Mechanism for nitrogen isotope fractionation during ammonium assimilation by Escherichia coli K12

Abstract

Organisms that use ammonium as the sole nitrogen source discriminate between [¹⁵N] and [¹⁴N] ammonium. This selectivity leaves an isotopic signature in their biomass that depends on the external concentration of ammonium. To dissect how differences in discrimination arise molecularly, we examined a wild-type (WT) strain of Escherichia coli K12 and mutant strains with lesions affecting ammonium-assimilatory proteins. We used isotope ratio mass spectrometry (MS) to assess the nitrogen isotopic composition of cell material when the strains were grown in batch culture at either high or low external concentrations of NH₃ (achieved by controlling total NH₄Cl and pH of the medium). At high NH₃ (≥0.89 µM), discrimination against the heavy isotope by the WT strain (−19.2‰) can be accounted for by the equilibrium isotope effect for dissociation of NH₄⁺ to NH₃ + H⁺. NH₃ equilibrates across the cytoplasmic membrane, and glutamine synthetase does not manifest an isotope effect in vivo. At low NH₃ (≤0.18 µM), discrimination reflects an isotope effect for the NH₄⁺ channel AmtB (−14.1‰). By making E. coli dependent on the low-affinity ammonium-assimilatory pathway, we determined that biosynthetic glutamate dehydrogenase has an inverse isotope effect in vivo (+8.8‰). Likewise, by making unmediated diffusion of NH₃ across the cytoplasmic membrane rate-limiting for cell growth in a mutant strain lacking AmtB, we could deduce an in vivo isotope effect for transport of NH₃ across the membrane (−10.9‰). The paper presents the raw data from which our conclusions were drawn and discusses the assumptions underlying them.

Fractionation of heavy and light isotopes of nitrogen (¹⁵N vs. ¹⁴N), carbon (¹³C vs. ¹²C), and hydrogen (D vs. H) can provide information about metabolic pathways and reaction mechanisms within living organisms (1–4). For example, fractionation between ¹⁵N and ¹⁴N during incorporation of ammonium N in a single-celled organism like Escherichia coli is determined by the rate-limiting step for assimilating it into glutamate, the precursor of 88% of cellular nitrogen-containing material (5). All nitrogen assimilated into this central metabolic intermediate goes on to be incorporated into cell material. Transfers from glutamate to other molecules are direct. Although transfers from glutamine, including transfers to glutamate, involve deamidation, the NH₃ released is carried directly to the assimilatory catalytic site through a tunnel and hence, cannot be protonated or diffused away (6). The overall ratio of ¹⁵N to ¹⁴N in biomass is, thus, controlled by steps at or before assimilation of ammonium N into glutamate.

In E. coli, the proteins participating in early and potentially rate-determining steps in the incorporation of NH₄⁺ into glutamate are (i) AmtB, its only membrane channel for NH₄⁺ (7), (ii) glutamine synthetase (GS), the first enzyme of the high-affinity ammonium-assimilatory pathway, (iii) glutamate synthase [glutamine(amide) 2-oxoglutarate amino transferase (GOGAT)], and (iv) glutamate dehydrogenase (GDH), the first enzyme of the low-affinity ammonium-assimilatory pathway (Fig. 1) (reviewed in ref. 8). To study effects of these proteins on the in vivo fractionation of ammonium N, we used WT and genetic mutant strains in which one or more was lacking or defective and studied these strains at high or low external concentrations of NH₃. To decrease the concentration of external NH₃ and still achieve significant cell yield, we lowered both the total concentration of NH₄Cl and the pH of the medium. Although E. coli lives in the human gut, which is nitrogen-rich, it also survives in fresh and brackish water, in which supplies of available ammonium can be limited (9, 10). Moreover, it acidifies its own environment by fermentation. Accordingly, the conditions that we have chosen to study the behavior of E. coli at low external NH₃ are pertinent to its normal life cycle.

Fig. 1.

Schematic diagram of nitrogen assimilation from ammonium by E. coli K12.

