FK506 Resistance of Saccharomyces cerevisiae Pdr5 and Candida albicans Cdr1 Involves Mutations in the Transmembrane Domains and Extracellular Loops

The 23-membered-ring macrolide tacrolimus, a commonly used immunosuppressant, also known as FK506, is a broad-spectrum inhibitor and an efflux pump substrate of pleiotropic drug resistance (PDR) ATP-binding cassette (ABC) transporters. Little, however, is known about the molecular mechanism by which FK506 inhibits PDR transporter drug efflux.

ABSTRACT

The 23-membered-ring macrolide tacrolimus, a commonly used immunosuppressant, also known as FK506, is a broad-spectrum inhibitor and an efflux pump substrate of pleiotropic drug resistance (PDR) ATP-binding cassette (ABC) transporters. Little, however, is known about the molecular mechanism by which FK506 inhibits PDR transporter drug efflux. Thus, to obtain further insights we searched for FK506-resistant mutants of Saccharomyces cerevisiae cells overexpressing either the endogenous multidrug efflux pump Pdr5 or its Candida albicans orthologue, Cdr1. A simple but powerful screen gave 69 FK506-resistant mutants with, between them, 72 mutations in either Pdr5 or Cdr1. Twenty mutations were in just three Pdr5/Cdr1 equivalent amino acid positions, T550/T540 and T552/S542 of extracellular loop 1 (EL1) and A723/A713 of EL3. Sixty of the 72 mutations were either in the ELs or the extracellular halves of individual transmembrane spans (TMSs), while 11 mutations were found near the center of individual TMSs, mostly in predicted TMS-TMS contact points, and only two mutations were in the cytosolic nucleotide-binding domains of Pdr5. We propose that FK506 inhibits Pdr5 and Cdr1 drug efflux by slowing transporter opening and/or substrate release, and that FK506 resistance of Pdr5/Cdr1 drug efflux is achieved by modifying critical intramolecular contact points that, when mutated, enable the cotransport of FK506 with other pump substrates. This may also explain why the 35 Cdr1 mutations that caused FK506 insensitivity of fluconazole efflux differed from the 13 Cdr1 mutations that caused FK506 insensitivity of cycloheximide efflux.

INTRODUCTION

Full-size fungal pleiotropic drug resistance (PDR) transporters (1, 2) belong to the large eukaryotic ABCG family of the ubiquitous ATP-binding cassette (ABC) transporter superfamily (3, 4). Eukaryotic ABC transporters can be divided into type I and type II ABC exporters with distinct transmembrane folds and reaction cycles (5). A typical PDR transporter consists of two homologous halves, each comprises a nucleotide-binding domain (NBD) that precedes a transmembrane domain (TMD) of six transmembrane spans (TMSs) separated by two intracellular and three extracellular loops (ILs and ELs, respectively) (1). The NBDs are the molecular motors transforming the chemical energy released by ATP binding and hydrolysis into the movement of compounds across biological membranes (2, 6).

Invasive fungal infections are a persistent public health problem (7, 8), with major contributing factors being the steady rise of the immunocompromised population caused by the AIDS epidemic and increasing numbers of organ transplant recipients (9), cancer patients (10), and ageing populations in developed countries. Azoles are still the first choice from the few antifungal classes available to treat invasive fungal infections (11), but treatment sometimes fails due to intrinsic or acquired resistance (7, 11). The overexpression of the archetypal fungal PDR transporters Pdr5 and Cdr1 are the most common causes of multidrug resistance in Saccharomyces cerevisiae (12) and Candida albicans (13).

An attractive approach to combat multidrug resistance is to use efflux pump inhibitors in combination with commonly used antifungals (14, 15). Several inhibitors of fungal efflux pumps, like FK506 (16), enniatin B (17), beauvericin (18), and milbemycins (19, 20), have been identified (Fig. 1). However, our understanding of how these compounds inhibit drug efflux of PDR transporters is rather limited. Previous research highlighted the importance of TMS11 in FK506 inhibition of Pdr5 (21, 22), and FK506 was later confirmed to be a substrate and a competitive inhibitor of Pdr5 drug efflux (23). Detailed biochemical characterization of one particular FK506-resistant variant, Pdr5-S1360F (TMS11), revealed no changes to the mutant’s ability to transport FK506 or rhodamine 6G (R6G) or its ability to bind or hydrolyze ATP, yet its R6G efflux was ∼25 times less sensitive to FK506 inhibition than that of wild-type (wt) Pdr5, possibly because of “an altered cross talk between the NBDs and the TMDs” (23).

FIG 1.

Structures of Pdr5 and Cdr1 efflux pump inhibitors. 23- and 16-membered-ring macrolides: (A) FK506 (molecular weight [MW], 804 Da) and (B) milbemycin α25 (MW, 572 Da). Depsipeptides: (C) enniatin B (MW, 639.8 Da) and (D) beauvericin (MW, 784 Da).

Despite these efforts, the molecular mechanism of FK506 inhibition of Pdr5/Cdr1 drug efflux remains unclear. A number of recently determined type I (24–30) and type II (31, 32) ABC exporter structures and several studies on S. cerevisiae Pdr5 (21–23, 33–37) and C. albicans Cdr1 (38–43) have improved our understanding of the structure-function relationship of fungal PDR transporters (2, 5, 44, 45). We have contributed to that knowledge by hyperexpressing and studying Pdr5, Cdr1, and a number of related fungal PDR transporters (19, 46–48) in the genetically modified S. cerevisiae host AD1-8u^– (AD) (49), a strain that is deleted in seven major multidrug efflux pumps and the transcription factor PDR3 and which has a gain-of-function mutant transcription factor, Pdr1-3 (19, 49). These modifications make AD exquisitely sensitive to most xenobiotics, and they enable exceptionally high levels of expression of heterologous multidrug efflux pumps that can result in up to 100- to 5,000-fold increased drug resistance. The fluconazole (FLC) susceptibilities (50–52) and inhibitor sensitivities (53–56) of C. albicans Cdr1 knockout strains and C. albicans clinical isolates overexpressing Cdr1 were remarkably similar to those of our azole-sensitive S. cerevisiae host strain and the azole-resistant S. cerevisiae strain overexpressing Cdr1 (18, 19, 55, 57). This validates the functional characterization of C. albicans Cdr1 in S. cerevisiae AD despite obvious differences in lipid composition, membrane potential, and bioenergetics between these two evolutionarily distant fungal species. A simple, yet highly selective, screen for naturally arising point mutations in C. albicans Cdr1 that overcome efflux pump inhibition by the Cdr1-specific d-octapeptide inhibitor RC21v3 revealed key amino acids that probably interact directly with RC21v3 (57). All six RC21v3-insensitive Cdr1 suppressor mutations were single-nucleotide changes in the CDR1 open reading frame (ORF) resulting in six amino acid changes in, or near, the large extracellular domains (EDs), indicating that RC21v3 acts as an efflux pump inhibitor by binding to the ED of Cdr1 and “freezing” the transporter (57).

