Objective: The voltage-gated K+ channel KCNQ1 associates with the small KCNE1 β subunit to underlie the IKs repolarizing current in the heart. Based on sequence homology, the KCNE family is recognized to comprise five members. Controversial data have indicated their participation in several K+ channel protein complexes, including KCNQ1. The expression level and the putative functions of the different KCNE subunits in the human heart still require further investigation.
Methods: We have carried out a comparative study of all KCNE subunits with KCNQ1 using the patch-clamp technique in mammalian cells. Real-time RT-PCR absolute quantification was performed on human atrial and ventricular tissue.
Results: While KCNQ1/KCNE1 heteromultimer reached high current density with slow gating kinetics and pronounced voltage dependence, KCNQ1/KCNE2 and KCNQ1/KCNE3 complexes produced instantaneous voltage-independent currents with low and high current density, respectively. Co-expression of KCNE4 or KCNE5 with KCNQ1 induced small currents in the physiological range of voltages, with kinetics similar to those of the KCNQ1/KCNE1 complex. However, co-expression of these inhibitory subunits with a disease-associated mutation (S140G-KCNQ1) led to currents that were almost undistinguishable from the KCNQ1/KCNE1 canonical complex. Absolute cDNA quantification revealed a relatively homogeneous distribution of each transcript, except for KCNE4, inside left atria and endo- and epicardia of left ventricular wall with the following abundance: KCNQ1 KCNE4 ≥ KCNE1>KCNE3>KCNE2>KCNE5. KCNE4 expression was twice as high in atrium compared to ventricle.
Conclusions: Our data show that KCNQ1 forms a channel complex with 5 KCNE subunits in a specific manner but only interactions with KCNE1, KCNE2, and KCNE3 may have physiological relevance in the human heart.
Voltage-gated K+ channels associate with at least three classes of auxiliary subunits, namely Kvβ, KCHIP (K-CHannel Interacting Protein) and KCNE. The Kvβ proteins assemble with many Kv channel α-subunits as cytoplasmic tetramers. They have been shown to alter voltage dependence, kinetics, inactivation and deactivation of the α-subunits. They are also involved in trafficking and chaperone behavior over the α-subunits . The KCHIP soluble peptides bind to N-terminus intracellular domain of the Kv channels to modulate channel presentation, achieving hence their chaperone-like tasks [2,3]. The single membrane-spanning domain KCNE peptides may interact with many Kv channels. KCNE1 was first shown to co-assemble with KCNQ1 to form the delayed rectifier K+ channel in the heart [4,5]. KCNQ1 protein is a Shaker type voltage-gated K+ channel that consists of six membrane-spanning domains, a pore loop, and two intracellular domains. While KCNQ1 mutations have been linked to human long QT (LQT1) syndrome  that can be associated with deafness in Jervell and Lange-Nielsen syndrome (JLN) , and to familial atrial fibrillation (AF) , KCNE1 protein has been implicated in LQT and JLN [9,10]. Cardiac rhythm abnormalities and deafness have also been observed in both kcnq1 and kcne1 knockout mice [11–15].
Data base mining has led to the identification of KCNE1-related genes in mammalian genome, encoding the KCNE-related peptides called MiRP1 to 3 (for MinK-related peptide 1 to 3) . The KCNE2 (MiRP1) protein is thought to co-assemble with the α-subunit human ether-à-go-go-related gene (HERG) to form the rapid delayed rectifier K+ channel responsible for IKr. Loss-of-function mutations in HERG cause LQT2 cardiac arrhythmia, while mutations in KCNE2 have been found in individual cases of inherited LQT . Allelic variants of KCNE2 may also contribute to drug-induced LQT syndrome . Recently, we identified a novel mutation in KCNE2 gene that was associated with AF . KCNE3 (MiRP2) was independently cloned by Schroeder et al. as a KCNQ1-associated subunit to form the basolateral K+ current in secretory epithelial cells . A mutation in the human KCNE3 gene causes periodic paralysis [20,21]. An association with Kv3.4 as pore-forming subunit was suggested to take place in the skeletal muscle. To date, KCNE4 has not been linked to any genetic or acquired disorder, however heterologous expression of this subunit produced a down-regulation of KCNQ1 as well as Kv1.1 and Kv1.3 channels [22,23]. Finally, KCNE5 was shown to produce similar effects than KCNE4 on KCNQ1 function and is also thought to associate with Alport syndrome, mental retardation, midface hypoplasia, and ellipticytosis (AMME) . However, the deleted region on chromosome X contains other candidate genes.
