Mutagenesis and structural analysis reveal the CTX-M β-lactamase active site is optimized for cephalosporin catalysis and drug resistance

Abstract

CTX-M β-lactamases are a widespread source of resistance to β-lactam antibiotics in Gram-negative bacteria. These enzymes readily hydrolyze penicillins and cephalosporins, including oxyimino-cephalosporins such as cefotaxime. To investigate the preference of CTX-M enzymes for cephalosporins, we examined eleven active-site residues in the CTX-M-14 β-lactamase model system by alanine mutagenesis to assess the contribution of the residues to catalysis and specificity for the hydrolysis of the penicillin, ampicillin, and the cephalosporins cephalothin and cefotaxime. Key active site residues for class A β-lactamases, including Lys73, Ser130, Asn132, Lys234, Thr216, and Thr235, contribute significantly to substrate binding and catalysis of penicillin and cephalosporin substrates in that alanine substitutions decrease both k_cat and k_cat/K_M. A second group of residues, including Asn104, Tyr105, Asn106, Thr215, and Thr216, contribute only to substrate binding, with the substitutions decreasing only k_cat/K_M. Importantly, calculating the average effect of a substitution across the 11 active-site residues shows that the most significant impact is on cefotaxime hydrolysis while ampicillin hydrolysis is least affected, suggesting the active site is highly optimized for cefotaxime catalysis. Furthermore, we determined X-ray crystal structures for the apo-enzymes of the mutants N106A, S130A, N132A, N170A, T215A, and T235A. Surprisingly, in the structures of some mutants, particularly N106A and T235A, the changes in structure propagate from the site of substitution to other regions of the active site, suggesting that the impact of substitutions is due to more widespread changes in structure and illustrating the interconnected nature of the active site.

Widespread antibiotic resistance gives rise to increasingly difficult-to-treat bacterial infections and is a significant threat to public health (1). β-lactams, including penicillins, cephalosporins, and carbapenems are among the most commonly used antibiotics worldwide (2). However, resistance to β-lactam drugs is widespread and most often due to the bacterial production of β-lactamases. These enzymes catalyze the hydrolysis of the amide bond in the β-lactam ring, resulting in a product that binds weakly to the target transpeptidase enzymes that are responsible for crosslinking the cell wall. β-lactamases are grouped into four classes, A, B, C, and D, based on homology of the primary amino acid sequence (3). Classes A, C, and D are serine hydrolases where a covalent, acyl-intermediate is formed and subsequently hydrolyzed by an activated water, while class B includes zinc metallo-enzymes where the β-lactam is hydrolyzed by direct attack of water (4, 5).

Class A β-lactamases function by nucleophilic catalysis with sequential acylation and deacylation steps where the catalytic Ser70 residue attacks the carbonyl carbon of the β-lactam ring to form a covalent acyl-enzyme intermediate (4, 6). Ser70 is activated by Lys73 and Glu166 and the reaction is facilitated by a proton transfer from Lys73 to Ser130, which protonates the nitrogen from the β-lactam amide leaving group during the acylation reaction (7). The active site residue Glu166 then acts as a general base to activate a water molecule for attack on the carbonyl carbon of the acyl-enzyme intermediate to release the hydrolyzed product (6, 8, 9). These catalytic residues are conserved in class A enzymes, consistent with their essential role in function (10). A minimal kinetic scheme for class A β-lactamase-mediated hydrolysis is shown in Figure 1A.

Figure 1.

Kinetic scheme and β-lactam antibiotics used in this study.A, kinetic scheme for class A β-lactamase hydrolysis of β-lactam antibiotics. B, structures of β-lactam antibiotics. The R1 and R2 groups vary between different β-lactam antibiotics.

Class A β-lactamases are widespread in bacteria and are a common source of resistance (11). These enzymes rapidly hydrolyze penicillins and many cephalosporins but have less activity toward extended-spectrum cephalosporins such as cefotaxime (12, 13). However, CTX-M β-lactamases are class A enzymes that rapidly hydrolyze penicillins and cephalosporins, including extended-spectrum cephalosporins such as cefotaxime. CTX-M β-lactamases emerged in Gram-negative bacteria in the late 1980s, likely due to the selective pressure of extended-spectrum cephalosporin use (14). They are now the most common extended-spectrum β-lactamases in Gram-negative bacteria (15). The CTX-M enzymes are divided into five subgroups based on amino acid sequence homology, including CTX-M-1, CTX-M-2, CTX-M-8, CTX-M-9, and CTX-M-25; where the name is from a prominent member of the subgroup (13, 14, 16). There is >10% sequence divergence between subgroups, but enzymes within a subgroup differ by only a few amino acids. CTX-M-14 is a member of the CTX-M-9 subgroup and is commonly found in Gram-negative pathogens.

The CTX-M-14 enzyme readily hydrolyzes the extended-spectrum cephalosporin cefotaxime with a k_cat/K_M value of ∼10⁶ M⁻¹ s⁻¹ (12, 17, 18, 19), which is 1500-fold faster than that observed for the canonical class A β-lactamase, TEM-1 (12, 20). Similar to TEM-1, CTX-M-14 rapidly hydrolyzes penicillins and early-generation cephalosporins with k_cat/K_M values of 10⁶ to 10⁷ M⁻¹ s⁻¹.

Here, we examine the contributions of active site residues of CTX-M-14 to the hydrolysis of ampicillin, cephalothin, and cefotaxime as representative penicillins, early-generation cephalosporins, and extended-spectrum cephalosporins, respectively (Fig. 1B). For this purpose, we created alanine substitutions for 11 active-site residues that are directly involved in catalysis or interact with substrates and determined steady-state kinetic parameters with the purified enzyme and each substrate (Fig. 2).

Figure 2.

CTX-M amino acid residues targeted for alanine mutagenesis. A ribbon diagram of the CTX-M-14 β-lactamase (PDB id: 1YLT) is shown with the active-site positions that were mutated to alanine shown in cyan. Ser70, the residue responsible for nucleophilic attack on the carbonyl carbon of the β-lactam to form the covalent intermediate and Glu166, which activates a water molecule for attack on the carbonyl of the ester covalent intermediate and catalysis of the deacylation reaction are shown in light green.