Ammonium (NH₄⁺ + NH₃) is the optimal nitrogen source for E. coli (i.e., the source which yields most rapid growth). Ammonium (pK_a = 9.25) enters cells in two forms (Fig. 1). NH₃, which is ∼2% of total ammonium at pH 7.4, crosses the cell membrane by unmediated diffusion, a process that cannot be altered genetically. When the pH is decreased to 5.5, NH₃ is only ∼0.02% of total ammonium. NH₄⁺, which is the bulk of total ammonium at both pH 7.4 and pH 5.5, can enter the cells if and only if the AmtB channel is expressed and functional. Its expression is controlled at the transcriptional level and regulated largely by the free-pool concentration of glutamine in the cell interior (8), whereas its activity is controlled by the regulatory protein GlnK, largely in response to the free-pool concentration of the precursor metabolite 2-oxoglutarate, which is an intermediate in the tricarboxylic acid cycle (ref. 11 and references cited therein). Expression of AmtB increases as the glutamine concentration declines, and the channel is activated as the concentration of 2-oxoglutarate rises.

Within the cell, N is assimilated into glutamate, the organic precursor of most cellular nitrogen, by a high-affinity cycle and a low-affinity enzyme (Fig. 1). The high-affinity cycle is constituted by the exquisitely regulated GS and by GOGAT. The low-affinity enzyme is biosynthetic (NADPH-dependent) GDH (8). Use of 1 mol ATP per glutamate synthesized in the GS/GOGAT cycle drives assimilation of N even at extremely low concentrations of ammonium, but it is apparently detrimental when ammonium is abundant and energy is limiting (12). Both GS and GDH use NH₃ as their substrate, because bond formation requires the lone pair of electrons on the N of NH₃ (6, 13). Hence, NH₄⁺ must be dissociated to NH₃ + H⁺ before either enzyme can use it. An equilibrium isotope effect associated with this spontaneous process leads to depletion of ¹⁵N in NH₃ relative to NH₄⁺. Its magnitude is −19.2‰ (weighted mean, SE = 0.4‰, n = 3) (14). That is, at equilibrium, ¹⁵N/¹⁴N in NH₃ is 19.2 parts per thousand lower than that ratio in NH₄⁺.

To control the site of rate limitation, we used well-characterized mutant strains (Table 1). One such strain (ΔamtB) lacked AmtB. A second strain (AmtBΔC-term) had a poorly active AmtB channel, in which the normal membrane pores lacked the usual carboxyl-terminal cytoplasmic extensions (15, 16). A third strain (ΔgdhA::kan) lacked GDH, the low-affinity ammonium-assimilatory enzyme. The last two strains (gltD::kan and gltD::kanΔamtB) lacked GOGAT (Table 1) (8). We did not use a mutant lacking GS, because this strain is auxotrophic for glutamine and requires that glutamine be added to the medium in high concentrations, even in the presence of high concentrations of ammonium (17). We studied the mutant strains and their parental strain, which is a physiologically robust E. coli K12 WT (18, 19), under both ammonium-excess and -limiting conditions. We determined doubling time, cell yield, residual ammonium in the medium, and isotopic fractionation associated with incorporation of ammonium into cell material (Materials and Methods). When the absence or alteration of one protein increased the doubling time of the strain (i.e., decreased the growth rate), we could determine the rate-limiting step in transport or assimilation.

Table 1.

Bacterial strains

Strain: genotype	Phenotype
NCM3722: E. coli K12 WT	WT
NCM4199: amtB, tesB::kan	AmtBΔC-term*
NCM4453: gltD::kan	GOGAT−†
NCM4454: ΔgdhA::kan	GDH−†
NCM4590: ΔamtB	AmtB−
NCM4701: gltD::kanΔamtB	GOGAT−, AmtB−

All strains were constructed in the background of a physiologically robust E. coli K12 WT strain, NCM3722 (Materials and Methods) (18, 19).

^*Residues from position 382 on were deleted (15, 16).

^†Provided by Dalai Yan, Indiana University School of Medicine, Indianapolis.

Results

Parental WT Strain.

