, in contrast to what has been concluded in numerous reports with regard to a redox analysis of HSCs and HPCs (Ito et al., 2004, 2006; Miyamoto et al., 2007; Piccoli et al., 2007b, 2005) (Table 1), DCF fluorescence cannot be used as a direct measure of cellular H₂O₂ levels, nor as direct evidence that H₂O₂ is involved in a signaling event (Kalyanaraman et al., 2012; Karlsson et al., 2010; Wardman, 2007; Winterbourn, 2014).

DCFH₂ has also been used as a general indicator of the redox status of HSCs and HPCs (Abbas et al., 2010; Chen et al., 2008a; Chuikov et al., 2010; Fan et al., 2007; Hosokawa et al., 2007; Jang and Sharkis, 2007; Jung et al., 2013; Juntilla et al., 2010; Liu et al., 2009; Pazhanisamy et al., 2011; Piccoli et al., 2007a; Shao et al., 2014; Yahata et al., 2011) (Table 1). However, using DCFH₂ as a probe for the overall cellular redox status is highly unreliable owing to three limiting factors. First, oxidation of DCFH₂ to DCF involves the intermediate radical DCFH• or DCF•⁻, which – in the presence of O₂ – can generate O₂•⁻ (Wrona and Wardman, 2006). The subsequent dismutase-mediated oxidation of O₂•⁻ produces additional H₂O₂, resulting in an artificial amplification of the initial DCF-induced fluorescence signal (Folkes et al., 2009). Second, the efficiency of oxidation of the intermediate DCFH₂ probe radical into the final product depends on the environmental oxygen concentration (Wrona and Wardman, 2006). This issue is of particular importance when DCFH₂ is used as a probe to study the effect of the O₂ concentration on ROS generation in HSCs and HPCs (Fan et al., 2007; Hao et al., 2011). Third, cellular conditions may also alter the oxidation efficiency of DCFH₂. For example, a higher detection sensitivity of DCFH₂ in HSCs was reported in the presence of an inhibitor of the P-glycoprotein multidrug-resistant 1 (MDR1) pump (Piccoli et al., 2005); here, an apparent higher sensitivity of DCFH₂ might be attributed to reduced expulsion of the probes by MDR1. In this context, it is worth noting that MDR1 is highly expressed in primitive HSCs (defined as cells that can in vivo provide long-term bone-marrow hematopoiesis or that can in vitro act as long-term culture-initiating cells) (Scharenberg et al., 2002).

Hydroethidine (HE) is commonly used for the detection of intracellular O₂•⁻ in studies of HSCs and HPCs (Jung et al., 2013; Nogueira-Pedro et al., 2014; Tesio et al., 2011; Wang et al., 2010). Although O₂•⁻ can directly oxidize HE to its radical, the reaction is relatively slow and is more likely to be performed by other cellular oxidants present (Winterbourn, 2014). Once formed, the HE radical combines rapidly with O₂•⁻ to generate the specific product 2-hydroxyethidium (2-OH-E⁺). However, HE can also react with other oxidants (e.g. ONOO⁻, H₂O₂, •HO) to generate the non-specific product ethidium (E⁺) (Kalyanaraman et al., 2012; Wardman, 2007; Winterbourn, 2014). Although 2-OH-E⁺ and E⁺ have slightly different fluorescence spectra, they cannot be distinguished easily (Robinson et al., 2006). Therefore, a simple fluorescence assay performed with HE cannot provide sufficient specific information regarding intra- and extracellular O₂•⁻ formation. Because the fluorescence signal can be generated from the non-specific reaction, any specificity of the fluorescent signal for O₂•⁻ needs to be confirmed with ROS-scavenging enzymes and NOS inhibitors (Zielonka et al., 2012b).