The goal of this study was to determine the contribution of individual active-site residues to penicillin and cephalosporin catalysis and whether residue contributions are similar between these classes of β-lactam antibiotics, i.e., penicillins versus cephalosporins. Further, we assessed if residues make similar contributions to catalysis within a β-lactam class, i.e., between the older cephalosporin, cephalothin, versus the extended-spectrum drug, cefotaxime. Finally, we investigated the structural impact of alanine mutations by determining the X-ray crystal structures of several mutant enzymes.

Results

Active-site residues make substrate-dependent contributions to β-lactam hydrolysis

To examine the impact of active-site residues of CTX-M β-lactamases on catalysis and substrate specificity, we created alanine substitutions for 11 residue positions in the active site of CTX-M-14 β-lactamase (Fig. 2). The rationale for choosing these residues is that several are known to be important for the hydrolysis of β-lactams by class A β-lactamases including Lys73, Ser130, Asn132, Lys234, and Thr235, based on results from other enzymes (21, 22, 23, 24, 25, 26, 27). Other chosen residues, including Asn104, Tyr105, Asn106, Asn170, Thr215, and Thr216, do not directly contribute to catalysis but may have an important influence on substrate specificity in that they are positioned in the active site to interact with the bound substrate or intermediates (Fig. 2) (12, 17, 19, 28). Certain residues, including Ser70, Glu166, and Ser237, although located in the active site, were not included due to extensive previous results on these positions.

Each of the alanine mutants was generated by oligonucleotide-directed mutagenesis and the enzyme was expressed and purified from Escherichia coli (Experimental procedures). Michaelis–Menten kinetic parameters were determined for WT CTX-M-14 and each mutant with ampicillin as a representative penicillin, with cephalothin and cefotaxime serving as representative cephalosporins (Figs. 1B and S1–S3; Table 1, Table 2, Table 3). As noted, WT CTX-M-14 is an excellent cephalosporinase (Tables 2 and 3) and also efficiently hydrolyzes ampicillin (Table 1).

Table 1.

Enzyme kinetic parameters for ampicillin hydrolysis by CTX-M-14 WT and alanine mutants

Enzyme	Ampicillin
	kcat	ΔΔG kcat	KM	kcat/KM	ΔΔG kcat/KM
	s−1	kcal/mol	μM	μM−1s−1	kcal/mol
WT	64 ± 2		110 ± 7	0.58 ± 0.04
K73A	1.6 ± 0.1	2.18	11,000 ± 900	0.00015 ± 0.00002	4.89
N104A	97 ± 2	−0.24	450 ± 20	0.22 ± 0.01	0.57
Y105A	53 ± 3	0.11	760 ± 100	0.07 ± 0.01	1.25
N106A	100 ± 4	−0.26	350 ± 30	0.29 ± 0.03	0.41
S130A	31 ± 20	0.43	33,000 ± 2 × 104	0.001 ± 0.0008	3.77
N132A	1.3 ± 0.1	2.31	390 ± 100	0.003 ± 0.0009	3.12
N170A	1.7 ± 0.1	2.15	140 ± 10	0.012 ± 0.001	2.30
T215A	85 ± 3	−0.17	220 ± 20	0.39 ± 0.04	−0.24
T216A	200 ± 40	−0.67	12,000 ± 4000	0.017 ± 0.007	2.09
K234A	>0.2	ND	ND	0.002 ± 0.001	3.36
T235A	150 ± 30	−0.50	18,000 ± 4000	0.008 ± 0.003	2.54
		0.63 avg			2.19 avg

Table 2.

Enzyme kinetic parameters for cephalothin hydrolysis by CTX-M-14 WT and alanine mutants

Enzyme	Cephalothin
	kcat	ΔΔG kcat	KM	kcat/KM	ΔΔG kcat/KM
	s−1	kcal/mol	μM	μM−1 s−1	kcal/mol
WT	730 ± 30		200 ± 20	3.7 ± 0.4
K73A	0.05 ± 0.001	5.68	58 ± 3	0.0009 ± 0.00005	4.93
N104A	400 ± 10	0.36	380 ± 20	1.1 ± 0.1	0.72
Y105A	550 ± 20	0.17	1100 ± 90	0.50 ± 0.05	1.18
N106A	520 ± 20	0.20	160 ± 20	3.3 ± 0.4	0.07
S130A	3.9 ± 0.1	3.10	190 ± 9	0.02 ± 0.001	3.09
N132A	0.35 ± 0.01	4.52	250 ± 20	0.001 ± 0.0001	4.86
N170A	340 ± 20	0.45	2800 ± 400	0.12 ± 0.02	2.03
T215A	1000 ± 70	−0.19	230 ± 40	4.4 ± 0.8	−0.10
T216A	790 ± 20	−0.05	1000 ± 70	0.79 ± 0.06	0.91
K234A	0.28 ± 0.01	4.66	2500 ± 100	0.0001 ± 0.00001	6.23
T235A	100 ± 5	1.18	11,000 ± 900	0.01 ± 0.001	3.50
		1.83 avg			2.49 avg

Table 3.

Enzyme kinetic parameters for cefotaxime hydrolysis by CTX-M-14 WT and alanine mutants

Enzyme	Cefotaxime
	kcat	ΔΔG kcat	KM	kcat/KM	ΔΔG kcat/KM
	s−1	kcal/mol	μM	μM−1 s−1	kcal/mol
WT	100 ± 1		66 ± 3	1.5 ± 0.1
K73A	0.02 ± 0.0001	5.04	670 ± 40	0.00003 ± 2 × 10−6	6.41
N104A	150 ± 10	−0.24	2000 ± 200	0.075 ± 0.01	1.77
Y105A	55 ± 2	0.35	1300 ± 70	0.042 ± 0.008	2.12
N106A	120 ± 3	−0.11	320 ± 20	0.38 ± 0.03	0.81
S130A	0.12 ± 0.002	3.98	250 ± 10	0.0005 ± 2 × 10−5	4.74
N132A	0.17 ± 0.004	3.78	900 ± 10	0.0002 ± 1 × 10−5	5.28
N170A	7.0 ± 0.1	1.57	580 ± 20	0.012 ± 0.001	2.86
T215A	53 ± 1	0.38	94 ± 3	0.56 ± 0.02	0.58
T216A	41 ± 2	0.53	2500 ± 200	0.016 ± 0.002	2.69
K234A	0.005 ± 0.0002	5.86	220 ± 30	0.00002 ± 3 × 10−6	6.65
T235A	0.78 ± 0.07	2.87	4900 ± 500	0.0002 ± 2 × 10−5	5.28
		2.18 avg			3.56 avg