Here, we employed the same powerful strategy to search for FK506 resistance-causing mutations of S. cerevisiae Pdr5 and C. albicans Cdr1 (Fig. 2). The isolation and functional characterization of 82 FK506-resistant Pdr5/Cdr1 mutants and molecular modeling of these mutants provided important new insights into how FK506 may inhibit drug efflux and the architectural changes that are necessary to overcome inhibition of drug efflux by FK506.

FIG 2.

Selection of naturally occurring FK506-resistant AD/ScPDR5 and AD/CaCDR1 mutants. Resistant mutants (R1 and R2) were isolated (step 1), and their PDR5- or CDR1-containing cassettes were PCR amplified from genomic DNA (step 2) and used to transform the FLC-sensitive host AD (step 3). In step 4, mutations in PDR5 or CDR1 from the 35 AD/ScPDR5 and 35 AD/CaCDR1 mutants that were confirmed, in step 3, to be FK506 resistant due to mutations in PDR5 or CDR1 (i.e., transformants T1 and T2, like R1 and R2, grew on YPD plates containing 40 µg/ml FLC and µg/ml FK506; inset) were determined by DNA sequencing (bottom left).

RESULTS

Inhibitor sensitivities of fungal multidrug efflux pumps.

FK506 inhibits different PDR transporters to different extents. In a preliminary experiment we compared the inhibition of PDR transporters by FK506 to that by other pump inhibitors. The inhibitor sensitivities of S. cerevisiae AD strains overexpressing the major multidrug efflux pumps of the human fungal pathogens Candida albicans (CaCDR1 and CaCDR2) or Candida glabrata (CgCDR1 and CgCDR2) or the prototype fungal PDR transporter S. cerevisiae PDR5 were investigated (Fig. 3). The fungal PDR transporter inhibitors tested were macrolides FK506 and milbemycin α25 and the depsipeptides enniatin B and beauvericin (Fig. 1). None of these inhibitors was toxic to S. cerevisiae AD cells (pABC3 control) (Fig. 3), and they did not inhibit FLC efflux by the MFS multidrug efflux pump, CaMdr1 (CaMDR1) (Fig. 3). However, they inhibited most, or all five, fungal PDR transporters tested, chemosensitizing the cells to FLC in the agar. Beauvericin inhibited all pumps, milbemycin α25 inhibited all but ScPdr5, FK506 inhibited all but CaCdr2, and enniatin B inhibited all but CaCdr2 and its C. glabrata orthologue, CgCdr2 (Fig. 3). This indicated that FK506, milbemycin α25, beauvericin, and enniatin B are effective inhibitors of CaCdr1 and CgCdr1, the major multidrug efflux pumps of two important human fungal pathogens, C. albicans and C. glabrata. In addition, apart from milbemycin α25, they also inhibit the prototype fungal PDR transporter S. cerevisiae Pdr5.

FIG 3.

FLC chemosensitization of S. cerevisiae AD overexpressing major fungal multidrug efflux pumps. YPD agar plates containing sub-MICs of FLC were seeded with AD/pABC3 (negative control) or AD/CaMDR1 (negative control), AD/ScPDR5, AD/CaCDR1, AD/CaCDR2, AD/CgCDR1, or AD/CgCDR2. Filter disks contained FK506 (F; 10 μg), enniatin B (E; 0.2 μg), beauvericin (B; 0.5 μg), or milbemycin α25 (M; 5 µg). After 2 days of incubation at 30°C, growth-inhibitory zones appeared around disks that were inhibited by the indicated efflux pump inhibitors.

Isolation of FK506-resistant Pdr5 mutants.

FK506 (16 µg/ml; 19.9 µM) inhibited S. cerevisiae Pdr5 and sensitized AD/ScPDR5 cells to FLC concentrations (2 µg/ml) that were comparable to the FLC susceptibilities of the negative-control strain AD/pABC3 (Table 1; see also Fig. S2A in the supplemental material). This enabled us to mimic, in vitro, treatment failure of an FLC-FK506 drug combination therapy of an invasive fungal infection and to determine how many Pdr5 mutations can prevent FK506 inhibition of FLC efflux without affecting its FLC efflux pump function. We searched for FK506-resistant colonies on yeast extract-peptone-dextrose (YPD) agar plates containing 10⁶ cells that were chemosensitized to 40 µg/ml FLC by 8 to 20 µg/ml FK506 (Fig. S1). We isolated 35 FK506-resistant AD/ScPDR5 mutants that had acquired 37 point mutations in PDR5. One mutant (M540I) was isolated three times, and four mutants (T550N, T552A/I, A723V, and F1310Y) were isolated twice (T552 was mutated to alanine and isoleucine) (Fig. 4). Of the 35 ScPDR5 mutants, 33 had single and two had two nonsynonymous single-nucleotide polymorphisms (nSNPs) (Fig. 4). Twenty-two of those mutations were in the N-terminal half and 15 were in the C-terminal half. The majority (26 or 73%) were in ELs, 9 (23%) of which were in the top half of individual TMSs and only 2 (5%) of which were in the cytosolic NBDs (A284V, between the Q-loop and the ABC signature of NBD1, and T1045N, just after the D-loop of NBD2) (Fig. 4). Interestingly, all 35 Pdr5 mutants also had slightly reduced FLC efflux pump function (i.e. 2- to 4-fold-reduced MIC_FLC), and some were also slightly (i.e., 2- to 4-fold-reduced MIC_R6G and MIC_CHX) (Table 1) reduced in their R6G and/or cycloheximide (CHX) transport. The mutants with the most severely reduced (up to 8-fold) Pdr5 efflux pump function were F683L (top of TMS5) and A747D (EL3) (Table 1).