The doubling time was 50 min at all values of c₀, the initial external concentration of ammonia (note, NH₃ not NH₃ + NH₄⁺) (Fig. 2A and Table 2). For 0.89 ≤ c₀ ≤ 280 µM, measured isotopic fractionations (ε_b) ranged from −16.1‰ to −23.8‰ (mean and SD; −19.2 ± 2.6‰, n = 7), with negative values of ε indicating depletion of ¹⁵N in the biomass relative to the dissolved inorganic N in the medium (Table 2). At lower concentrations (c₀ ≤ 200 nM; i.e., at 1.0 or 0.5 mM total ammonium and pH 5.5), the isotopic fractionation decreased to −8.1‰ or −5.4‰, respectively (Fig. 2B and Table 2). AmtB is highly expressed, even when the concentration of external NH₃ is 5–10 times higher, namely 1 μM (5 mM total ammonium at pH 5.5) (11); however, its activity is not needed for optimal growth, and it is inhibited by the regulatory protein GlnK (20–22). AmtB is not expressed when external NH₃ is 10 μM (0.5 mM total ammonium at pH 7.4) or higher (11, 23).

Fig. 2.

(A) Growth and (B) determination of ε_b for assimilation of external ammonium (c₀ = 0.089 µM NH₃, 0.5 mM total NH₄Cl, 0.1% glucose, pH 5.5) into biomass in WT (□) and AmtB⁻ strains (△). Cultures in B were different from cultures in A, which were sampled more frequently.

Table 2.

Summary of cultures and measured isotopic fractionations

Experiment	Strain*	c0 (NH3)† (μM)	DT‡ (min)	n§	εb¶ (‰)
17	NCM3722	280	50	8	−20.4 ± 2.1
23	NCM3722	280	50	7	−16.8 ± 1.9
16	NCM3722	140	50	8	−16.1 ± 0.7
22	NCM3722	140	50	7	−18.2 ± 1.1
8	NCM3722	70	50	9	−19.1 ± 1.4
10	NCM3722	7	50	5	−23.8 ± 1.4
5	NCM3722	0.89	50	6	−20.0 ± 1.0
26	NCM3722	0.18	50	6	−8.1 ± 0.3
2	NCM3722	0.089	50	5	−5.4 ± 0.3
9	NCM4590	70	50	9	−19.9 ± 1.7
11	NCM4590	7	50	4	−22.2 ± 2.9
3	NCM4590	0.89	50	8	−20.3 ± 0.5
27	NCM4590	0.18	>100	5	−30.2 ± 0.7
1	NCM4590	0.089	>200	6	−29.9 ± 0.9
18	NCM4454	280	50	9	−22.1 ± 0.7
21	NCM4454	280	50	8	−19.4 ± 2.2
19	NCM4454	140	50	7	−25.4 ± 2.9
20	NCM4454	140	50	8	−18.8 ± 0.5
15	NCM4454	70	50	7	−19.3 ± 1.4
14	NCM4454	7	50	4	−20.4 ± 1.4
6	NCM4454	0.89	50	8	−23.9 ± 0.4
7	NCM4454	0.089	50	5	−6.1 ± 0.6
13	NCM4453	70	50	9	−10.3 ± 1.0
12	NCM4453	7	65	5	−11.2 ± 0.4
29	NCM4453	7	65	5	−11.5 ± 0.5
28	NCM4701	7	75	5	−10.3 ± 0.1
24	NCM4199	0.89	50	8	−23.3 ± 1.2
25	NCM4199	0.089	110	4	−17.2 ± 0.6
30	NCM4199	0.089	110	4	−17.8 ± 0.4

^*Strain number (phenotypes listed in Table 1).

^†Concentrations of NH₃ determined from total concentrations of NH₄Cl, which were 71-fold higher than the concentrations of NH₃ at pH 7.4 and 5,620-fold higher at pH 5.5. Cultures with c₀ ≤ 0.89 were grown at pH 5.5.

^‡Doubling time.

^§Number of points used in the determination of ε_b.

^¶Reported uncertainty is the SE of the slope derived from the regression described in Calculations.

GDH⁻ Strain.

The GDH⁻ strain (ΔgdhA::kan) lacks the low-affinity pathway for ammonium assimilation and hence, is completely dependent on the high-affinity pathway for synthesis of glutamate and glutamine. Its doubling time was indistinguishable from that of WT under all conditions of nitrogen availability, and its isotopic fractionations were very similar to those of the WT strain (Table 2). For 0.89 ≤ c₀ ≤ 280 µM, ε_b ranged from −18.8‰ to −25.4‰ (mean and SD; −21.3 ± 2.6‰, n = 7). For c_o = 89 nM, ε_b decreased to −6.1‰. Because the ranges of ε_b for cells having or lacking GDH overlap, these observations strongly support earlier reports that the GS/GOGAT cycle is the primary means for incorporating ammonium into biomass at all concentrations of NH₃ (12, 24, 25).