The distribution pattern of KCNQ1 and KCNE subunits does not always overlap. In the heart, although KCNQ1 and KCNE1 expression is well documented, it is still unclear how this expression takes place in different cardiac compartments. It is not clear either how many distinct KCNE subunits are expressed and at which extent. Interactions of KCNQ1 with KCNE subunits have been reported in a controversial manner. The aim of this study is to systematically screen for interactions between KCNQ1 and all KCNE subunits in the same system. To help identify which potential KCNQ1/KCNE interaction may take place in the human heart, we conducted absolute quantitative measurements of gene expression of all subunits in left atrium and in left ventricular endo- and epicardium. Given the abundance of KCNE1 and KCNE4 mRNA, we have investigated the possibility of the formation of a complex made of KCNQ1 and both auxiliary subunits.
COS-7 cells were transiently transfected by DEAE-Dextran precipitate method using 1.5 μg pCI-WT-KCNQ1 (or S140G-KCNQ1, or R243H-KCNQ1 mutants) DNA and 1.5 μg pIRES-KCNE(1–3)-CD8 or pXOOM-KCNE(4, 5)-GFP DNA per 35-mm culture dish. The pCI, pIRES and pXOOM plasmids drive cDNA expression under the control of CMV promoters. Currents were recorded 48 h after transfection in the whole-cell configuration as described . Deactivating current traces were fitted using a one a0+a1*exp(−t/τ1), or two-exponential function: I(t)=a0+a1*exp(− t/τ1)+a2*exp(− t/τ2), where I(t) is the current, a0, a1, and a2 current amplitude, t is the time, τ1 and τ2 are the time constants of deactivation.
2.2. Human cardiac samples
Healthy human hearts were obtained from organ donors whose hearts were explanted to obtain pulmonary and aortic valves for transplant surgery. These experiments complied with the Helsinki Declaration of the World Medical Association and were approved by the Albert Szent-Gyorgyi Medical University Ethical Review Board. Left ventricular wall samples were obtained from the base. Left epicardial and endocardial tissues were obtained by cutting 1-mm-thick slices from the epicardial and endocardial surfaces, respectively. Samples from each region matched one donor. Six to ten donors were investigated. Then, the tissues were fast-frozen in liquid nitrogen. The donors did not show any sign of cardiac abnormalities and did not receive any medication. Absence of both hypertrophy and end-stage heart failure was validated by expression level of the natriuretic peptides ANP and BNP .
2.3. TaqMan Real-time RT-PCR absolute quantification
Total RNA was isolated using Trizol reagent (Invitrogen) and DNase treated using the RNeasy Fibrous Tissue Mini Kit (QIAGEN) following the manufacturer's instructions. Four dilutions of each RNA sample were prepared. For each dilution, quantification of total RNA was determined by repeated optical density (OD) measurements (n = 8) at 260 nm. Lack of DNA contamination was checked by PCR without prior cDNA synthesis.
Absolute quantification of KCNQ1, and KCNE mRNA copies was determined using the calibration curve method, based on a recombinant double stranded plasmid DNA (dsDNA) molecule calculation . The calibration curves were obtained as the following: RT-PCR products of each investigated gene were cloned into the pCRII-TOPO plasmid vector (Invitrogen), and sequenced in order to confirm 100% homology with the GeneBank sequences. Four dilutions of linearized and purified fragments were quantified by repeated OD260nm measurements (n = 8), and plasmid copy number was determined. For each investigated gene, one of the 4 dilutions was used to prepare new 10-fold serial dilutions ranging from 50 up to 5.106 copies of dsDNA (from 5 up to 5.105 copies of dsDNA for KCNE5). Amplifications were performed in duplicate using FAM-labelled fluorogenic TaqMan Gene Expression Assays (Applied Biosystems). The assay references were as follows: KCNQ1, Hs00923523_m1; KCNE1, Hs00264799_s1; KCNE2, Hs00270822_s1; KCNE3, Hs00538801_m1; KCNE4, Hs00758199_g1; KCNE5, Hs00273381_s1; HPRT, Hs99999909_m1. Data were collected with instrument spectral compensations by the Applied Biosystems SDS 2.1 software. The calibration curve so obtained was used to evaluate PCR efficiency, which reached 100% for each gene.