TABLE 1.

Drug susceptibilities and inhibitor sensitivities of FK506-resistant suppressor mutants of AD/ScPDR5

Straina	Mutation location	MIC (µg/ml)			Inhibitor sensitivity
					FK506 effect		Growth-inhibitory zone sizee (mm)
		FLC	R6G	CHX	MICFLCc (µg/ml)	% R6G effluxd	F	E	B
AD/pABC3		0.5	0.5	0.016	1	ND	ND	ND	ND
AD/ScPDR5		256	256	1	2	18	25	30	26
A284V	NBD1	128	128	1	8	75	14	24	27
S539L	TMS1	128	128	1	32	40	0	30	28
M540I (3)	TMS1	64	256	0.5	32	15	0	0	18
T550N (2)	EL1	64	128	0.5	32	42	0	13	21
T552A	EL1	128	128	0.5	16	49	0	0	22
T552I	EL1	128	128	1	8	21	0	20	22
F622I	TMS3	64	128	0.25	32	24	0	24	22
F683L	TMS5	64	32	0.5	4	38	0	22	24
A684S	TMS5	128	128	1	16	31	0	25	27
S694A/V741L	EL3	64	128	0.5	16	36	0	20	22
W698L	EL3	128	256	0.5	16	31	0	18	21
F715Y	EL3	128	256	1	8	18	19	26	24
A723V (2)	EL3	128	128	1	64	40	17	27	26
V726F	EL3	128	128	1	64	56	0	18	22
A747D	EL3	32	32	0.125	8	56	0	14	21
G750D	EL3	128	128	0.5	32	55	0	15	20
Y764Sb	EL3	128	128	1	64	27	14	30	28
T1045N	NBD2	128	128	0.5	2	82	17	34	28
Q1235E	EL4	64	128	0.5	64	42	0	21	24
Q1237R	EL4	64	128	0.5	32	55	0	20	23
G1297C	TMS9	64	128	0.5	2	44	0	24	25
F1310Y (2)	EL5	128	128	0.5	64	26	0	27	27
T1364S/M1379V	TMS11/EL6	128	256	0.5	32	55	0	21	13
M1373T	EL6	64	256	0.5	64	25	0	14	20
P1376T	EL6	128	256	1	32	44	0	16	23
A1378S	EL6	128	128	1	4	35	17	30	27
P1380R	EL6	64	128	0.5	64	23	0	24	26
W1383L	EL6	64	128	0.5	16	28	15	33	25
Y1414N	EL6	64	128	0.5	32	36	0	23	22
G1441C	EL6	128	128	0.5	4	31	15	25	23

^aNumbers in parentheses indicate the frequency of isolation. Pdr5-A284 is equivalent to Cdr1-A278, i.e., A284/A278, S539/S531, M540/V532, T550/T540, T552/S542, F622/F612, F683/F673, A684/V674, S694/S684, W698/N688, F715/F705, A723/A713, V726/V716, V741/V731, A747/S737, G750/G740, Y764/Y754, T1045/T1036, Q1235/Q1226, Q1237/Q1228, G1297/G1288, F1310/L1301, T1364/T1355, M1373/L1364, P1376/P1367, A1378/V1369, M1379/L1370, P1380/P1371, W1383/W1374, Y1414/R1405, and G1441/G1431.

^bStrains listed after Y764S contain C-terminal mutations, while those preceding and including Y764S contain N-terminal mutations.

^cMIC_FLC was measured in the presence of 16 µg/ml FK506 (the checkerboard results for FK506 concentrations ranging from 0 to 64 µg/ml are shown in Table S1).

^dPercent R6G efflux at 20 µg/ml FK506 relative to that of the no FK506 controls; the percent R6G efflux values for 0, 5, 10, and 20 µg/ml FK506 are listed in Table S2.

^eGrowth inhibitory zones around discs containing FK506 (F; 10 µg), enniatin B (E; 1 µg), or beauvericin (B; 0.5 µg) placed on YPD agar plates supplemented with a 1/4 MIC_FLC and inoculated with 10⁵ cells. Plates were incubated at 30°C for 48 h, after which the diameters of the growth-inhibitory zones were measured.

FIG 4.

Location of 37 FK506 resistance causing Pdr5 mutations. (A) Topology model of the 35 FK506 resistance-causing TMD mutations of Pdr5. Two additional mutations were found in the NBDs of Pdr5 (A284V in NBD1 and T1045N in NBD2). Colored circles depict various amino acid groups: blue, hydrophilic (A, S, and T); orange, hydrophobic aliphatic (V, I, L, and M); brown, aromatic (W, F, and Y); green, amide group-containing (Q and N); black and pink, helix-breakers G and P, respectively. Numbers in parentheses indicate how many times the same mutation was isolated. (B) Model of Pdr5 based on the ABCG5-G8 structure. TMS1 to TMS12 are numbered, and individual domains are color coded. The 37 FK506 resistance-causing mutations are shown as sticks and use the same color codes as those shown in panels C and D. (C and D) Stick models (C, magenta for N-terminal and green for C-terminal amino acids; O, red; N, blue; S, yellow) of all (D) or almost all (C) FK506 resistance-causing mutations of the TMDs with masking parts of Pdr5 as the transparent background. Panel C is a view from the front, and panel D is a view from the top. Distances (Å) are in the same color as the dashed lines connecting proximal amino acids (TMD1 connections, yellow; TMD2 connections, blue; TMD1-TMD2 connections, red). Hot spot 1 (within green circles) includes T550-T552 (EL1) and F715-A723-V741-Y764 (EL3); hot spot 2 (red circles) includes S539-M540 (top of TMS1), W1383-F683 (top of TMS5), M1373 (top of TMS11), T552 (EL1), Y764 (EL3), and Q1235 (EL5) proximal amino acids (see the text for further details).

Susceptibilities of R6G and FLC efflux to FK506 for FK506-resistant Pdr5 mutants.