GOGAT⁻ Strains.

The GOGAT⁻ strain (gltD::kan) lacks the high-affinity ammonium-assimilatory cycle and depends on the linear, low-affinity pathway (biosynthetic NADPH-dependent GDH) for synthesis of glutamate. GS converts ∼12% of the glutamate product of GDH to glutamine to meet biosynthetic needs. When initial external concentrations of NH₃ were decreased from 70 to 7 μM, the doubling time of the strain increased from 50 to 65 min (Table 2 and Fig. S1A), and hence, the activity of GDH [E. coli has only a biosynthetic GDH (26)] was apparently rate-limiting for cell growth under these conditions. The weighted-mean isotopic fractionation was −11.2‰ (SE = 0.3‰, n = 3) at both concentrations of NH₃ (Fig. S1D). This strain did not grow at all at c₀ ≤ 1 μM (23).

In agreement with its dependence on the low-affinity enzyme GDH for synthesis of glutamate, gltD::kan has an abnormally low internal free-pool concentration of glutamate at low NH₃ (27). The gltD::kan strain also has an unusually high free-pool concentration of glutamine (27–29), the primary metabolic indicator of nitrogen sufficiency and hence, the primary metabolic regulator of the transcriptional response to nitrogen availability (8, 27). This strain fails to express a number of proteins under the control of nitrogen regulatory protein C (NtrC), which is active at low internal concentrations of glutamine (8, 30).

Although we presumed that AmtB, which is one of the proteins controlled by NtrC, was poorly expressed (23, 31), we constructed a double-mutant strain (gltD::kanΔamtB) to be certain that AmtB was completely absent. The doubling time of the gltD::kanΔamtB strain was slightly longer than the doubling time of the gltD::kan strain, but N and C utilization rates (Fig. S1 B and C) and the isotopic fractionation was unchanged at −10.3‰ (Table 2 and Fig. S1D).

AmtB⁻ and AmtB-Defective Strains.

Finally, the AmtB⁻ strain (ΔamtB) lacks the NH₄⁺ channel. For acquisition of ammonium, it must depend on unmediated diffusion of NH₃ across the cytoplasmic membrane. At high external concentrations of NH₃, both the doubling time of the ΔamtB strain and its isotopic fractionation were identical to those of the WT and the ΔgdhA strain (Table 2). In fact, under these conditions, the WT does not transcribe the glnKamtB operon (11, 23, 31). When the external concentration of NH₃ was decreased to 1 μM, the ΔamtB strain grew very slowly; its doubling time was initially 200 min, and growth slowed even more as the strain consumed ammonium and thereby decreased the external NH₃ concentration further (11, 23). The AmtB⁻ strain also failed to consume all of the NH₃ supplied in the media and it displayed a high C/N ratio in its biomass (see Supporting Information and Figs. S2 and S3). Under these conditions, the weighted-mean isotopic fractionation was −30.1‰ (SE = 0.6‰, n = 2), markedly larger than the value observed in other strains.

For the strain in which the AmtB protein was modified at the carboxyl terminal (AmtBΔC-term) (15, 16), both the doubling time and isotopic fractionation at c₀ = 0.89 µM did not differ from those of the WT, GDH⁻, and AmtB⁻ strains (Table 2). However, when c₀ was decreased to 89 nM, AmtBΔC-term had a doubling time of 110 min, more than two times as long as that of the WT and approximately one-half as long as that of ΔamtB. Under this condition, the activity of AmtB seemed to be rate-limiting for growth, and the weighted-mean ε_b was −17.6‰ (SE = 0.3‰, n = 2).

Process-Related Summary of Isotopic Observations.

