Transmembrane protein western blotting: Impact of sample preparation on detection of SLC11A2 (DMT1) and SLC40A1 (ferroportin)

Western blotting has been widely used for investigation of protein expression, posttranslational modifications, and interactions. Because western blotting usually involves heat-denaturation of samples prior to gel loading, clarification of detailed procedures for sample preparation have been omitted or neglected in many publications. We show here the case that even excellent primary antibodies failed to detect a specific protein of interest due to a routine heating practice of protein samples. We performed western blotting for transmembrane iron transporter proteins; SLC11A2 (divalent metal transporter 1, DMT1), SLC40A1 (ferroportin 1, Fpn1), and transferrin receptor-1 (TfR1), along with cytoplasmic iron storage protein ferritin H. Our results in 12 human culture cell lysates indicated that only unheated samples prior to gel loading gave rise to clear resolution of DMT1 protein, while heated samples (95°C, 5min) caused the loss of resolution due to DMT1 protein aggregates. Unheated samples also resulted in better resolution for Fpn1 and TfR1 western blots. Conversely, only heated samples allowed to detect ferritin H, otherwise ferritin polymers failed to get into the gel. Neither different lysis/sample loading buffers nor sonication improved the resolution of DMT1 and Fpn1 western blots. Thus, heating samples most critically affected the outcome of western blotting, suggesting the similar cases for thousands of other transmembrane and heat-sensitive proteins.


Introduction
More than 3,000 human genes encode transmembrane proteins that have been classified into receptors, transporters, enzymes, and other proteins [1]. Transmembrane proteins such as solute carrier (SLC) transporters play vital roles in import and export of nutrients, metabolites, and ions. Examples of SLC substrates are sugars, amino acids, vitamins, neurotransmitters, urea cycle metabolites, and metals [2,3] including iron [4]. The human SLC transporter families comprise more than 400 genes and 52 subfamilies under the SLC family clusters, including MFS (major facilitator superfamily) and APC (amino acid-polyamine-organocation) superfamily [3,5,6]. The majority of SLC transporter proteins has 12-transmembrane segment (TMS) with diversities typically ranging from 9 to 14 TMS [7].

Chemicals, and reagents
All chemicals and reagents used in this study, their suppliers, and catalog numbers are listed in Table 1. Ferric ammonium citrate (FAC) and desferrioxamine (DFO) were dissolved in distilled water at 100mM and 25mM, respectively. Hemin was dissolved in 0.1M NaOH at 50mM. Human bioactive hepcidin was dissolved in water at 0.5mM, added to undifferentiated Caco-2 culture at 0.5uM, and incubated for 18-24 hr.
The 2.5kb DMT1 cDNA was isolated after digestion with NcoI and HindIII followed by filling-in with DNA pol. Klenow fragment, and cloned into SmaI site of pCMV and pCMVFlag vectors. These plasmid DNAs, verified by DNA sequencing (Eurofins), were transiently transfected into HEK293 cells by electroporation (X-Cell, Bio-Rad) as described previously [36]. Human Fpn1 siRNA was purchased from Sigma-Aldrich (GAUGGAACUUGGUAUCCAU[dT] [dT], and AUGGAUACCAAGUUCCAUC[dT][dT]) and 5ul of 20uM solution was transfected into Caco-2 cells (grown in 2ml media/35 mm plate) with 200ul of Opti-MEM containing 6ul of Lipofectamine RNAiMax (Invitrogen) for 18hr, followed by treatment with 250 and 500uM FAC for additional 8hr in the presence of siRNA prior to harvest. Knockdown of Fpn1 in Caco-2 cells was first verified by qPCR as described previously [37] using three sets of human Fpn1 primers.