In stem cell redox studies, mitochondrial O₂•⁻ has been measured by mitochondrial-targeted HE (Mito-HE or MitoSOX) (Cho et al., 2014; Hole et al., 2010; Taniguchi Ishikawa et al., 2012; The et al., 2013). MitoSOX reacts with O₂•⁻ in a similar manner as HE and, therefore, has the same limitations (Winterbourn, 2014). But as a positively charged molecule, MitoSOX has an increased affinity to react with O₂•⁻, making the reaction more specific than the HE reaction with O₂•⁻. Oxidation of MitoSOX by O₂•⁻ generates the specific product red-fluorescent 2-hydroxymitoethidium (2-OH-Mito-E⁺), whereas its reaction with other oxidants results in the non-specific Mito-E (Zielonka and Kalyanaraman, 2010). However, because of the overlap of their fluorescence spectra, MitoSOX cannot be utilized on its own as a reliable indicator of mitochondrial O₂•⁻ formation (Zielonka and Kalyanaraman, 2010).

The peroxidase substrate Amplex Red is used to detect H₂O₂ that has been extracellularly generated, as well as any that is produced inside the cell and has diffused out (Hole et al., 2010; Skinner et al., 2008; Thompson et al., 2008; Vlaski et al., 2014). Amplex Red can be oxidized to the fluorescent product resorufin by H₂O₂ or other electron oxidants (e.g. O₂•⁻ and NO•). However, the specificity of Amplex Red for H₂O₂ depends on the presence of sufficient amounts of horseradish peroxidase (HRP). In this case, HRP catalyzes the oxidation of its substrate Amplex Red by H₂O₂ to resorufin in two one-electron oxidation steps (Gorris and Walt, 2009). However, there are some limitations, in particular the light-induced reduction of resorufin in the presence of biological reducing agents, such as NAD(P)H or glutathione, which results in an artificial amplification of the fluorescent signal (Zhao et al., 2011). Thus, in light-protected condition, Amplex Red can be used to specifically measure the extracellular formation of H₂O₂ in the presence of HRP (Kalyanaraman et al., 2012).

The probe diaminofluorecscein (DAF-2) and its cell-permeable diacetate (DFA-2DA) are used to monitor the formation of cellular NO• based on the generation of the resulting green fluorescent product DFA-2T (Zielonka et al., 2012b). DAF-2 does not react directly with NO•, but with N₂O₃, which is generated upon NO• oxidation and can be attenuated in the presence of the ascorbic acid (Ye et al., 2004). This method for NO• detection has been used in hematological studies (Čokić et al., 2009; Gorbunov et al., 2000). However, under conditions that generate both NO• and O₂•⁻, it is not a suitable method to measure NO• selectively, because O₂•⁻ contributes to the oxidation of DAF-2 into fluorescent DFA-2T. Therefore, NOS inhibitors or ROS-scavengers should be included in the assay in order to discriminate oxidation that is mediated by O₂•⁻ from that mediated by NO• (Zielonka et al., 2012b).

Chemiluminescent assays have been used to detect ROS in HSCs and HPCs (Hole et al., 2010) due to their high sensitivity and the fact they only require a relatively small number of cells (Winterbourn, 2014). However, these assays have the same limitations as oxidant-sensitive fluorescent probes. For instance, an assay based on luminol is unable to discriminate among individual ROS species present in biological systems or provide any mechanistic information (Zielonka et al., 2013). The luminol radical that is generated in the initial, non-specific oxidation can react in numerous interactions that can influence the resulting overall chemiluminescence, as described in detail elsewhere (Wardman, 2007). Luciferin analogues are more specific for the chemiluminescent detection of extracellular O₂•⁻ (Teranishi, 2007), and this approach could, thus, be used to confirm the production of O₂•⁻ that has been detected by using an alternative procedure.

Cell-permeable boronate probes contain a blocking group that is attached to the fluorophore. First, an oxidant reacts with and oxidizes the blocking group without changing the oxidation state of the fluorophore. The oxidized form of the blocking group is then released, resulting in fluorescence. These probes are thus referred to as ‘non-redox’ probes. This mechanism avoids several of the problematic steps that affect the probes discussed above, such as oxidation of the probe, generation of probe-radical intermediates and complex chemical reactions (Winterbourn, 2014). The first application of a non-redox probe was the detection of H₂O₂ in living cells (Dickinson et al., 2013; Lin et al., 2013; Lippert et al., 2011; Woolley et al., 2012). It was later shown that these probes (particularly in the case of aromatic boronate probes) react with ONOO⁻ much faster than with H₂O₂ (about a million-fold) (Sikora et al., 2009). Thus, the use of the ROS-scavenging enzymes [catalase, polyethylene glycol (PEG) catalase, SOD or PEG-SOD] and a NOS inhibitor (N^ω-nitro-L-arginine methyl ester), is necessary to distinguish between ONOO⁻-mediated and H₂O₂-mediated-fluorescence signals (Zielonka et al., 2012a).