Alanine substitutions at the 11 active-site residues have a wide range of effects on the hydrolysis of the various substrates (Table 1, Table 2, Table 3 and Figs. S1–S3). k_cat is representative of events occurring on the enzyme and for the β-lactamase mechanism it reflects the magnitude and relationship of the acylation (k₂) and deacylation (k₃) rates (Fig. 1A) (12, 29). Alanine substitutions at several positions, including Lys73, Ser130, Asn132, Asn170, Lys234, and Thr235, result in large decreases in k_cat, suggesting that these residues contribute directly to the acylation and/or deacylation reactions (Table 1, Table 2, Table 3). In contrast, substitutions at Asn104, Tyr105, Asn106, Thr215, and Thr216 result in only modest effects on k_cat for all three substrates, suggesting that these residues do not directly contribute to the acylation or deacylation reactions for penicillins or cephalosporins (Table 1, Table 2, Table 3).

k_cat/K_M is the second-order rate constant for substrate-binding times the probability that the bound substrate is converted to product and thus sets a lower limit on the substrate binding rate (k₁) (30, 31, 32). k_cat/K_M reflects rate constants up to the first irreversible step (33, 34) and in the β-lactamase mechanism, it includes events up to the formation of the acyl-enzyme, including the K_S for substrate binding and the acylation rate (k₂) (12, 29) (Fig. 1A). Nearly all of the 11 residues mutagenized contribute to k_cat/K_M values (Table 1, Table 2, Table 3 and Figs. S1–S3). The substitution of alanine for Lys73, Ser130, Asn132, Asn170, Lys234, and Thr235 results in >2 kcal/mol positive ΔΔG values for all three substrates, indicative of the >100× decreases in k_cat/K_M (35, 36, 37). Alanine substitution of Asn104, Tyr105, Asn106, and Thr216 also results in decreased k_cat/K_M values, particularly with cefotaxime as substrate (Table 1, Table 2, Table 3 and Figs. S1–S3). Only the T215A substitution has a modest effect on k_cat/K_M for all three substrates. These results suggest that nearly all of the active-site residues mutated contribute to substrate-binding affinity and/or the rate of acylation.

The findings that Lys73, Ser130, Asn132, Asn170, Lys234, and Thr235 make large contributions to k_cat and k_cat/K_M for all three substrates is consistent with their proposed roles in catalysis. Lys73 is proposed to activate Ser70 for nucleophilic attack on the β-lactam and also to serve as a proton shuttle and, via Ser130, to the β-lactam nitrogen leaving group. Asn132 and Thr235 make key hydrogen bonding interactions with the side chain acylamide or C3/C4 carboxylate that are present in penicillins and cephalosporins. Further, Asn170 hydrogen bonds to a catalytic water that participates in acylation and deacylation reactions and the charge on Lys234 facilitates interactions with the substrate C3/C4 carboxylate and influences the pKa of the Ser130 proton shuttle.

Similarly, the observation that Asn104, Tyr105, Asn106, and Thr216 contribute only to k_cat/K_M is consistent with a more peripheral location around the active site to facilitate substrate binding but not a direct contribution to catalysis (Fig. 2).

It is of interest that, although k_cat and/or k_cat/K_M for all substrates are affected by alanine substitutions of many residues, the relative impact of the substitutions is dependent on the substrate. The change in free energy (ΔΔG) associated with k_cat for the substitutions varies greatly for ampicillin versus the cephalosporins. The average ΔΔG for k_cat for ampicillin hydrolysis, defined as the sum of the individual ΔΔG values for the mutants divided by the total number of mutants, is 0.63 kcal/mol (Table 1). In contrast, for the cephalosporins cephalothin and cefotaxime, the average ΔΔG is much higher, with values of 1.83 and 2.18 kcal/mol, respectively (Tables 2 and 3). These results indicate that, on average, k_cat for the alanine mutants is reduced >10-fold more for cephalosporin substrates compared to ampicillin. Therefore, the active-site residues, on average, make a larger contribution to the acylation and/or deacylation rates for cephalosporin versus ampicillin hydrolysis. The underlying basis of this observation is that the K73A, S130A, N132A, K234A, T235A substitutions result in a much larger decrease in k_cat for cephalothin and cefotaxime versus ampicillin hydrolysis while N104A, Y105A, N106A, N170A, T215A, and T216A have a modest effect (Table 1, Table 2, Table 3 and Figs. S1–S3). Of note, alanine substitutions at Ser130 and Thr235 have no effect on k_cat for ampicillin but a large effect for cephalosporins.

A similar trend is observed for the effect of alanine mutations on k_cat/K_M. The average ΔΔG values for k_cat/K_M are 2.19 for ampicillin, 2.49 for cephalothin, and 3.56 kcal/mol for cefotaxime (Table 1, Table 2, Table 3). Thus, as observed for k_cat, the average ΔΔG for the mutants is lowest for ampicillin and highest for cefotaxime. The underlying basis for this observation is that all alanine mutants exhibit higher ΔΔG values for cefotaxime and higher or equivalent values for cephalothin than ampicillin (Table 1, Table 2, Table 3). Taken together, the results indicate that the active-site residues in CTX-M-14 make a significantly larger contribution to substrate binding and turnover of cefotaxime than cephalothin and especially compared to ampicillin. These findings suggest the CTX-M-14 active site is highly optimized for the hydrolysis of cefotaxime, which is consistent with the fact that the hallmark of CTX-M enzymes compared to other class A enzymes is the efficient hydrolysis of cefotaxime (13, 14).

X-ray structures indicate some alanine mutants propagate changes through the active site

We determined X-ray crystal structures for the apo-enzymes of several of the mutants, including N106A, S130A, N132A, N170A, T215A, and T235A, to assess the impact of the substitutions on the active-site conformation. As noted, the N132A substitution greatly decreases k_cat and k_cat/K_M for all substrates (Table 1, Table 2, Table 3). The X-ray structure of the apo-enzyme of the N132A mutant was also determined and it shows the β-carbon of Ala132 is in the same position as the β-carbon of Asn132 in WT CTX-M-14 (Fig. 3, A–C and Table 4). However, the Asn104 side chain forms a hydrogen bond with Asn132 in the WT enzyme and, in the absence of this hydrogen bond due to the N132A mutation, a new water molecule is present near the position of the Asn132 side chain in WT (Fig. 3, B and C). The alanine substitution precludes hydrogen bonding to the acylamide side chain of penicillins and cephalosporins. The large decrease in k_cat and k_cat/K_M associated with N132A for all substrates suggests this hydrogen bond makes a large contribution to substrate binding and catalysis (7, 38, 39). In addition, the new hydrogen bond to Glu166 from the water present between Asn104 and Glu166 of the N132A enzyme could lower the basicity of Glu166 and thereby slow the acylation and deacylation of penicillins and cephalosporins (40, 41).