Routine practice of heating samples in Fpn1 (SLC40A1) western blot
According to the citation history for Fpn1 antibody in CiteAb, the rabbit polyclonal anti-Fpn1 antibody NBP1-21502 (Novus Biologicals) has been most successfully used in research article publications (https://www.citeab.com/). After preliminary testing of optimum dilutions of this primary antibody with either 5% skim milk or 1% BSA in Tween/TBS buffer, we decided to use 1,000-1,500-fold dilution in 1% BSA in Tween/TBS. We prepared WCL samples from several culture cell lines, heated at 95˚C for 5min, loaded on a 10% acrylamide SDS-PAGE gel, and performed western blotting. The blot gave rise to multiple bands including the putative Fpn1 band � (63 kDa) between 50 and 75 kDa size markers (Fig 1A).
The expression of Fpn1 can be increased by high iron levels through dissociation of IRE-binding protein 1 and 2 (IRP1 and IRP2) from the iron-responsive element (IRE) in 5'-UTR of Fpn1 mRNA [22]. Additionally, another Fpn1 mRNA (termed FPN1B) transcribed from an alternative promoter lacks the IRE but has the same open reading frame, therefore FPN1B expression is not responsive to cellular iron status [38]. To characterize the putative Fpn1 band further, we treated SW480 cells with ferric ammonium citrate (FAC) or cisplatin that directly inhibits IRP2 binding to IRE [39] and performed western blotting. The putative 63 kDa faint band was not increased by FAC but slightly increased by cisplatin treatment in SW480 cells (Fig 1B). In contrast, ferritin H was significantly increased by FAC and cisplatin because ferritin H mRNA also contains a functional 5'-IRE [40,41], suggesting that FAC and cisplatin treatments are working to drive 5'-IREdependent translational upregulation. Collectively, we demonstrate that the most popular anti-Fpn1 antibody gave rise to the band around 63 kDa but it did not show clear regulation through the IRP-IRE system in SW480 cells. Thus, without taking a positive control for western blotting, the identity of the 63 kDa band in Fig 1 remained undetermined.

No solubilization problem but heating samples may compromise Fpn1 western blot
To validate the putative Fpn1 band and anti-Fpn1 antibody further, we made WCL from HEK293 cells transiently transfected with a human Fpn1 expression plasmid and performed western blotting in a routine sample preparation and heating protocol (see Materials and Methods). The overexpressed Fpn1 (untagged) gave rise to one specific band that was stuck on the top of the separation gel (Fig 2A).
This result indicates that the Fpn1 antibody (NBP1-21502, Novus Biologicals) can detect at least overexpressed human Fpn1 protein by western blotting, in which the Fpn1 protein appeared to be insolubilized or aggregated. To solve the potential solubilization problem first, we used three different cell lysis conditions; RIPA lysis buffer, RIPA plus sonication, and This experiment used sample loading buffer containing 5% β-mercaptoethanol, heated at 95˚C for 5 min, anti-Fpn1 antibody (Novus, rabbit polyclonal, 1,000-fold diluted with 1% BSA in T/TBS) at 4˚C overnight followed by anti-rabbit IgG-HRP conjugated (8,000-fold diluted with 1% BSA in T/TBS) at room temperature for 1.5hr. Unless otherwise noted, the rest of western blots in this study were performed in the same procedures with or without sample heating prior to gel loading. B) SW480 cells were treated with 0.1-0.5mM FAC (ferric ammonium citrate), 2-100ug/ml cisplatin, or FAC 0.25mM plus cisplatin 10ug/ml for 18hr. 20ug of WCLs were subjected to Western blotting with anti-Fpn1 antibody (top). The membrane was soaked in the stripping solution at room temperature for 1hr, washed with T/TBS, and incubated with anti-ferritin H mouse monoclonal antibody (sc-135667, Santa Cruz Biotechnology). In A) and B), the asterisks indicate the predicted position (63 kDa) of the endogenous Fpn1. The experiments were repeated three times and the representative western blots are shown. transmembrane extraction reagent (Five Photon Biochemicals). To solve the aggregation prob-three different lysis conditions, transfected Fpn1 was stuck on the top of the gel when they were heated at 95˚C prior to gel loading. In contrast, no heating samples gave rise to a specific Fpn1 band between 50 and 75kDa markers regardless of their different lysis conditions ( Fig  2B). Furthermore, HEK293 WCL in RIPA mixed with three different 2X SDSPAGE sample loading buffers containing 5% β-mercaptoethanol (the regular sample loading buffer, see Materials and Methods), 4M urea, or both 5% β-mercaptoethanol and 4M urea did not improve the migration problem of overexpressed Fpn1 protein when they were heated ( Fig  2C). By contrast, no heating of these WCL samples showed the right size of Fpn1 band regardless of the use of different loading buffers (Fig 2C). Taken together, we concluded that no heating of protein samples is the most critical factor for the right resolution and detection of transfected Fpn1 protein by western blotting with this anti-Fpn1 polyclonal antibody.