First strand cDNA was synthesized from 2 μg of total RNA using the High-Capacity cDNA Archive Kit (Applied Biosystems). PCR reactions were performed with 40 ng of cDNA using the same probes and primers as indicated above. The hypoxanthine guanine phosphoribosyl transferase (HPRT) reference gene was used for normalizing the expression data. The number of mRNA molecules per 40 ng of reverse-transcribed total RNA was calculated from the linear regression of the standard curve.
2.4. Statistical analysis
All data are expressed as mean ± S.E.M. Statistical differences between samples were tested using a t-test or one-way analysis of variance. A value of P<0.05 was considered as statistically significant.
3.1. KCNEs shape KCNQ1 current
In order to monitor and compare interactions between KCNQ1 and the 5 KCNE subunits, we co-transfected COS-7 cells with DNA encoding for all subunits under CMV promoter. Positively transfected cells, identified with CD8 beads (associated with KCNE1, KCNE2, or KCNE3 expression) or GFP fluorescence (associated with KCNE4 or KCNE5 expression), were depolarized to various potentials from a − 80 mV holding potential. Fig. 1 illustrates representative current recordings. Current traces obtained from co-transfected cells with WT-KCNQ1 and KCNE1 had slower activation kinetics and higher current density than WT-KCNQ1 alone (Fig. 1A and B). When KCNE2 and KCNE3 are used for co-transfection, they yield instantaneous currents with low and high density, respectively (Fig. 1C and D). Finally, KCNE4 and KCNE5 produced almost undetectable currents between − 100 mV to + 40 mV when co-transfected with KCNQ1, pointing out to an inhibitory action of these two subunits on KCNQ1 function. Increasing test potentials from + 50 mV to + 90 mV revealed currents with slow activation kinetics (Fig. 1E and F).
3.2. Naturally occurring mutations reveal KCNQ1/KCNE interactions
R243H-KCNQ1 and S140G-KCNQ1 have been identified as disease-causing mutations leading to LQT syndrome and AF, respectively [8,27]. Since these two mutations had dramatic effects on KCNQ1 behavior when co-expressed with KCNE1 subunit, they were used as probes for screening KCNQ1/KCNE interactions. Expression of S140G-KCNQ1 mutation by itself produced a small current (Fig. 2A). However, co-expression of S140G-KCNQ1 mutation with KCNE1 led to very large instantaneous inward and outward currents (Fig. 2B). Introduction of both S140G-KCNQ1 and KCNE2 subunits into COS-7 cells produced a large and fast-activating current (Fig. 2C) similar to that recorded with S140G-KCNQ1 and KCNE1. S140G-KCNQ1/KCNE3 complex had current characteristics undistinguishable from those of WT-KCNQ1/KCNE3 (Fig. 2D). The current traces are large and instantaneous at all voltages tested. Co-expression of S140G-KCNQ1 with either KCNE4 or KCNE5 subunits enhanced current density when compared to co-expression of the same subunits with WT-KCNQ1 (Fig. 2E and F).
While functional expression of R243H-KCNQ1 alone yield a normal WT-KCNQ1 current, co-expression with KCNE1 revealed a 110 mV right shift in the activation curve . Co-expression of R243H-KCNQ1 with KCNE4 demonstrated an obvious association of these 2 subunits: the resulting current had slow activation and deactivation kinetics similar to those of Iks (Fig. 2G). Surprisingly, the association with KCNE4 reduced the inhibitory effect of the R243H-KCNQ1 mutation observed with the R243H-KCNQ1/KCNE1 pairing (Fig. 2H).
Current density drawn at various potentials showed that S140G-KCNQ1 mutation enhanced current density with all KCNE subunits (Fig. 3A–D). However, only KCNE1, KCNE2 and KCNE3 exhibited current enhancement in inward and outward components. To better illustrate the effect of S140G-KCNQ1 mutant, current density was compared at + 40 mV (Fig. 4). Clearly, KCNE1 and KCNE3 were activators of the WT-KCNQ1, while KCNE2, KCNE4 and KCNE5 were rather inhibitors. S140G-KCNQ1 current enhancement was evident for all KCNE subunits but KCNE3 that remained at the same level.