Figure 3.

X-ray crystal structures of the active site of WT CTX-M-14 β-lactamase and alanine mutants.A, diagram of WT CTX-M-14 active site (PDB ID:1YLT). Relevant amino acid residues are labeled. Carbons are colored light blue, nitrogen is dark blue, and oxygen is red. Hydrogen bonds are indicated by dotted lines. B, structure of N132A mutant enzyme active site (tan). C, structure alignment of WT and N132A mutant enzymes. Hydrogen bonds are not shown for clarity. Waters from CTX-M-14 WT are shown in red and waters from N132A are shown in tan. D, active site of CTX-M-14 enzyme (light blue). E, active site of N170A mutant enzyme (light green). F, structure alignment of WT and N170A mutant enzymes. Waters from CTX-M-14 WT are shown in red and waters from N170A are shown in green. G, active site of CTX-M-14 enzyme (light blue). H, active site of S130A mutant enzyme (pink). I, structure alignment of WT and S130A mutant enzymes.

Table 4.

X-ray crystallography data collection and refinement statistics for CTX-M-14 mutant enzymes

Crystal (PDB ID)	N106A (8DOE)	S130A (8DOD)	N132A (8ELB)	N132A/MES (8ELA)	N170A (8DPQ)	T215A (8DON)	T235A (8DP4)
Data Collection
Space group	P 1 21 1	P 32 2 1	C 2 2 21	C 2 2 21	P 32 2 1	P 32 2 1	P 41 21 2
a, b, c (Å)	49.84, 37.91, 65.46	41.49, 41.49, 230.78	59.23, 60.41, 264.1	59.37, 60.430, 264.22	41.42.42, 42.42, 230.49	41.43, 41.43, 231.26	41.96, 41.96, 260.53
a, β, γ (deg)	90.00, 103.44, 90.00	90, 90, 120	90, 90, 90	90, 90, 90	90, 90, 120	90, 90, 120	90, 90, 90
Resolution range (Å)	34.86–1.5 (1.55–1.5)	35.5–1.61 (1.67–1.61)	42.29–1.50 (1.52–1.50)	30.00–1.50 (1.53–1.50)	34.25–1.67 (1.73–1.67)	34.27–1.36 (1.41–1.36)	29.48–1.40 (1.45–1.40)
Rmerge (%)	12.3 (61.5)	5.5 (23.9)	18.3 (11.8)	6.8 (24.0)	6.7 (16.8)	9.3 (19.6)	10.6 (34.9)
Rpim (%)	9.7 (49.3)	2.9 (12.4)	8.4 (5.8)	3.9 (11.7)	2.3 (6.2)	5.9 (12.2)	3.6 (9.1)
I/σ(I)	9.0 (5.5)	19.8 (5.8)	6.0 (7.9)	13.2 (6.0)	24.00 (19.10)	11.3 (7.5)	9.40 (7.80)
CC(1/2)	0.982 (0.563)	0.999 (0.983)	0.975 (0.975)	0.975 (0.916)	0.997 (0.982)	0.991 (0.976)	0.988 (0.978)
Multiplicity	4.4 (4.5)	7.8 (8.0)	5.9 (4.7)	5.9 (6.6)	7.4 (9.1)	6.6 (6.8)	11.9 (15.5)
Completeness (%)	98.46 (97.22)	99.5 (99.3)	90.0 (98.78)	76.0 (90.8)	97.30 (93.75)	99.57 (99.98)	98.00 (99.67)
Wilson B-factor (Å2)	8.4	17.4	10.4	12.6	12.98	7.9	13.0
No. of unique reflections	37,858 (3704)	31,100 (3050)	69,260 (7451)	57,378 (6290)	27,171 (2567)	50,984 (5041)	46,441 (4598)
Refinement
Rwork, Rfree (%)	14.1, 16.4	22.2, 25.9	18.8, 21.9	17.5, 20.9	17.9, 23.9	14.5, 16.0	16.8, 18.2
No. of protein residues	262	263	523	518	262	262	263
No. of water molecules	418	207	551	627	280	430	321
Ramachandran favored (%)	98.46	96.92	97.69	98.07	98.46	98.45	98.46
Ramachandran outliers (%)	0.00	1.15	0.39	0.39	0.38	0.39	0.00
Bond lengths (Å)	0.01	0.031	0.020	0.006	0.026	0.009	0.006
Bond angles (deg)	1.49	1.66	2.01	0.972	2.16	1.43	1.30
Overall average B-factor (Å2)	12.36	30.33	13.21	15.88	21.57	12.81	17.33
Protein average B-factor (Å2)	9.75	29.97	11.5	14.38	20.34	10.38	15.01

Values in parentheses in the body of the table indicate the highest resolution shell.

We also determined a second structure of the N132A enzyme under different crystallization conditions and in a different space group. This structure contained a zinc ion, a Cl⁻ ion, an SO₄ molecule, and a 2-(N-morpholino) ethanesulfonic acid (MES) molecule from the buffer located in the active site (Fig. S4, A–C). These ligands were accommodated by the absence of the Asn132 sidechain as well as movement of the side chains of Ser70, Asn104, Glu166, and Asn170 (Fig. S4D). The structure reveals that, although the apo-N132A structure described above superimposes with the WT enzyme, the positions of the side chains in the active site of the N132A enzyme are flexible to accommodate substrate or other ligands.

The N170A substitution also decreases k_cat and k_cat/K_M for all substrates (Table 1, Table 2, Table 3). We determined the X-ray structure of the N170A apo-enzyme. The structure is very similar to WT CTX-M-14 with the exception of the removal of the Asn170 side chain (Fig. 3, E and F and Table 4). Despite the removal of the hydrogen bond from Asn170, the catalytic water remains in essentially the same position in the N170A active site. Interestingly, a new water is near the former position of the Nδ of the Asn170 side chain in the WT structures and makes hydrogen bonds to Asn104, Asn132, and Glu166 and may stabilize the active-site structure and preclude more extensive changes such as observed with N132A. Nevertheless, substrate binding and catalysis are decreased for all substrates, indicating the presence of the Asn170 side chain is required for efficient hydrolysis.