Heating samples compromised DMT1 (SLC11A2) western blot
Another key SLC protein handling cellular iron traffic is DMT1 (SLC11A2). According to the citation data in CiteAb, there are more than 10 different anti-DMT1 antibodies used for publications. Among them, we used a rabbit monoclonal anti-DMT1 antibody (D3V8G, Cell Signaling Technology). To test whether heating protein samples also affects the resolution of DMT1 western blots, we prepared WCLs in RIPA or cytoplasmic and nuclear fractions from HEK293 cells transiently transfected with human DMT1 (untagged) or Flag-DMT1. Samples were mixed with 2x SDSPAGE sample loading buffer (containing 5% β-mercaptoethanol), heated at 95˚C for 5min or left at room temperature, and performed SDS-PAGE and western blotting with anti-DMT1 or Flag antibody. As observed in the Fpn1 western blots in Fig 2, heating protein samples significantly impaired the DMT1 protein bands (Fig 3A). In clear contrast, no heating of samples (left at room temperature until sample loading) showed good resolution of~75kDa and additional larger and smaller bands detected with anti-Flag (Fig 3A, top) as well as anti-DMT1 antibodies (Fig 3A, bottom). DMT1 mRNA isoforms contain 3'-UTR IRE [42] that should be subject to mRNA destabilization when IRP1 and/or IRP2 dissociates from the IRE in iron overloaded conditions or by inactivation of IRP2 following cisplatin treatment [39]. Indeed, endogenous DMT1 proteins were detected by western blotting in unheated WCLs isolated from SW480 cells treated with FAC (0.1-0.5 mM) or cisplatin (2-100 ug/ml), in which FAC or cisplatin treatment slightly decreased DMT1 protein levels ( Fig 3B). Collectively, no heating of protein samples prior to loading to SDS-PAGE gel is critical for detection of transfected and endogenous DMT1 proteins by western blotting.

Effects of heating samples from 12 human cell lines on DMT1 and Fpn1 western blots
To assess the effect of heating samples on detection of endogenous DMT1 and Fpn1 proteins in various cell types by western blotting, we prepared WCLs from 12 human cell lines (lysed in RIPA), mixed with 2x SDSPAGE sample loading buffer, heated at 95˚C for 5min or not heated (left at room temperature), and performed SDSPAGE and Western blotting. Given the results of transfected Fpn1 and DMT1 in Figs 2 and 3, only unheated WCLs from HEK293 cells transfected with empty vector, DMT1, and Fpn1 were loaded together as positive controls of western blots. As shown in Fig 4A, endogenous DMT1 proteins were detected in most of these human cell lines; however, heating samples at 95˚C for 5min reproducibly impaired the major DMT1 band, some of which were stuck on the top of the gel (arrow on the right). Fpn1 western blots were more complicated because of multiple major bands in addition to no clear difference between heated and unheated samples (Fig 4B). This was an unexpected result because transiently transfected Fpn1 protein in HEK293 cells was detected only when the samples were not heated prior to loading to SDSPAGE gel (Fig 2). The Fpn1 western blotting was working in Fig 4B because transiently transfected Fpn1 used as a positive control was detected in the range between 50 and 75kDa size markers (Fig 4B). The endogenous protein showing the same migration as transfected Fpn1 was also detected in several cell types such as Caco-2 cells ( Fig 4B).
As our Fpn1 western blots gave rise to multiple non-specific bands, we further verified the endogenous band migrating together with transfected Fpn1. First, transfection of Fpn1 siRNA into undifferentiated Caco-2 cells reduced Fpn1 mRNA level to 35.3% (p = 0.007 in unpaired t-test, FAC 0) and consistently diminished the endogenous protein band co-migrating with transfected Fpn1 (Fig 5A). Second, as hepcidin binds Fpn1 and induces Fpn1 internalization and degradation [43] in Caco-2 cells [44], we treated Caco-2 cells (3 plates each) with bioactive hepcidin at 500 ng/ml for 24hr and performed Fpn1 western blotting, in which hepcidintreated cells appeared to have decreased Fpn1 protein levels ( Fig 5B). Collectively, we concluded that the band co-migrating with transfected Fpn1 is the endogenous Fpn1 protein.