3.3. Absolute quantification of KCNQ1 and KCNE subunits in human heart
KCNQ1 and other KCNE subunits are expressed in different compartments of the heart, but their relative distribution remains unclear. To correlate our functional study with the distribution of KCNQ1 and KCNEs in the normal human heart, we performed RNA quantification of each subunit. Expression levels of KCNQ1 and each KCNE gene were assayed in 4 types of human cardiac samples: left atrium (LA), left ventricle (LV), left ventricular endocardium (LV Endo), and left ventricular epicardium (LV Epi) from 6 donors. Fig. 5 shows the absolute expression levels of all subunits. Overall, all subunits are expressed in the heart except KCNE5 that it is found at a very low level. In each investigated cardiac region, KCNQ1 is expressed at a very high level compared to any other KCNE transcript: KCNQ1≫KCNE4 ≥ KCNE1>KCNE3>KCNE2>KCNE5. For instance, the number of transcripts in the left atrium for KCNQ1 is 20833 ± 2003, and that for KCNE1 to KCNE5 is about 7, 56, 14, 4, and 436 times less, respectively. We also observed relatively homogeneous expression levels for each subunit within the 4 cardiac compartments assayed. Interestingly, KCNE1 and KCNE4 are the most abundant KCNE transcripts in the heart. They were found at the same expressing level in the ventricular samples, KCNE4 has a significantly higher copy number than KCNE1 in the LA (5621 ± 1283, n = 10, for KCNE4, and 2806 ± 397, n = 6, for KCNE1; P<0.05).
3.4. Does KCNE4 contribute to the IKs current in the heart?
In the light of the expression level and distribution of KCNE1 and KCNE4 in the heart, it appears that at least the two transcripts are present at high levels. If one assumes that these two transcripts are efficiently translated, how could these subunits affect the KCNQ1 current if they are expressed in the same cell? We transfected COS-7 cells with KCNQ1, KCNE1, and KCNE4 DNA with a 1:1:1 ratio. Using IRES expressing vectors, we were able to detect cells that were expressing both KCNE1 (CD8 beads) and KCNE4 (GFP) subunits. Recordings from cells expressing KCNQ1, KCNE1, and KCNE4 exhibited currents with characteristics similar to those recorded from cells expressing KCNQ1/KCNE1 complex (Fig. 6, Table 1). Currents activated at normal physiological ranges, tail current had a hook, mid-points of activation curves were not significantly different (V0.5 for KCNQ1/KCNE1 was 26.5 ± 2.6 mV, n = 10; V0.5 for KCNQ1/KCNE1/KCNE4=23.6 ± 2.7 mV, n = 8, p = 0.7), deactivation time constant was fitted with a single-exponential function and was not significantly different (p = 0.9) from that of KCNQ1/KCNE1 complex. For comparison, mid-point of activation obtained from KCNQ1/KCNE4 complex recordings was shifted towards depolarizing potentials (V0.5=118.6 ± 3.8 mV), and tail currents decayed rapidly and could be well fitted with a two-exponential function.
|Complex||Deactivation τ fast (ms)||Deactivation τ slow (ms)||Hook|
|KCNQ1/KCNE1||ND||208 ± 18 (9)*||+|
|KCNQ1/KCNE4||32 ± 5 (8)||223 ± 41 (8)*||−|
|KCNQ1/KCNE1/KCNE4||ND||205 ± 15 (11)*||+|
|S140G-KCNQ1/KCNE4||7.3 ± 1.1 (7)||79.5 ± 10.2 (7)||+/−|
|Complex||Deactivation τ fast (ms)||Deactivation τ slow (ms)||Hook|
|KCNQ1/KCNE1||ND||208 ± 18 (9)*||+|
|KCNQ1/KCNE4||32 ± 5 (8)||223 ± 41 (8)*||−|
|KCNQ1/KCNE1/KCNE4||ND||205 ± 15 (11)*||+|
|S140G-KCNQ1/KCNE4||7.3 ± 1.1 (7)||79.5 ± 10.2 (7)||+/−|
ND, not determined.
Not significantly different, P = 0.9.