The S130A and T235A substitutions greatly decrease k_cat and k_cat/K_M for cephalothin and cefotaxime and also greatly decrease k_cat/K_M for ampicillin but only modestly impact k_cat for ampicillin hydrolysis (Table 1, Table 2, Table 3). We determined the X-ray structure of the S130A apo-enzyme and found that it is superimposable with that of WT CTX-M-14 (PDB id: 1YLT) (Fig. 3, G–I). The only significant changes in the structure compared to WT is the absence of the hydroxyl group on residue 130 and the rotation of the Ser70 hydroxyl toward the gap created by the S130A substitution (Fig. 4, H and I). In principle, a water molecule could occupy the former position of the Ser130 hydroxyl group and partially compensate for its loss by acting in the proton shuttle between Lys73 and the β-lactam nitrogen leaving group. However, there is no water molecule in a suitable position in the S130A apo-enzyme structure (Fig. 3H). Since Ser130 is proposed to serve as a proton shuttle to protonate the β-lactam nitrogen leaving group during the acylation reaction, this result suggests that the Ser130 hydroxyl group is not necessary for protonation of the ampicillin nitrogen.

Figure 4.

X-ray crystal structures of the active site of WT CTX-M-14 β-lactamase and alanine mutants.A, diagram of WT CTX-M-14 active site. Relevant amino acid residues are labeled. Carbons are colored light blue, nitrogen is dark blue, and oxygen is red. Hydrogen bonds are indicated by dotted lines. B, structure of T235A mutant enzyme active site (salmon). C, structure alignment of WT and T235A mutant enzymes. Hydrogen bonds are not shown for clarity. Waters from CTX-M-14 WT are shown in red and waters from T235A are shown in salmon.D, active site of CTX-M-14 enzyme (light blue). E, active site of N106A mutant enzyme (white). F, structure alignment of WT and N106A mutant enzymes. G, active site of CTX-M-14 enzyme (light blue). H, active site of T215A mutant enzyme (yellow). I, structure alignment of WT and T215A mutant enzymes.

The X-ray structure of the apo-enzyme of the T235A mutant clearly shows the absence of the hydroxyl group at residue 235 (Fig. 4, A–C). In addition, the substitution results in several other changes in the active site. The hydrogen bond between the residue 235 hydroxyl and the hydroxyl of Thr216 is lost, resulting in a movement of Thr216 and other residues in the Gly213-Ser220 loop that resides above Ser70 in the active site (Fig. 4B). Further, there is a minor change in the position of the Lys73 Nζ that results in new hydrogen bonds of this atom with the main chain O of Ser130 as well as the carboxylate of Glu166. A more dramatic change is observed in the Val103-Asn106 loop where Asn106 is rotated and no longer forms hydrogen bonds with the main chain N and O of Val103, which results in a change in conformation of the loop so that Asn104 is rotated out of the active site and is not in a position to interact with the substrate. The basis for this long-range effect of the T235A substitution is not clear but it emphasizes that structural changes can reverberate through the active site, and changes beyond residue 235 could influence the effect of the substitution on kinetic parameters.

The N106A substitution results in moderate changes in k_cat/K_M, with values decreased less than 2-fold for cephalothin, 4-fold for cefotaxime, and 2-fold for ampicillin while k_cat values are unchanged for all substrates (Table 1, Table 2, Table 3). Asn106 resides at the base of the Val103-Asn106 loop and is not in a position to interact directly with the bound substrate (38) (Fig. 2). Rather, Asn106 forms hydrogen bonds with the main chain N and O of Val103. These bonds influence the conformation of the loop. Many class A enzymes, including the prototypical TEM-1 β-lactamase, contain Ser106, which has a different hydrogen-bonding pattern that results in an altered loop conformation where residue 104 is no longer in position to bind substrate (38). We previously showed that an N106S substitution in CTX-M-14 results in a TEM-like conformation of the Val103-Asn106 loop where Asn104 is pointed out of the active site and is associated with a 5-fold decrease in k_cat/K_M for cefotaxime hydrolysis due to a large increase in K_M (38). The N106A structure is superimposable with WT CTX-M-14 except for the 103-106 loop where the substitution eliminates the hydrogen bonds between Asn106 and the main chain of Val103, which alters the position of the loop such that the Asn104 side chain, which normally makes a hydrogen bond with bound cefotaxime, is not pointed toward the active site (Fig. 4E). This indicates the N106A substitution propagates a change in structure beyond the site of mutation and could explain the decrease in k_cat/K_M for cefotaxime hydrolysis.

Finally, Thr215 is located on the Gly213-Ser220 loop in the helix 10 region of CTX-M-14 that resides above the active site Ser70 and Ser130 residues (Fig. 2). It is not in position to interact directly with bound substrate and the T215A substitution has only modest effects on kinetic parameters for all substrates. The X-ray structure of the T215A apo-enzyme is superimposable with the WT CTX-M-14 structure (Fig. 4, G and H). Beyond the replacement of the Thr215 side chain with alanine, there are only minor changes in active-site residues (Fig. 4H), consistent with the modest effect of the mutation on the catalytic activity of the mutant.

Discussion

The CTX-M β-lactamases are an important source of resistance to extended-spectrum cephalosporins in Gram-negative bacteria (13, 15). Here we investigated the source of the high-catalytic activity of CTX-M toward cefotaxime and, more generally, the contributions of active-site residues to cephalosporin and penicillin hydrolysis. Alanine substitutions were introduced at 11 positions and determining kinetic parameters for hydrolysis of cefotaxime, cephalothin, and ampicillin were determined. The results revealed the active site of CTX-M-14 β-lactamase is highly optimized for cefotaxime hydrolysis in that all 11 positions contribute to either substrate binding or turnover or both. Further, for nearly all positions examined, the alanine substitution has the largest impact on cefotaxime hydrolysis, suggesting relaxed active site requirements for efficient hydrolysis of ampicillin compared to cephalosporin substrates. Interestingly, a similar pattern was observed in a deep mutational scanning study of the active site of the NDM-1 metallo-β-lactamase where sequence requirements for ampicillin hydrolysis were less stringent than for imipenem or cefotaxime (42). Thus, ampicillin hydrolysis appears to be an easier catalytic task for both a serine and a metallo-β-lactamase, despite differences in enzyme mechanism and structural fold-type between these enzymes.