Effects of heating samples for single-transmembrane TfR1 and cytoplasmic ferritin western blots
We elucidated the effect of sample heating on western blots for two other key iron transport and storage proteins; single-transmembrane TfR1 and cytoplasmic ferritin. We prepared WCL in RIPA from Caco-2 cells treated with hemin, FAC, and an iron chelator deferoxamine (DFO) for 24hr. We compared heated samples with unheated samples for detection of each endogenous protein by western blotting. In Fpn1 western blotting, transiently transfected Fpn1 in HEK293 was reproducibly detected in unheated samples, whereas heated samples failed to show the right size of Fpn1 band between 50-75kDa protein size markers, rather showing very faint bands stuck on the top of the gel (arrow, Fig 6A right). FAC or hemin treatment in Caco-2 cells increased a band migrating similar to transfected Fpn1, in which the unheated group of samples showed slightly stronger bands compared to heated samples at 95˚C for 5min (Fig 6A). Conversely, we anticipated that the iron chelator DFO would inhibit translation of Fpn1 mRNA via the 5'-UTR IRE; however, we repeatedly failed to observe it (Fig 5A, DFO compared to none). The endogenous DMT1 was consistently detected only when samples were not heated (Fig 6B). In contrast, the H subunit of ferritin (ferritin H) was detected as increased by hemin or FAC treatment only in heated samples ( Fig 6C). As ferritin is composed of 24 subunits of ferritin H and L, heat denaturation of samples is necessary for ferritin western blot otherwise we often observed no ferritin bands or ferritin stuck on the top of the gel (S1 Fig). Interestingly, western blot for an 85 KDa homodimer of a plasma single-transmembrane protein TfR1 showed that unheated samples gave much stronger bands in an iron-responsive manner (iron-mediated downregulation) than samples heated at 95˚C for 5min (Fig 6D). The unheated samples gave rise to an additional higher molecular weight band, probably a dimer of TfR1, in an iron-responsive manner (asterisk in Fig 6D). The cytoplasmic protein GAPDH as a protein loading control was detected equally in both heated and unheated samples ( Fig  6E). Similar results were observed in HepG2 WCLs except for the endogenous