As expected, the FLC efflux (i.e., MIC_FLC) of all Pdr5 mutants was more resistant to FK506 than wt Pdr5 (Table 1). Pdr5-M540I, -T550N, -F622I, -V726F, -T1364S/M1379V, -M1373T, and -P1380R were the most FK506-resistant mutants (Table S1). However, the FK506 susceptibilities were substrate dependent, e.g., the R6G efflux of Pdr5-M540I, -M1373T, and -P1380R was as sensitive to FK506 as that of wt Pdr5 (Table 1 and Table S2). The exact opposite was true for the two FK506-resistant NBD mutants, Pdr5-A284V and -T1045N. Their FLC efflux was only slightly more resistant to FK506 than wt Pdr5, yet their R6G efflux was almost completely FK506 resistant (Table 1 and Tables S1 and S2).

Cross-resistance of the FK506-resistant Pdr5 mutants to the broad-spectrum inhibitors enniatin B and beauvericin.

Beauvericin, and possibly enniatin B, appear to be true efflux pump inhibitors that freeze PDR transporters in a particular conformation, e.g., beauvericin had very low (2.8 nM and 4.2 nM) 50% inhibitory concentration values for the Pdr5 ATPase activity and R6G transport (53). FK506, on the other hand, appears to be a competitive inhibitor of the Pdr5/Cdr1 efflux pump function. To further investigate these differences, we determined the enniatin B and beauvericin susceptibilities of all 35 FK506-resistant AD/ScPDR5 mutants. The FLC efflux of about half (18) of them was also slightly more resistant to enniatin B, but only four (three M540I and T552A) were completely resistant to enniatin B (i.e., no growth-inhibitory zones) (Table 1). Their response to beauvericin, however, was practically unchanged; only M540I and T1364S/M1379V were slightly more resistant to beauvericin (Table 1). Thus, Pdr5-M540I was the only FK506-resistant mutant for which FLC efflux was, at least partially, cross-resistant to all three Pdr5 inhibitors, FK506, enniatin B, and beauvericin. However, the same three mutants, and F715Y, were the only isolates for which R6G efflux remained sensitive to FK506 (Table 1 and Table S2). In other words, FK506 inhibition of Pdr5 is substrate dependent.

A Pdr5 model reveals two FK506 resistance hot spots.

To obtain an indication of the possible spatial arrangement of these mutations, homology models of Pdr5 and Cdr1 were generated based on the crystal structure of ABCG5/G8 (31). The structure of the ED of Cdr1 was refined by molecular dynamics (MD) simulations (Fig. S3). The average local QMEANBrane scores (58), a model quality estimation metric for membrane proteins, of 0.69 and 0.72 for Pdr5 and Cdr1, respectively, was in the same range as that of the template crystal structure (0.72), which indicates a good overall quality of the models (domain average local QMEANBrane scores: NBD1, 0.64/0.66; TMD1, 0.73/0.75; NBD2, 0.70/0.72; TMD2, 0.71/0.74). The model revealed two FK506 resistance hot spots for Pdr5 (Fig. 4B to D, areas delimited by green and red circles). Hot spot 1, on top of TMD1 near the center of EL3 (Fig. 4B, green circle), included the following residues positioned close (3 to 5 Å) (Fig. 4C, yellow dashed lines) to each other: T550 and T552 of EL1, F715 at the end of PDRB (1), and A723, V741, and Y764 of EL3 (Fig. 4C and D). FK506 resistance hot spot 2, in the center near the top of the TMDs (Fig. 4B red circle), included the following residues that “connect” (Fig. 4C and D, red dashed lines) TMD1 with TMD2: S539, M540, and W1373 connect the top of TMS1 with the EL6 motif (1) (Fig. 4C and D), F683 and M1373 connect the top of TMS5 with the top of TMS11, and T552, Y764, and Q1235 connect EL1 with EL3 and EL4 (Fig. 4C and D). Most remaining FK506 resistance residues were found in or near those hot spots, including Q1237 and F1310, which connected EL4 with EL5 (Fig. 4C and D, blue dashed line). Notable exceptions with no nearby partners were S694 and W698 of PDRA (1), near the extracellular TMS5-TMS7 boundary, and G1297 and T1364, near the centers of TMS9 and TMS11, respectively.

Isolation of FK506-resistant Cdr1 mutants.

Thirty-four FK506-resistant AD/CaCDR1 mutants were also isolated and characterized. The search for FK506-resistant Cdr1 mutants had almost reached saturation, with three-quarters (26; 74%) of the 35 mutations isolated multiple times: T540I, S542P, R546T, and A713P were isolated four times, F1235C and V1322A were isolated three times, and S542L and C712S were isolated twice each (Fig. 5). Thirty-three had a single and one had two nSNPs (Y544F/R546Y; top of TMS2) in the CDR1 ORF (Fig. 5). Twenty-five of the 35 mutations were in TMD1, and 10 were in TMD2 (Fig. 5). As for Pdr5, the majority (20, or 57%) of FK506 resistance-causing mutations were located in ELs, 19 in the N-terminal half and one in the C-terminal half of Cdr1. The remaining 15 (43%) mutations were near the center or close to the extracellular boundary of individual TMSs, and no mutation was in an NBD (Fig. 5). There were obvious similarities, but also significant differences, in the distribution of the 35 FK506-resistant Cdr1 mutations compared to the 37 Pdr5 mutations. The most obvious similarity was 14 mutations (T540I [4], S542P [4], S542L [2] of EL1 and A713P [4] of EL3; Fig. 5) that were in positions equivalent to those of the six Pdr5 FK506 resistance hot spot 1 mutations T550N (2), T552A/I, and A723V (2) (Fig. 4). Thus, almost a third (20) of all (72) FK506-resistant Pdr5/Cdr1 mutations were generated in just three equivalent amino acid positions of Pdr5/Cdr1 hot spot 1: T550/T540 and T552/S542 (EL1) and A723/A713 (EL3).

FIG 5.

Location of 35 FK506 resistance-causing Cdr1 mutations. Color coding and labeling conventions are the same as those used for Fig. 4. (A) Topology model of the 35 FK506 resistance-causing TMD mutations of Cdr1. (B) Model of Cdr1 based on the ABCG5-G8 structure. The green and red circles highlight the FK506 resistance hot spot 1 (EL1-EL3 contacts) and hot spot 2 (TMD2), respectively. (C and D) Hot spots 1 (front view) and 2 (top down view) at higher magnification with masking parts removed. Distances (Å) are in the same color (yellow) as the dashed lines connecting proximal amino acids T540-S542-Y544 (EL1) and C712-A713-Q714 (EL3).