Altogether, the studies presented above suggest that currently used techniques for ROS and RNS detection are not sufficiently specific. In addition, with exception of the boronate probe, they cannot be used in vivo. In light of these limitations, which are all related to the complexity of cellular redox chemistry, a better approach to examine ROS and RNS in vitro and in vivo – and specifically in HSCs as opposed to HPC systems – is needed.

A possible assay that could be applied to stem cell biology came from a recent study suggesting that fluorescence-based techniques using HE, boronate probes, DAF-DA, can be used for initial screening before – in a second step – specificity of these probes for certain reactive species is confirmed with NOS inhibitors and ROS-detoxifying enzymes (Zielonka et al., 2012b). It should be noted that there is always uncertainty as to whether the ROS scavengers reach the site of oxidation. Because all these probes allow analysis at single-cell level by flow cytometry, we propose the following procedure to measure ROS and RNS in LSK cell populations that contain HSCs and HPCs (Fig. 2A). A cell population is assayed by using any of these probes (step 1) and specificity of fluorescence signal confirmed before cells are sorted into subpopulations on the basis whether or not they respond to a particular ROS or RNS probe (‘ROS/RNS low’ or ‘ROS/RNS high’) (step 2). The following step 3 could then include the sorting of individual cells by their respective phenotypic properties. In the examples shown in Fig. 2A, individual cells within the resulting subpopulations (i.e. ROS/RNS low and ROS/RNS high) are sorted according to whether they express the SLAM family phenotypic markers, CD150 and CD48. The CD150⁺CD48⁻LSK population that is enriched in HSCs can so be distinguished from a population that contains mainly MPPs (CD150⁻CD48⁻LSK) or HPCs (CD150⁺CD48⁺LSKs) (Oguro et al., 2013). This step could then be followed by the functional analysis of the resulting single cells in order to determine whether they are true HSCs because this cannot be determined with certainty on the basis of surface markers analysis alone (step 4). Such a functional analysis could involve the testing of single cells for their capacity for hematopoietic reconstitution or their proliferative potential (Fig. 2A). This type of approach would not only help to overcome the problem of functional heterogeneity in phenotypically homogeneous HSCs/HPCs populations (such as LSK cells), but also allow to determine the ROS and/or RNS profiles of functional HSC subpopulations.

Furthermore, there are other approaches to determine the redox state of HSCs compared with that of HPCs in their physiological environment. A major breakthrough of ROS detection in vivo within cellular and subcellular compartments, as well as in vitro, was the development of genetically encoded fluorescent-protein-based redox sensors that are based on the introduction of cysteines into yellow or green fluorescent protein (rxYFP or roGFP) (Dooley et al., 2004; Østergaard et al., 2001). Signal detection is based on the change in the chromophore spectrum upon oxidation or reduction of these cysteines (Lukyanov and Belousov, 2014). The extent of the redox change can be estimated from the relative fluorescence intensity of the two fluorescence excitation maxima of the sensor, making the measurement ratiometric (Lukyanov and Belousov, 2014). These sensors enable specific, real-time, reversible, ratiometric detection of the changes in the redox state (Lukyanov and Belousov, 2014). In addition, these probes enable the specific detection and quantification of H₂O₂ or of the glutathione redox potential (GSH:GSSG ratio) at the single-cell level in vitro (Shimi et al., 2011) and in vivo (Albrecht et al., 2011; Kojer et al., 2012). For instance, using roGTP transgenic flies, this approach has enabled the in vivo mapping of H₂O₂ and of oxidized glutathione during development and aging (Albrecht et al., 2011). Another H₂O₂-sensitive sensor, Hyper, was developed by inserting the circularly permuted YFP (cpYFP) into the regulatory domain of the E. coli H₂O₂-sensing protein OxyR (Choi et al., 2001). This probe enabled in vivo imaging of H₂O₂ gradients at tissue-scale as well as their quantification (Bilan et al., 2013; Love et al., 2013). Furthermore, this method was used to follow redox changes specifically in Hyper-expressing neutrophils in vivo (Pase et al., 2012).