KCNE4 behaved as an inhibitory subunit of WT-KCNQ1 or S140G-KCNQ1, as compared to KCNE1. It appears that, in a complex associating both KCNE1 and KCNE4, the dominant effect is carried by the activator subunit (e.g. KCNE1). Unlike the behavior of KCNE4 with WT-KCNQ1 or S140G-KCNQ1, KCNE4 acted rather as an activator with the R243H-KCNQ1 mutant (Fig. 2F). Therefore, one may hypothesize that upon triple co-expression of KCNE1, KCNE4, and the R243H-KCNQ1 mutant, the KCNE4 activating behavior would be dominant. Fig. 6E and F do not support such a hypothesis. Again, the currents were undistinguishable from those obtained with the pair R243H-KCNQ1/KCNE1. Our data suggest that if both KCNE1 and KCNE4 subunits coexist in the same cell, KCNQ1 is more likely to form a functional complex with KCNE1 subunit over KCNE4.
To date, association of KCNQ1 with KCNE1 remains the most documented among the five members of this family. However, associations of KCNQ1 with other KCNE subunits have been controversial. The aim of this study was to compare in the same background expressing system the behavior of KCNQ1 when co-expressed with either of the KCNE subunits. These in vitro functional interactions were weighted with in vivo quantification of KCNQ1 and KCNE transcripts in four compartments of the human heart.
In a first series of experiments, we reproduced previous data showing that KCNE1 associated with KCNQ1 to slow its activation kinetics and greatly enhance current density at the cell membrane. KCNE2 produced a small current that had instantaneous shape when co-expressed with KCNQ1. KCNE3 also induced instantaneous current with KCNQ1, but with high density. The full functional channels had a constitutively active behavior as described earlier in intestine . Both KCNE4 and KCNE5 produced a small current with KCNQ1 that could hardly be studied at physiological ranges. This work showed that, at higher depolarizing pulses (+ 50 mV to + 90 mV), KCNQ1/KCNE4 or KCNE5 currents had slow activation kinetics that looked like KCNQ1/KCNE1 combination.
In another series of experiments, we took advantage of disease-causing mutations to probe for interactions between KCNQ1 and other KCNE subunits. S140G-KCNQ1 mutation had a gain-of-function effect when co-expressed with KCNE1 and KCNE2. Unlike the current recorded from KCNQ1/KCNE2 complex, S140G-KCNQ1/KCNE2 complex had a large and instantaneous current. This drastic effect of KCNE2 on the S140G-KCNQ1 expression strongly supported the formation of KCNQ1/KCNE2 heteromeric channels. The gain-of-function mutation increased current density for KCNE4 or KCNE5 leading to a large slowly activating, rapidly deactivating current in the physiological range of membrane potentials.
It has been shown that KCNE1 and KCNE3 exert specific gating control on KCNQ1 through a single site of interaction located in the transmembrane domain [28–30]. A threonine residue at position 58 of KCNE1 was associated with slow gating kinetics, while a valine residue at the corresponding KCNE3 position (V72) drastically accelerated gating and caused a fraction of the channels to remain open at polarized voltages. Surprisingly, the corresponding amino acids in KCNE2, KCNE4 and KCNE5 proteins are different (I64, L50, and G75, respectively). Hence, it is hard to predict the kinetic behavior of the channel complex from the nature of the amino acid at this site. Moreover, S140G-KCNQ1 mutation, located in the external S1–S2 loop on KCNQ1, far away from the pore region, conferred to the KCNE1-containing complex a current profile typical of KCNE3-containing channel, but only changed the current–voltage relationship for KCNE4 and KCNE5. Clearly, the control of KCNQ1 channel gating by the KCNE accessory subunits is complex and requires further investigations.
Our data clearly show that all 5 KCNE subunits alter KCNQ1 current in a specific manner. However, the physiological relevance of all these protein interactions described in vitro is still poorly understood. We believe that the KCNQ1/KCNE1 association is convincingly documented in vivo. The KCNE2 and KCNE3 interactions with KCNQ1 in vivo are also very likely as they recapitulate in vitro all the properties of native currents in stomach parietal cells  and colonic crypt , respectively. In vivo evidences are still lacking for the KCNQ1 interactions with KCNE4 and KCNE5.