However, there were no Cdr1 mutations found in positions equivalent to Pdr5 hot spot 2 residues. Instead, nine (26%) FK506-resistant Cdr1 mutants had mutations in just four residues (F1235C and N1240D of TMS8, W1295S of TMS9, and V1322A/G of TMS10) near the center of TMD2 that were facing each other but facing away from the substrate binding pocket (Fig. 5D). Cdr1-F1235C and -V1322A/G, right next to each other (∼4 Å) (Fig. 5D, red dashed line), were isolated three times and four times, respectively; each mutation substituted smaller residues. We called this cluster of residues Cdr1 FK506 resistance hot spot 2 (Fig. 5B and D, areas delimited by a red circle).

In vitro ATPase activities of the FK506 suppressor mutants.

The FK506-resistant Cdr1 mutant ATPase activities ranged from 53 to 205 nmol Pi/min/mg (Fig. S4). Cdr1-Q714P was the only mutant with unchanged ATPase activity. Most mutants (i.e. N535I, T540I, S542P, S542L, Y544F/R546Y, R546T, G547R, C712S, A713P, N1240D, W1295S, V1322A, V1322G, F1235C, and Q1388P) had ATPase activities ∼50% to 70% lower than that of wt Cdr1. Changes of such magnitude, however, generally bear little relationship to the uncoupled substrate transport of PDR transporters. There was no obvious correlation between the reduced ATPase activities of the F506-resistant Cdr1 mutants and their efflux pump activities. Although most mutants had somewhat reduced ATPase activities, the majority maintained wt transport function for at least one of the three substrates tested (Table 2). The only notable exceptions were N1240D, at the center of TMS8, which had severely reduced (i.e. 16-fold) R6G efflux pump activity, and N535I (EL1), with a significant (4-fold) drop in FLC and CHX efflux pump function. However, N1240D had an even higher FLC transport ability than wt Cdr1 (Table 2). We conclude that the variations in Cdr1 ATPase activities of the 34 FK506-resistant Cdr1 mutants had little impact on the observed substrate transport and inhibitor sensitivity profiles.

TABLE 2.

Drug susceptibilities and inhibitor sensitivities of FK506-resistant suppressor mutants of AD/CaCDR1

Straina	Mutation location	MIC (µg/ml)			Inhibitor sensitivity
					FK506 effect on MICsc (µg/ml)		Growth-inhibitory zone sized (mm)
		FLC	R6G	CHX	FLC	R6G	F	E	B
AD/pABC3		0.5	0.5	0.016	1	ND	ND	ND	ND
AD/CaCDR1		512	64	4	16	8	26	43	28
N535I	EL1	128	32	1	128	32	15	52	35
T540I (4)	EL1	512	64	4	128	64	13	39	26
S542L (2)	EL1	512	64	2	32	64	15	40	28
S542P (4)	EL1	512	64	2	256	64	0	39	30
Y544F/R546Y	EL1/TMS2	256	32	2	256	32	0	34	29
R546T (4)	TMS2	512	64	4	128	64	0	39	28
G547R	TMS2	256	32	2	128	16	12	18	30
C712S (2)	EL3	512	64	2	128	32	0	40	28
A713P (4)	EL3	512	64	2	128	64	0	40	28
Q714Pb	EL3	512	32	1	128	32	0	39	28
F1235C (3)	TMS8	512	64	4	256	64	0	35	29
N1240D	TMS8	1,024	4	2	256	4	0	32	26
W1295S	TMS9	512	64	2	128	32	0	38	30
V1322G	TMS10	512	64	2	256	32	0	29	25
V1322A (3)	TMS10	512	64	2	256	64	0	37	27
Q1388P	EL6	256	64	1	128	64	0	40	38

^aNumbers in parentheses indicate the frequency of isolation. Cdr1-G521 is equivalent to Pdr5-G529, i.e., G521/G529, N535/K543, T540/T550, S542/T552, Y544/Y554, R546/R556, G547/G557, C712/C722, A713/A723, Q714/E724, F1235/F1244, N1240/N1249, W1295/Y1304, V1322/S1331, and Q1388/Q1397.

^bStrains listed after Q714P contain C-terminal mutations, while those preceding and including Q714P contain N-terminal mutations.

^cMIC_FLC and MIC_R6G in the presence of 16 µg/ml FK506 (the checkerboard results for FK506 concentrations ranging from 0 to 32 µg/ml are shown in Tables S3 and S4).

^dGrowth-inhibitory zones around discs containing FK506 (F; 10 µg), enniatin B (E; 1 µg), or beauvericin (B; 0.5 µg) placed on YPD agar plates supplemented with 1/4 MIC_FLC and inoculated with 10⁵ cells. Plates were incubated at 30°C for 48 h, after which the diameters of the growth-inhibitory zones were measured.

Susceptibilities of R6G and FLC efflux to FK506 and enniatin B and beauvericin cross-resistance of FK506-resistant Cdr1 mutants.

The FK506 sensitivities of Cdr1-mediated FLC and R6G efflux were quantified with the 2-dimensional checkerboard assay. However, unlike the Pdr5 mutants that had quite different FK506 sensitivities for FLC and R6G, all FK506-resistant Cdr1 isolates also exhibited FK506-resistant R6G efflux (Table 2 and Tables S3 and S4). It would seem that FK506 resistance of FLC efflux was much easier to achieve for Cdr1 than for Pdr5 (i.e., the FLC efflux of Cdr1 mutants was generally much less susceptible to FK506 than the FK506-resistant Pdr5 mutants; compare Table S1 with Table S3). Apart from Cdr1-G547R with an ∼60% reduced growth inhibitory zone for enniatin B, there was no obvious cross-resistance to enniatin B or beauvericin for any of the 34 FK506-resistant Cdr1 mutants (Table 2).

A different set of mutations for AD/CaCDR1 cells resistant to CHX chemosensitization with FK506.