The substitution of a side chain with alanine or a change in substrate structure affects k_cat/K_M when it results in the loss of stabilizing interactions between the enzyme and the transition state of the rate-limiting step. Because k_cat/K_M reflects rates up to formation of the acyl-enzyme, this could be the transition state for substate binding or for the acylation reaction. When viewed in the light of their effects on transition state binding (k_cat/K_M), several trends are observed. There is a reasonable correlation between the effects of the mutations on (ΔΔG_kcat/KM) for cefotaxime and cephalothin hydrolysis (Fig. 5). The slope is near one, indicating that the overall magnitude of the effects on transition state binding is very similar for the two cephalosporins tested. For ampicillin and cefotaxime, however, the correlation between ΔΔG_kcat/KM is also reasonably linear but the slope is smaller (0.58). On average, the effects of mutation on ampicillin hydrolysis are about 1.7 times smaller than their effects on cefotaxime or cephalothin. Thus, the transition state for cephalosporin binding and/or acylation is bound more tightly to the CTX-M-14 enzyme than that for ampicillin hydrolysis and the tighter transition state binding is correlated with increased sensitivity of cephalosporin hydrolysis to active site mutations.

Figure 5.

Free energy relationships. Linear free energy relationship between the effect of the CTX-M-14 alanine mutations on the k_cat/K_M for cefotaxime (labeled CTX on x-axis) and the k_cat/K_M’s for cephalothin (red) and ampicillin (blue). Nonweighted least square linear fits are shown.

The 11 active site alanine mutants can be placed in two groups based on their kinetic parameters with β-lactam substrates (Fig. 6). The first group consists of mutants with alanine substitutions at Lys73, Ser130, Asn132, Asn170, Lys234, and Thr235, which result in large decreases in both k_cat and k_cat/K_M values, suggesting these residues contribute strongly to the acylation and deacylation rates and also to substrate-binding affinity. Although not a part of this study, this group also includes Ser70 and Glu166, which serve as the nucleophile for attack on the β-lactam carbonyl carbon and the general base for deacylation, respectively (4, 6, 9, 43). The second group consists of mutants with alanine substitutions at Asn104, Tyr105, Asn106, Thr215, and Thr216, which result in only modest effects on k_cat but more significantly impact k_cat/K_M, particularly for cefotaxime, indicating they contribute to the substrate specificity of CTX-M enzymes by facilitating cefotaxime binding. As detailed below, previous studies have shown that Ser237 can be included in this group, as the S237A mutant enzyme displays a similar kinetic profile (44). It is noteworthy that this second group of residues, which influence substrate binding, are located more peripherally on the active site and surround the first group, which are key to both k_cat and k_cat/K_M (Fig. 6).

Figure 6.

Schematic illustration of the positions of CTX-M-14 active-site residues. Shown in red are positions that, when substituted, decrease both k_cat and k_cat/K_M (red) and thus are important for catalysis and substrate binding. Shown in blue are residues that, when substituted, lower only k_cat/K_M (blue) and contribute to substrate binding. A, ribbon diagram showing the 11 residues mutated in this study as well as Ser70 and Glu166, which have previously been shown to be critical for catalysis, and Ser237, which has been shown to contribute to substrate binding. B, surface diagram of CTX-M-14 with the residues that are highlighted in panel A shown in red and blue.

The residues Lys73, Ser130, Asn132, Asn170, Lys234, and Thr235 are proposed from structures, mutagenesis, molecular dynamics, and QM/MM simulations to participate in the mechanism of catalysis in class A β-lactamases (6, 7, 23, 24, 25, 26, 27, 28, 45, 46, 47). Alanine substitutions at these positions in CTX-M-14 result in very large decreases in k_cat and k_cat/K_M, with ΔΔG values >3 kcal/mol for cephalothin and cefotaxime hydrolysis. This is entirely consistent with the proposed key roles of these residues in catalysis. An interesting observation, however, is that the detrimental impact of these substitutions is partitioned differently between k_cat and k_cat/K_M depending on the substrate. This is particularly seen for ampicillin versus cephalosporins for the S130A and Thr235A mutants. For both of these mutants k_cat/K_M values are strongly decreased but k_cat is only modestly changed for ampicillin. In contrast, both k_cat/K_M and k_cat are strongly decreased for the cephalosporins, suggesting the substitutions impact different steps in the catalytic pathway for penicillins versus cephalosporins. A similar observation was made previously for TEM-1 for the analogous residue Ser235, where a S235A substitution strongly reduced k_cat for cephalosporins but not penicillins (27). This was attributed to the relatively higher importance of the hydrogen bond between Ser235 and the cephalosporin C4 carboxylate than the analogous interaction with penicillins and a similar conclusion is consistent with the results for CTX-M-14 (27).

Alanine substitutions at Asn104, Tyr105, Asn106, Thr215, and Thr216 have a minimal impact on k_cat but decrease k_cat/K_M, particularly for cefotaxime. These residues are located on either on the 103-106 or 213-220 active-site loop. We have previously shown that the conformation of the 103-106 loop is important to conserve interactions between the side chains of Asn104 and Tyr105 with cefotaxime (38). There are no direct interactions between the 213-220 loop and cefotaxime in the CTX-M-14 S70G/cefotaxime structure (44). However, a hydrogen bond between Thr216 and the Thr235 hydroxyl groups likely contributes to the proper positioning of Thr235 for hydrogen bonding to the C4 carboxylate of cefotaxime.

The Ser237 residue of CTX-M enzymes is also in the active site and in position to hydrogen bond with the C4 carboxylate of cefotaxime. In previous studies, we showed that a S237A substitution has no effect on k_cat but decreases k_cat/K_M by 3-fold compared to WT CTX-M-14 (44). Therefore, Ser237 should be classified along with Asn104, Tyr105, Asn106, Thr215, and Thr216 as a residue that contributes to cefotaxime binding but not the acylation or deacylation rates.