In vitro interactions of two subunits may find a physiological relevance if they co-localize in the same tissue and are present at a reasonable amount. We investigated not only the distribution of all subunits in well-defined heart territories but we determined also the absolute level of each transcript. It turns out that all subunits are homogeneously expressed within each region, i.e. atria, ventricle, epi- and endocardium except for KCNE4 that showed higher expression level in the atrium. Since KCNE5 has a very low copy number of transcripts in all compartments, we conclude that KCNE5 may not be a molecular determinant in shaping cardiac repolarization in human adults. Expression pattern of KCNE5 in developing mouse embryo suggests that this subunit may play a role in heart development . Interestingly, we found that KCNE4 is as abundant as KCNE1. The number of KCNE4 transcripts is even higher in the atrium than those of KCNE1. One should keep in mind that mRNA expression level does not necessarily reflect protein expression pattern. What is the physiological significance of the regional co-localization of two subunits that have opposite effects on KCNQ1 current? What are the functional consequences on cardiac action potential? Since it is proposed that each KCNQ1 channel complex may contain two or even four KCNE subunits [32,33], one may imagine that heteromultimerization of KCNE subunit exists in native KCNQ1 channels. Given the abundance of KCNE1 and KCNE4, co-expression of WT-KCNQ1 (or S140G-KCNQ1, or R243H-KCNQ1), KCNE1 and KCNE4 induced currents that were undistinguishable from the KCNQ1/KCNE1 (or S140-KCNQ1/KCNE1, or R243H-KCNQ1/KCNE1) current. Lowering the amount of KCNE1 DNA, and increasing that of KCNE4 ruled out the hypothesis that KCNE1 might have been expressed more efficiently than KCNE4. Hence, it is likely that hetero-KCNE complexes do not form in physiological conditions, or at least that the presence of KCNE1 is sufficient to confer a full KCNE1 behavior. These findings do not exclude, but strongly question, the possibility that KCNE4 may have an important role in vivo on the KCNQ1 current, particularly in pathophysiological situations leading to a reduced KCNE1 expression. Formation of heteromultimers with KCNE1, KCNE2 and KCNE3 is a less crucial matter. First, the KCNE2 and KCNE3 transcripts were only found at relatively low copy numbers and these transcripts may well be confined to particular cells. Second, it is known from expression in epithelial and parietal cells that the KCNQ1 complexes containing these different KCNEs had different localization in the cells. The KCNQ1/KCNE1 channel was found in apical location in kidney [34,35] or inner ear , the KCNQ1/KCNE2 complex was found in the tubulovesicular and canaliculi compartment in parietal cells , and the KCNQ1/KCNE3 channel harbor a basolateral location in enterocytes of large and small intestine . Hence, even though the KCNE1, KCNE2 and KCNE3 proteins can co-exist in the same cell, they may not have the same spatial distribution (sarcolemma, intercalated discs, transverse tubule or elsewhere) during physiological situations. These trafficking mechanisms could be extrapolated to KCNE1 and KCNE4 subunits, which could localize KCNQ1 in different specialized areas of cardiac myocytes. However, our mRNA expression data cannot distinguish between two cells expressing different subunit combinations.
Growing evidences have shown that KCNE peptides do not restrict their interactions to KCNQ1. KCNE1 has been reported to interact with HERG, data that need to be confirmed in vivo . KCNE2 has been described to associate with HERG, HCN1, Kv4.2, Kv4.3, and KCNQ2/KCNQ3 complex [17,39–41]. KCNE3 interacts with Kv3.4, Kv2.1, Kv3.1, HERG and KCNQ4 [19,20,42]. KCNE4 associates with Kv1.1 and Kv1.3 . We have previously shown that Kv1.1 subunit is statistically over expressed in nodal tissues than in working myocytes isolated from mouse heart . The presence of KCNE4 in pacemaking cells should be explored. To date, no partner other than KCNQ1 was described for KCNE5. The KCNE family offers a growing number for K+ channel diversity. Association of this class of auxiliary subunits with different partners and the plethora of the functional consequences generated through these interactions may be useful for cell adaptation to a variety of signals. Such signals may vary among specialized tissues to induce differential distribution of the KCNE transcripts.
We thank all individuals who participated in providing tissue samples. We are grateful to Andras Varro for his help in sample collection. We thank Nathalie Gaborit and Olivier Bignolais for technical assistance. This work was supported by grants from CNRS, INSERM, Ouest Genopole, and the Association Française contre les Myopathies (AFM).