As shown above, there were obvious substrate-specific differences in the responses of individual FK506-resistant Pdr5 mutants to FK506. We therefore designed an experiment to answer the following question: what FK506-resistant mutations will be isolated when AD/CaCDR1 cells are chemosensitized to a completely different efflux pump substrate, the protein translation inhibitor CHX? We isolated 13 FK506-resistant AD/CaCDR1 mutants, all with single point mutations in Cdr1 (Fig. 6). No mutations were in the NBDs, 7 (54%) were in the ELs, and 6 (46%) were near the center of TMS8 or TMS9 or near the intracellular (TMS11) or extracellular (TMS1 and TMS5) TMS boundaries (Fig. 6). Although the set of FK506-resistant Cdr1 mutations for CHX was different from that for FLC (compare Fig. 5 with Fig. 6), all 13 FK506-resistant mutations for CHX were also within or near Cdr1 FK506 resistance hot spots 1 and 2, described above for FLC efflux of Cdr1. Cdr1-T539I (EL1) and Cdr1-Y749S and Cdr1-A753P (EL3) were near hot spot 1 residues T540 (EL1) and Q714 (EL3) (Fig. 6B), respectively. The ten remaining FK506-resistant hot spot 2 mutations for CHX efflux were all clustered around TMS11 (Fig. 6C, within red circle), distinctly different from the ten Cdr1 hot spot 2 mutations for FLC efflux (Fig. 5D). Two were within TMS11 (A1347V and T1351I), five were proximal to TMS11 (V616F [TMS2], F693L [TMS5], A1365D [EL6], and F1383L/I [EL6 helix 1]) and I528T, at the top of TMS1, and helix-breaker mutations P1238Q and G1288C, at the center of TMS8 and TMS9, respectively, were nearby.

FIG 6.

Location of 13 Cdr1 mutations causing FK506 resistance of CHX transport. Color coding and labeling conventions are the same as those used for Fig. 4 and 5. (A) Topology model of the 13 Cdr1 mutations. (B) Model of the three ED mutations with masking parts of Cdr1 removed. Amino acids C712-A713-Q714, part of FK506 resistance hot spot 1, are shown as gray sticks. Y749 is right next (2.8 Å) to Q714. (C) Model of the 10 TMD mutations of Cdr1 with the masking EDs removed (view from the extracellular side). Cdr1 hot spot 2, surrounding TMS11, is within the red circle (see the text for further details).

DISCUSSION

We employed a simple but very powerful and selective screen to isolate and characterize 85 FK506 resistance-conferring mutations of the archetype fungal multidrug efflux pumps Pdr5 and Cdr1. By overexpressing Pdr5 and Cdr1 in the sensitive host S. cerevisiae AD to exceptionally high levels (19), a large difference (200- to 1,000-fold) of drug susceptibilities was obtained between the sensitive host and the efflux pump-overexpressing strain. We isolated a significant number (82) of naturally generated suppressor mutants that remained resistant to FLC or CHX and, in addition, had acquired FK506 resistance. Forty µg/ml FLC or 0.36 µg/ml CHX and 8 to 20 µg/ml FK506 were optimal for the isolation of FK506-resistant AD/ScPDR5 and AD/CaCDR1 mutants at a surprisingly high rate of ∼1 suppressor mutant per 10⁵ to 10⁶ cells (see Fig. S1 in the supplemental material).

Isolating and characterizing 69 naturally selected FK506-resistant mutants with FLC as an efflux pump substrate indicated the range of mutations that can cause Pdr5 and Cdr1 to become resistant to FK506 while maintaining their FLC efflux pump function. In total, we isolated 72 mutations, 70 of which were spread across just 40 different residues of the TMDs and EDs of Pdr5 and Cdr1 (Fig. 7). This suggests that our search for FK506-resistant Pdr5/Cdr1 mutants was approaching saturation, especially for Cdr1, with 35 FK506-resistant mutations spread across just 14 mutated residues (Fig. 7). While ten Pdr5/Cdr1 mutations (Fig. 7, pink/green) were near the center (TMS8, TMS9, TMS10, and TMS11) and 15 near the top (TMS1, TMS2, TMS3, TMS5, TMS8, and TMS9) of individual TMSs, the majority (45) were in the ELs. Although the FK506 resistance-causing mutations for Pdr5/Cdr1 were clustered in similar regions and almost one-third (23) of all mutations were in or near FK506 resistance hot spot 1 (T550/T540, T552/S542 and A723/A713) (Fig. 7, red), an obviously important contact point between EL1 and EL3 (Fig. 4D and 5C), there were also clear differences between Pdr5 and Cdr1. The five FK506 resistance hot spot 2 mutations for Cdr1 were clustered at the intersection of TMS8 with TMS9 and TMS10 near the center of the transporter (Fig. 5D). All mutations reduced the size of the amino acids, and seven of them were in just two residues that, according to the homology model, faced each other, one in TMS8 (F1235C) and the other in TMS10 (V1322A or V1322G) (Fig. 5D). However, the 11 FK506 resistance hot spot 2 mutations for Pdr5 were clustered near the top of the TMDs at the intersection between the two halves of the transporter (Fig. 4). These could be further divided into three smaller clusters (Fig. 4C and D, red dashed lines) of mutated residues: (i) S539L-M540I, at the top of TMS1, were near (∼3 Å) W1383L, an EL6 motif 1 residue, and nearby mutations P1376T, A1378S, T1364/M1379V, and P1380R were in the loop connecting TMS11 with EL6 motif 1; (ii) F683L of TMS5 was near (∼5 Å) M1373T of TMS11 at the top of these two centrally positioned TMSs; (iii) Q1235E of EL4 was near hot spot 1 residues T552A/I of EL1 and Y764S of EL3. Another possible contact point, outside hot spots 1 and 2, was Q1237R of EL4, in close proximity (4.4 Å) to Y1310Y of EL5 (Fig. 4C and D, blue dashed lines). In addition, Pdr5 had two FK506 resistance-causing mutations in the cytosolic NBDs (A284V and T1045N). Interestingly, those two mutations were the most FK506-resistant mutations for R6G efflux, yet the FK506 sensitivity of their FLC efflux hardly differed from that of wt Pdr5.

FIG 7.