We also determined the X-ray structures of the N106A, S130A, N132A, N170A, T215A, and T235A apo-enzymes. The structures of the S130A, N132A, N170A, and T215A enzymes reveal only minor changes in the active site beyond the site of the alanine substitutions (Figs. 4 and 5). In contrast, the N106A and T235A structures reveal conformational changes at sites beyond the site of the alanine substitution. Asn106 forms hydrogen bonds with the main chain N and O of Val103, which stabilizes the structure of the Val103-Asn106 loop. We previously showed that the natural drug-resistance mutation N106S changes the hydrogen bonding pattern in the loop that results in a 180° flip in the peptide bond between Asn104 and Tyr105. This alters the conformation of the loop to move Asn104 out of the active site, while the carbonyl oxygen of Asn104 hydrogen bonds to Asn132 in a conformation identical to that observed in many class A enzymes, including TEM-1 (38). The N106A substitution eliminates the hydrogen bonds with Val103 but does not result in a flip in the 104-105 peptide bond. Therefore, Ser106 appears to be required for this change. However, the side chain of Asn104 is moved out of the active site in the N106A mutant as it is in the N106S enzyme. Consistent with this change, both the N106A and N106S substitutions result in a ∼5-fold decrease in the k_cat/K_M value for cefotaxime hydrolysis (38).

Further, the T235A substitution also causes active site structural changes away from the site of mutation. The elimination of the hydrogen bond between Thr235 and Thr216 results in movement of the Gly213-Ser220 loop that resides above Ser70 in the active site. In addition, the T235A substitution is associated with a conformational change of the Val103-Asn106 loop where Asn106 no longer forms hydrogen bonds with the Val103 N and O main chain and Asn104 is rotated out of the active site. It is not clear how the T235A substitution triggers this change but it does illustrate that the active site is a highly interconnected unit where changes at a position can propagate to more distant sites.

Alanine substitutions at active-site residues have also been observed to alter the positioning of other active-site residues for E. coli alkaline phosphatase (48). Interestingly, the residues mutated were in a cooperative unit with the residues that showed alternate positioning, i.e., co-mutation of the residues resulted in nonadditive effects on the catalytic activity. It is not surprising that alterations of active-site residues, even to small side chains such as alanine, can alter the positioning of other residues in that active sites are densely packed with functional groups to facilitate transition state stabilization and often contain extended hydrogen bond networks. Nevertheless, it is noteworthy that such structural effects do not impact the hydrolysis of all substrates equally in that, for CTX-M enzymes, such changes are much more damaging for cefotaxime than ampicillin hydrolysis. More generally, alanine substitutions at CTX-M active-site residues most strongly impact cefotaxime hydrolysis, regardless of whether the mutations structurally impact other residues.

It is likely that the clinical use of cefotaxime and other oxyimino-cephalosporins led to the emergence of CTX-M enzymes by gene transfer into pathogenic bacteria (13, 16). The genes for CTX-M enzymes naturally reside on the chromosome of Kluyvera species and the movement of the genes to transferable plasmids has been described (16, 49). It is not clear, however, what selective pressures led to the evolution of CTX-M enzymes, which are highly optimized for cefotaxime hydrolysis. Cefotaxime is a semisynthetic antibiotic, and presumably, Kluyvera species, which reside in the soil, water, and sewage, have not been exposed to cefotaxime in nature for the long-term evolution of this catalytic specificity. It is possible that natural product β-lactams resembling cefotaxime are found in the environment and led to the evolution of the observed CTX-M substrate specificity.

Experimental procedures

Strains and plasmids

Plasmid pET28a-CTX-M-14 was used for the expression of CTX-M-14 β-lactamase for purification and as the template plasmid for the construction of alanine mutants. This plasmid contains the gene for CTX-M-14 without the signal sequence. An N-terminal His-tag is present as well as a tobacco etch virus protease cleavage site. The construction of this plasmid has been described (50, 51). The E. coli XL1-Blue [recA1, endA1, gyrA96, thi-1, hsdR17, supE44, relA1, lac, [F9 proAB lacIq lacZΔM15, Tn10 (tetr)]] (Stratagene, Inc) was used as the host for the construction of CTX-M-14 alanine mutant variants. E. coli BL21(DE3) (fhuA2 [lon] omp Tgal (λDE3) [dcm] ΔhsdS λ DE3 = λ sBamHIo ΔEcoRI-B int::(lacI:: PlacUV5::T7gene1) i21 Δnin5) was used as the host for protein expression subsequent to purification of CTX-M alanine mutants.

Site-directed mutagenesis

The CTX-M-14 residues were mutated to alanine by site-directed mutagenesis using the QuikChange method and the oligonucleotides shown in Table S1 (52). The Pfu turbo polymerase was used to replicate the pET28a-CTX-M-14 plasmid to introduce the desired alanine mutations. The DpnI restriction enzyme was then added to the solution to remove the parental strands. The product of the QuikChange reaction was introduced into E. coli XL-1 Blue cells by electroporation (53). DNA sequencing was used to confirm the presence of the designed mutations and to ensure that no extraneous mutations were present in the CTX-M-14 gene.

Protein expression and purification

WT CTX-M-14 and the alanine mutant variants were expressed with an N-terminal His6-tag and purified after expression from E. coli BL21(DE3) as described (51). Briefly, E. coli containing the pET28a-CTX-M-14 plasmid were cultured in LB medium at 37 °C until A reached 1.0 to 1.2. Protein expression was induced by 0.2 mM isopropyl β-D-1-thiogalactopyranoside for 20 h at 23 °C. Cells were then collected by low-speed centrifugation and resuspended in buffer A (25 mM sodium phosphate, pH 7.4, 300 mM NaCl, and Xpert protease inhibitor mixture) supplemented with 20 mM imidazole. The E. coli cells were disrupted by sonication and cell debris was removed by centrifugation at 8000g for 15 min. The supernatant was loaded on a column-containing Talon resin (Takara Bio, Inc) and the flow-through was collected and reloaded on the column. This was repeated twice to increase protein binding to the resin. The column was washed with buffer A with 20 mM imidazole and protein was eluted with buffer A supplemented with 40, 60, and 80 mM imidazole, respectively. The protein fractions were combined, concentrated, and buffer-exchanged with buffer A using an Amicon Ultra-15 centrifugal filter unit (MilliporeSigma). The N-terminal His-tag of the purified protein was removed by treatment with tobacco etch virus protease overnight at 4 °C. The CTX-M-14 enzyme without the tag was purified by treatment with a nickel-Sepharose 6 Fast Flow resin (GE Health care Life Sciences) for 1 h at 4 °C. Protein purity and His-tag cleavage were assessed by SDS-PAGE followed by Coomassie Brilliant Blue staining.