By using these genetically modified approaches, it should be possible to measure the redox state of a HSC in situ. For instance, a mouse expressing roGFP or Hyper could be generated, from which HSC-enriched subpopulations are obtained according to the flow cytometry-based sorting procedure outlined above in order to determine and analyze their ROS levels (Fig. 2B). Alternatively, boronate-based fluorescent probes that enable measurement of real-time changes in H₂O₂ in tissue, cells and subcellular compartments could be used to assess the HSC redox status in situ (Cochemé et al., 2011; Dickinson et al., 2011; Van de Bittner et al., 2010). Here, boronate probes could be administered to mice via a tail-vein injection before their bone marrow cells are extracted, analyzed and sorted by flow cytometry with respect to their redox status and surface markers (Fig. 2B).

Other promising tools for the ROS and RNS detection in HSCs are fluorescence resonance energy transfer (FRET)-based redox sensors. In FRET, energy is transferred from a light-excited fluorophore (donor) to a longer-wavelength fluorophore (acceptor), whose absorption spectrum overlaps with the emission spectrum of the donor (described in detail by Lukyanov and Belousov, 2014). The main advantage of FRET-based probes lies in the fact that the ratio between the two wavelengths is measured. Unfortunately, these sensors exhibit a low dynamic range and are pH sensitive. Nevertheless, FRET probes have proven very useful and can also be used to probe subcellular compartments or to measure NADPH-oxidase-generated H₂O₂ (Enyedi et al., 2013; Kolossov et al., 2008). However, to date, FRET-based redox sensors have not been used in vivo.

It is of particular interest to measure ROS levels (e.g. H₂O₂) under physiological or stress conditions in a hematopoietic niche, as this would help to shed light on the relationship between the state of stem cells and the microenvironment in which they reside. This has been recently achieved for the measurement of the absolute pO₂ concentration in the bone marrow hematopoietic niche by using a metallo-porphyrin-based two-photon-enhanced platinum-porphyrin phosphorescent oxygen-sensitive probe (PtP-C343) (Spencer et al., 2014).

Genetically encoded fluorescent probes or boronate-based fluorescent probes have been used to detect H₂O₂ in various tissues (Bilan et al., 2013; Logan et al., 2014; Love et al., 2013) but not in bone marrow. We believe that their use for H₂O₂ detection in bone marrow should be considered. Another possibility is the use of nanoparticle-based sensors, which have shown to be useful for ROS detection in whole-animal studies (Uusitalo and Hempel, 2012). Nanoparticles can be conjugated with different moieties for their targeting to certain cells or subcellular compartments (Uusitalo and Hempel, 2012). Recently, intravenous injection of specially designed, liver-targeting semiconducting polymer nanoparticles enabled the simultaneous detection of fluorescence and chemiluminescence in real-time to discriminate between ONOO⁻ and H₂O₂ generation (Shuhendler et al., 2014), making this a particular promising approach for the detection of specific ROS in HSCs. However, further studies are needed to determine whether these probes can be used in HSCs.

In view of the data demonstrating that an anaerobic metabolic character is related to the maintenance of stemness, methods to accurately measure the levels of ROS and RNS in stem cells are of paramount importance. These highly reactive molecules have regulatory roles in energy metabolism that might impact on the HSC fate. Current evidence indicates that variations in the levels of ROS and RNS affect the balance between HSC self-renewal and commitment. As discussed above for commonly used techniques for the measurement of reactive species, there are clear imprecisions in the currently used approaches that must be overcome in order to help us to better understand how ROS and RNS impact on HSC biology. We suggest here that, by combining new advances in the detection of specific ROS and RNS with protocols that allow to achieve a higher enrichment of HSCs, it will be possible to confirm and elucidate any of functional roles of ROS and RNS in stem cell maintenance.

We are thankful to Mrs Elisabeth Doutreloux-Volkmann (EFS-Aqli, Bordeaux), Dr Jacqueline R. Wyatt (Wyatt Scientific Editing Service) and Mr Jerold Morgan (Biorad Laboratories, Seattle) for the language corrections.