Compilation of all 70 FK506 resistance-causing mutations found in the transmembrane domains of Pdr5 (magenta) or Cdr1 (green). Amino acid positions that had FK506 resistance-causing mutations in both Pdr5 and Cdr1 are highlighted in red. There were two additional FK506 resistance-causing mutations in the NBDs of Pdr5 (A284V in NBD1 and T1045N in NBD2). Numbers indicate how many times the same mutation was isolated in Pdr5 and/or Cdr1. The tables underneath summarize how many residues (gray background) were mutated how many times (white background) in the four major N- and C-terminal domains TMD1 and ED1 (i.e., EL1, EL2, and EL3) on the left and TMD2 and ED2 (i.e., EL4, EL5, and EL6) on the right. There were a total of 40 residues mutated a total of 70 times.

Thus, based on the distribution of the 37 FK506-resistant Pdr5 mutations, it would seem that for FLC efflux of Pdr5 to become resistant to FK506, there are two main possibilities: (i) modify the critical EL1-EL3 contact or (ii) modify contacts between the N-terminal and C-terminal halves near the top of the transporter. To elucidate the substrate-specific effect of the mutations, we undertook another set of experiments using CHX instead of FLC as a pump substrate to search for FK506-resistant Cdr1 mutants. Although there were some similarities between the FK506-resistant mutants obtained for CHX and those for FLC, there were obvious differences. The set of FK506-resistant Cdr1 CHX mutants indicated that for CHX efflux of Cdr1 to become resistant to FK506, there are two main possibilities: (i) modify the critical EL1-EL3 contact or (ii) slightly adjust, directly or indirectly, the positioning of TMS11 at the center of the transporter. However, for FLC efflux of Cdr1 to become resistant to FK506, the two main possibilities, again, are (i) the modification of the obviously very important EL1-EL3 contact or (ii) a slight adjustment of the positioning of TMS8-TMS9-TMS10 near the center of Cdr1.

To learn more about how enniatin B and beauvericin inhibit fungal drug efflux pumps, we tested all 69 FK506-resistant Pdr5/Cdr1 mutants for cross-resistance to enniatin B and beauvericin. Although some mutants showed reduced enniatin B and beauvericin sensitivities, most remained unchanged in their sensitivities. The only FK506-resistant Pdr5 mutants cross-resistant to enniatin B and beauvericin were three strains with the M540I mutation at the top of TMS1. We also found that no FK506-resistant Cdr1 mutant was cross-resistant to enniatin B and beauvericin. This indicates that FK506 inhibits fungal PDR transporters differently from enniatin B and beauvericin. Recent evidence suggested that beauvericin, effluxed by the S. cerevisiae ABCC type ABC transporter Yor1, blocks Pdr5 and Cdr1 transport by freezing the transporter in a particular conformation (53, 54). Interestingly, a search for beauvericin-resistant mutants of a beauvericin-sensitive S. cerevisiae YOR1 knockout strain gave five beauvericin-resistant suppressor mutants, all of which had gain-of-function mutations in Pdr5, with three being mutations in G538 at the top of TMS1 (53). These gain-of-function-mutations (G538R/C), however, led to reduced efflux of many other Pdr5 efflux pump substrates, such as FLC and R6G. G538R/C at the top of TMS1 is on the TMS face opposite that of M540I, the only enniatin B and beauvericin cross-resistant mutation identified in our FK506-resistant Pdr5 mutant search. It would seem that subtle changes to the architecture of Pdr5 near the top of TMS1 can dramatically affect the inhibitor sensitivities and the efflux pump function of Pdr5.

As our search for possible FK506-resistant Pdr5/Cdr1 mutants was close to saturation, we obtained a comprehensive catalogue of amino acids involved in FK506 interactions and could identify those most critical for FK506 inhibition of FLC efflux by fungal PDR transporters. Although there are obvious differences between how Pdr5 and Cdr1 respond to FK506, we propose that, unlike true efflux pump inhibitors enniatin B and beauvericin, the Pdr5/Cdr1 efflux pump substrate, FK506, inhibits drug efflux by slowing transporter opening and/or substrate release, and FK506 resistance of Pdr5/Cdr1 drug efflux can be achieved by modifying critical intramolecular contact points that enable large conformational changes necessary for the cotransport of FK506 with other efflux pump substrates, such as FLC, CHX, or R6G. Achieving cotransport of FK506 with other efflux pump substrates, however, comes at a cost, i.e., the reduced transport efficiency of other efflux pump substrates. To test this possible cotransport hypothesis, we performed 2-dimensional checkerboard experiments of select Cdr1 efflux pump substrates. The FLC-nigericin combination showed that these two drugs were indifferent to each other, and they did not affect each other’s resistance level in AD/CDR1 cells (Fig. S2B). The cotransport hypothesis would also provide an attractive alternative for the hypothesis concerning altered cross talk between the NBDs and the TMDs in the Pdr5-S1360F mutant (23). We hope that the collection of valuable mutants generated in this study, and our hypothesis, will help direct future research into verifying the proposed molecular mechanism of FK506 inhibition of fungal PDR transporters.

MATERIALS AND METHODS

Chemical reagents.

FK506 was kindly provided by Astellas Pharma, Inc. (Tokyo, Japan). Milbemycin α25 was a generous gift from Daiichi Sankyo, Co., Ltd. (Tokyo, Japan). Enniatin B was purchased from Alexis Biochemicals (San Diego, CA, USA). Beauvericin and rhodamine 6G (R6G) were obtained from Sigma-Aldrich Japan, Inc. (Tokyo, Japan). FLC was purchased from LKT Laboratories, Inc. (St. Paul, MN, USA), and cycloheximide (CHX) was purchased from Nacalai Tesque, Inc. (Kyoto, Japan).

Isolation of inhibitor-resistant mutants of AD/ScPDR5 and AD/CaCDR1.