Enzyme kinetics

Michaelis–Menten steady-state kinetic parameters for CTX-M-14 WT and alanine mutants were determined for ampicillin, cephalothin, and cefotaxime (Figs. S1–S3). The wavelengths and extinction coefficients used for detection were as follows: ampicillin, 235 nm, Δε = 900 M⁻¹ cm⁻¹; cephalothin, 262 nm, Δε = 7660 M⁻¹ cm⁻¹; cefotaxime, 264 nm, Δε = 7250 M⁻¹ cm⁻¹. Enzyme reactions for each substrate were performed at 25 °C in buffer containing 50 mM sodium phosphate (pH 7.0) and 1 μg/ml bovine serum albumin. Substrate hydrolysis was monitored using a Beckman Coulter DU 800 spectrophotometer (Beckman Coulter). Initial velocities of substrate hydrolysis were plotted as a function of substrate concentration and fit to the Michaelis–Menten equation using GraphPad Prism 9 (GraphPad Software, www.graphpad.com) to obtain k_cat and K_M values (Figs. S1–S3). In cases where the K_M was too high to obtain V_max, k_cat/K_M values were determined from progress curves using the equation, v = (k_cat/K_M)[E][S], where [S] << K_M (37). Error was calculated for k_cat/K_M values by propagating the error on the k_cat and K_M values using Equation 1 below where SEM indicates standard error of the mean. The change in free energy (ΔΔG) for k_cat and k_cat/K_M was calculated for each mutant using Equation 2 where R is the gas constant and T is temperature in kelvin. The error on the ΔΔG values was calculated using Equation 3.

$$\frac{{\mathrm{k}}_{\mathrm{c}\mathrm{a}\mathrm{t}}}{{\mathrm{k}}_{\mathrm{m}}}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r}=\frac{{\mathrm{k}}_{\mathrm{c}\mathrm{a}\mathrm{t}}}{{\mathrm{k}}_{\mathrm{m}}}\sqrt{{\left({\mathrm{S}\mathrm{E}\mathrm{M}}_{\mathrm{k}\mathrm{c}\mathrm{a}\mathrm{t}}{/\mathrm{k}}_{\mathrm{c}\mathrm{a}\mathrm{t}}\right)}^2+{\left({\mathrm{S}\mathrm{E}\mathrm{M}}_{\mathrm{K}\mathrm{m}}{/\mathrm{K}}_{\mathrm{m}}\right)}^2}$$ (1)

$$\Delta \Delta \mathrm{G}=\mathrm{R}\mathrm{T}\mathrm{I}\mathrm{n}\frac{\left(\frac{\mathrm{k}\mathrm{c}\mathrm{a}\mathrm{t}}{{\mathrm{K}}_{\mathrm{m}\mathrm{m}\mathrm{u}\mathrm{t}}}\right)}{\left(\frac{{\mathrm{k}}_{\mathrm{c}\mathrm{a}\mathrm{t}}}{{\mathrm{K}}_{\mathrm{m}\mathrm{w}\mathrm{t}}}\right)}$$ (2)

$$\Delta \Delta \mathrm{G}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r}=\Delta \Delta \mathrm{G}\sqrt{{\left({\mathrm{S}\mathrm{E}\mathrm{M}}_{\mathrm{k}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{w}\mathrm{t}}{/\mathrm{k}}_{\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{w}\mathrm{t}}\right)}^2+{\left({\mathrm{S}\mathrm{E}\mathrm{M}}_{\mathrm{k}\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{m}\mathrm{u}\mathrm{t}}{/\mathrm{k}}_{\mathrm{c}\mathrm{a}\mathrm{t}\mathrm{m}\mathrm{u}\mathrm{t}}\right)}^2}$$ (3)

X-ray crystallography

Crystal screening was performed using commercially available crystal screens PEGs and PACT from Qiagen and the hanging-drop vapor diffusion method. Crystal screens were standardly set up using 0.1 μl of 25 mg/ml CTX-M-14 mutant enzyme in 200 mM NaCl, 20 mM NaPO4, pH 7.4 with 0.1 μl of precipitant utilizing a TTP LabTech Mosquito instrument (TTP Labtech Ltd). The buffer conditions under which the crystals used for structure determinations were obtained as follows: CTX-M-14 N106A, 0.2 M Magnesium chloride, 0.1 M Tris pH 8, 20% (w/v), PEG 6000; S130A, 0.1 M Tris–HCl pH 8.5, 25% (w/v) PEG 6000; N132A-MES, 0.01 M Zinc chloride, 0.1 M MES pH 6, 20% (w/v) PEG 6000; N132A, 0.1 M Tris–HCl pH 8.5, 20% (w/v) PEG 10000; N170A, 0.1 M Sodium Hepes pH 7.5, 25% (w/v) PEG 8000; T215A, 0.2 M Calcium chloride, 0.1 M Hepes pH 7, 20% (w/v) PEG 6000; and T235A, 0.1 M SPG (succinic acid, sodium dihydrogen phosphate, and glycine in molar ratios 2:7:7) buffer pH 6, 25% (w/v) PEG 1500. A solution of 20% glycerol, 80% mother liquor was added to crystals as cryo-protectant and the crystals were picked and frozen using liquid nitrogen.

Diffraction data were collected at the Berkeley Center for Structural Biology using the Advanced Light Source synchrotron beamline 821. The reflection data were indexed, integrated, and scaled using the HKL2000, iMosflm, and the CCP4i Suite (54, 55). Molecular replacement was performed using CTX-M-14 (PDB entry 1YLT) as the structural model. The structures were refined further for several rounds with Phenix.refine and Coot density fitting (56, 57). The refined coordinates have been deposited in the Protein Data Bank. The data collection and refinement statistics for all structures are listed in Table 4. The UCSF Chimera and ChimeraX programs were used to construct structure figures (58, 59).

Data availability

Accession codes: Coordinates and structure factors have been deposited in the Protein Data Bank under accession codes CTX-M-14 N106A, 8DOE; S130A, 8DOD; N132A-MES, 8ELA; N132A, 8ELB; N170A, 8DPQ; T215A, 8DON; T235A, 8DP4. All relevant data associated with the paper are available upon request from the corresponding author.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.

Reviewed by members of the JBC Editorial Board. Edited by Chris Whitfield