Discussion
Western blotting techniques were improved and seemingly have established, which may make us often omit in publications describing critical steps and tips for detection of protein of interest in western blotting. This is partly because a standard protocol for sample preparation has a routine protein denaturation step by heating at 95-100˚C in a sample loading buffer containing SDS and a reducing agent such as β-mercaptoethanol or dithiothreitol. This heating step is understood important for antibodies to find epitopes efficiently as well as to detect a specific protein in multiple protein binding complexes or a subunit of multimeric proteins as exemplified by the detection of ferritin in this study (Fig 6C). However, detection of transmembrane proteins by western blotting seems to be more challenging and complicated. There are more than 3,000 human genes encoding transmembrane proteins, in which~1,300 receptors including seven-transmembrane GPCR (G protein coupled receptor) and single-pass transmembrane receptors,~800 transporters such as SLC, ABC (ATP binding cassette) transporters, and channels, and~500 enzymes including transferases, hydrolases, and oxidoreductases  [1]. We sometimes find researcher's troubleshooting discussions via internet medias about unsuccessful western blotting for detection of particular transmembrane proteins. Along with our initial failure to detect SLC family iron transporters in this study, we anticipate that routine heat denaturing protocols may not work or significantly loose band signals in western blotting to detect many more transmembrane and heat-sensitive proteins, perhaps due to heat-induced protein aggregations and/or conformational changes.
In this study, we observed that non-heated samples solely solved the migration and resolution problems of DMT1 and overexpressed Fpn1 western blots, whereas three different cell lysis conditions or three different 2X sample loading buffers did not improve the situation (Fig  2). Despite the fact that both DMT1 and Fpn1 are 12-transmembrane proteins [13,15,28,29], heating samples at 95˚C for 5min ruined DMT1 western blot (Figs 3, 4A and 6B) whereas only partially impaired endogenous Fpn1 western blot (Figs 4B and 6A). However, heating samples completely impaired the migration of transfected Fpn1 (untagged) being stuck on the top of the separation gel (Fig 2A). We do not have good explanations for these different outcomes in sample heating between DMT1 and Fpn1 12-TMS proteins and between transfected and endogenous Fpn1 proteins. However, these results alert us to be aware of potentially different effects of sample heating on western blotting of any other SLC transporters and transmembrane proteins. It is therefore important to find the optimum condition of sample preparation for each transmembrane protein western blotting. This was also the case in TfR1 western blotting, in which non-heated samples tended to give rise to stronger and consistent iron statusdependent TfR1 band despite showing the additional TfR1 dimer band (Fig 6D, S2 Fig). In contrast, heating samples was essential for detection of cytoplasmic ferritin H subunit protein in western blotting because ferritin is assembled of 24 subunits of H and L proteins [11]. These results suggest that a single western blot loaded only either heated or non-heated protein samples cannot be used for successive reprobing with DMT1, Fpn1, TfR1, and ferritin antibodies. Another cytoplasmic protein GAPDH, often used as a protein loading control of western blotting, was not affected by heated or unheated protein samples. It should be noted that, although we observed no improvement for Fpn1 western blotting by switching from RIPA to RIPA plus sonication or commercially available membrane protein extraction lysis buffer (Fig 2A), the lysis buffer components together with mechanical destruction of membranes may also significantly improve western blotting for some other transmembrane proteins more than the impact of heating protein samples.
It may be a routine practice for laboratories and researchers primarily working on transmembrane proteins to avoid heat-denaturing protein samples prior to gel loading. Alternatively, they may have additional technical tips for particular transmembrane protein western blotting that we have not tested in this study. Indeed, no sample heating prior to gel loading for Fpn1 western blotting was noted in some earlier publications [22,26,28], while we observed relatively few attempts of DMT1 western blotting in earlier publications or no particular notes for western sample preparations except examples such as pre-incubation at 37˚C for 15min by Nam et al [45]. CiteAb records more than 50 citations for several applications with the rabbit polyclonal anti-Fpn1 antibody (NBP1-21502, Novus Biologicals) that we used in this study, in which more than 30 publications used this antibody for western blotting. Most of them did not clarify sample heating conditions or heated at 90-100˚C, except 75˚C for 10min [46] or no heating [39]. Even if protein samples might have been preheated in most of these publications, they might have observed the right size of the endogenous Fpn1 band as we observed in this study (Figs 4B and 6A). CiteAb also lists more than 10 publications for DMT1 western blotting using Proteintech, Abcam, or Novus Biologicals anti-DMT1 antibody, only some of which noted that they treated samples with gentle heating at 37˚C for 30min [47][48][49] or 50˚C for 15min [50]. Anti-DMT1 antibodies that have different epitopes from our anti-DMT1 antibody (Cell Signaling Technology, near the N-terminus of human DMT1) may also affect the optimum conditions of protein sample preparation for better resolution of western blotting.

Conclusion
In summary, only non-heated protein samples worked for western blotting with the anti-DMT1 (SLC11A2) monoclonal antibody having the epitope near the N-terminus. Non-heated protein samples gave rise to better resolution and/or stronger band signals in western blotting with anti-Fpn1 (SLC40A1) and anti-TfR1 polyclonal antibodies. Three different lysis buffers, three sample loading buffers, or sample sonication we tested did not improve the resolution of DMT1 and Fpn1 western blots. In contrast, only heated protein samples worked for ferritin western blotting with the anti-ferritin H monoclonal antibody, while sample heating did not affect western blotting for another cytoplasmic protein GAPDH frequently used for a protein loading control. Collectively, either heated or unheated samples most critically determined the quality and outcome of transmembrane protein western blotting we tested, suggesting that there should be many similar cases for western blotting of other transmembrane and heat-sensitive proteins.