A detailed description of the various strains that overexpress efflux pumps, as well as the sensitive control strain AD/pABC3, was provided by Lamping et al. (19). YPD agar plates containing high (40 µg/ml), but sub-growth-inhibitory (i.e., 1/6 MIC_FLC for AD/ScPDR5 and 1/12 MIC_FLC for AD/CaCDR1), FLC concentrations and 8 to 20 µg/ml FK506 were seeded with 10⁶ cells and incubated for 2 to 5 days at 30°C until clearly visible FK506-resistant AD/PDR5 or AD/CDR1 colonies appeared (Fig. 2, step 1, white arrows; see also Fig. S1 in the supplemental material). This experimental design forced AD/ScPDR5 and AD/CaCDR1 cells to maintain sufficient FLC efflux pump function but, at the same time, become resistant to FK506. To distinguish FK506-resistant mutants with mutations in PDR5 or CDR1 (Fig. 2, red cross) from those with mutations in other portions of the genome, nuclear DNA containing the cassette that was used to introduce PDR5 or CDR1 into S. cerevisiae AD, comprising the entire PDR5- or CDR1-URA3 region (Fig. 2, blue), was PCR amplified from genomic DNA and used to transform the FLC-sensitive host S. cerevisiae AD (Fig. 2, step 2). Transformation cassettes with point mutations in PDR5 or CDR1 gave FK506-resistant transformants that overexpressed Pdr5 or Cdr1 (Fig. 2, step 3). Once it was confirmed that the FK506 resistance phenotype was due to point mutations within the PDR5 or CDR1 ORF, individual mutations (Fig. 2, red arrow in step 4) were determined by DNA sequencing.

To test whether FK506 inhibition of the Cdr1 efflux pump function was substrate dependent, we also selected FK506-resistant AD/CDR1 mutants on YPD agar plates supplemented with high (0.36 µg/ml), but sub-growth-inhibitory (i.e. 1/10 MIC_CHX for AD/CaCDR1), concentrations of CHX in the presence of 10 µg/ml FK506.

Determination of antifungal MICs.

The susceptibilities of yeast to FLC, R6G, and CHX were determined according to the CLSI broth microdilution method, except that complete supplement mixture (CSM) was used instead of RPMI, as S. cerevisiae AD cells do not grow in RPMI medium. Logarithmic-phase cells (∼10⁴ cells/ml) were incubated in CSM liquid medium in the presence of drugs at 30°C for 2 days (59). The MIC for an antifungal was defined as the lowest concentration of the drug that inhibited growth by >80%.

R6G efflux activity.

The energy-dependent R6G efflux from yeast cells was measured as described previously (60). In short, starved, logarithmic-phase cells preloaded with 15 µM R6G in the presence of 5 mM 2-deoxyglucose were incubated at 30°C for 20 min with various FK506 concentrations in a total of 100 µl HEPES buffer (50 mM, pH 7.0). R6G efflux from cells was started by adding 100 µl 40 mM glucose solution. After incubating the cell suspension at 30°C for 8 min, R6G accumulation in the supernatant (expressed as nmol/10⁷ cells) was determined spectrophotometrically after removing the cells by filtration through a 2-µm glass-fiber filter microtiter plate (Pall Corporation, East Hills, NY, USA) and collecting the supernatant in individual wells of 96-well black flat-bottom microtiter plates (BMG Labtechnologies GmbH, Offenburg, Germany).

Agar diffusion FLC chemosensitization assay.

Inhibition of FLC transport by wt and FK506-resistant AD/PDR5 or AD/CDR1 mutants was determined semiquantitatively with agar diffusion assays. YPD agar plates containing FLC at a 1/4 MIC_FLC of the individual test strains were seeded with 2 × 10⁶ logarithmic-phase cells suspended in 5 ml melted (50°C) top agar (i.e., YPD plus 0.6% [wt/vol] agarose and FLC at a 1/4 MIC_FLC). Filter discs containing optimized amounts (i.e., giving good-sized growth-inhibitory zones) of FK506, milbemycin α25, enniatin B, or beauvericin were placed onto the solidified top agar, and the plates were incubated at 30°C for 48 h. The diameters of growth-inhibitory zones surrounding the filter discs were measured, and the increased resistance of individual mutants, relative to the inhibitor-sensitive AD control strains overexpressing wt PDR5 or CDR1, was expressed as percent reduced growth-inhibitory zone.

Checkerboard chemosensitization assay.

FK506 inhibition of FLC and R6G efflux of wt and FK506-resistant AD/ScPDR5 or AD/CaCDR1 mutants was quantified with the checkerboard chemosensitization assay as previously described (59). In brief, 2-dimensional checkerboard assays were performed in microtiter plates with individual wells of 100 µl YPD medium containing 2-fold serial dilutions of FK506 (0, 1, 2, 4, 8, 16, 32, and 64 µg/ml) in one dimension and FLC (0, 1, 2, 4, 8, 16, 32, 64, 128, 256, 512, and 1,024 µg/ml) or R6G (0, 0.125, 0.25, 0.5, 1, 2, 4, 8, 16, 32, 64, and 128 µg/ml) in the second dimension. Each of these wells was then inoculated with 100 µl YPD-cell suspensions of ∼4,000 logarithmic cells and incubated at 30°C with shaking at 150 rpm for 48 h, after which cell growth was monitored spectrophotometrically at 600 nm.

Isolation of plasma membranes and determination of Cdr1 ATPase activities.

Yeast cells overexpressing various Cdr1 mutants were grown in YPD at 30°C to mid-logarithmic growth phase, partially purified plasma membranes were isolated, and their Cdr1 ATPase activities were determined as previously described (59).

Homology modeling of Pdr5 and Cdr1.

Homology models of C. albicans Cdr1 (residues 169 to 1501) and S. cerevisiae Pdr5 (residues 175 to 1511) were constructed using a four-step procedure (details are given in the supplemental material). First, using the crystal structure of the human sterol transporter ABCG5/G8 (PDB entry 5DO7) (31) as a template (sequence identity/similarity/coverage, ∼24%/32%/92%), three candidate models of Cdr1 were generated by three different modeling approaches (61–63) and evaluated with respect to their suitability for further refinement of the extracellular domain through molecular dynamics (MD) simulations. The predominantly unstructured extracellular domain next was refined during 5.0 µs of restrained implicit solvent MD simulations (64). The final Cdr1 model was then obtained by replacing the ED of the initial model with the refined, MD-derived structure and minimizing the resulting structure in an implicit membrane environment to resolve steric clashes introduced by the assembly procedure. The Pdr5 model was constructed with the SWISS-MODEL server (61) by using the final Cdr1 model as a template structure (sequence identity/similarity/coverage, ∼58%/75%/99%).