Table 2 includes 11 different observations of ε_b for cells equilibrating NH₃ by diffusion and having both the low- and high-affinity pathways for its assimilation (Fig. 1). These observations include seven WT cultures with 0.89 ≤ c₀ ≤ 280 µM, three ΔamtB cultures with 0.89 ≤ c₀ ≤ 70 µM, and one AmtBΔC-term strain with c₀ = 0.89 µM. Values of ε_b range from −16.1‰ to −23.8‰. The weighted mean is −19.6‰ (SE = 0.7‰). There were seven different observations for cells equilibrating NH₃ by diffusion and lacking the low-affinity GDH pathway for its assimilation. For these ΔgdhA cultures, as for WT, 0.89 ≤ c₀ ≤ 280 µM. Values of ε_b range from −18.8‰ to −25.4‰. The weighted mean is −22.2‰ (SE = 0.9‰). In total, there are 18 different observations of ε_b for cells equilibrating NH₃ by diffusion and using the high-affinity GS/GOGAT cycle to incorporate ammonium N into organic molecules. Values of ε_b range from −16.1‰ to −25.4‰. The weighted mean is −21.1‰ (SE = 0.6). Variations of ε_b are not correlated with c₀ (r² = 0.13). Three values of ε_b were obtained for cells relying on the AmtB channel for transport of NH₄⁺ and assimilating ammonium N by GS + GOGAT. In those cases (two WT cultures with c₀ = 0.089 µM and c₀ = 0.18 µM and one ΔgdhA culture with c₀ = 0.089 µM), ε_b ranged from −5.4‰ to −8.1‰. In four cases, cells incorporated N using GS + GOGAT but had either no AmtB channel or an AmtB channel that was impaired by deletion of the C-terminal extensions. When AmtB was entirely absent (two cases, c₀ = 0.18 µM and c₀ = 0.089 µM), weighted-mean ε_b was −30.1‰. When AmtB was only modified (c₀ = 0.089 µM; two cases), the weighted-mean ε_b was −17.6‰. Finally, when cells lacked GOGAT (three cultures of ΔgltD and one culture of ΔgltDamtB::kan), values of ε_b varied from −10.3‰ to −11.5‰ with a weighted mean of −10.5‰ (SE = 0.2‰).

Discussion

When external concentrations of NH₃ exceed 0.18 µM, NH₃ can rapidly equilibrate across the cytoplasmic membrane of many bacteria by unmediated diffusion, and NH₄⁺ channels are not expressed (7, 11, 23, 32–34). The interior pool of NH₃ (in equilibrium with intracellular NH₄⁺) has only one input: diffusion of NH₃ from the external medium. Although it has three potential outputs—assimilation of NH₃ by GS, assimilation of NH₃ by GDH, and leakage from the cell—our results are not significantly affected by the absence of GDH (Process-Related Summary of Isotopic Observations), which reduces the outputs to two. The isotopic budget is summarized in Fig. 3A. Balancing the input against the outputs, we can write

where the δ-values are defined in Fig. 3A, ε_at is the isotope effect associated with transport of NH₃ across the membrane, and g is the fraction of the input that is incorporated into biomass. When equilibration of NH₃ across the membrane is rapid in comparison to the rate of assimilation, g → 0 (n.b.; g is not the amount of NH₃ that is assimilated but instead the fraction that is assimilated) and δ_ae = δ_ai. Because δ_b = δ_ai + ε_GS, the isotopic composition of the external NH₃ is related to the isotopic composition of the NH₃ within the cells by δ_ae = δ_b − ε_GS. Because essentially all external N is in the form of NH₄⁺, δ_ae = δ_e + ε_h, where δ_e is the measured isotopic composition of the N supplied to the medium and ε_h is the equilibrium isotope effect relating NH₄⁺ and NH₃. Finally, recalling that measured values of ε_b are equal to δ_b − δ_e, we obtain ε_GS = ε_b − ε_h. As noted above and summarized in Table 3, 18 experiments yielded ε_b ∼ −21.1‰. The 95% confidence interval of that value overlaps with the 95% confidence interval for ε_h. Accordingly, there is no evidence for fractionation by GS.

Fig. 3.

Flows of N across the cell membrane and within cells. Isotopic compositions are denoted by δ-terms, and isotope effects are denoted by ε-terms. In subscripts, a is NH₃, b is biomass, e is external, h is NH₄⁺ or protonation, i is internal, and t is transport. Isotopic compositions of the N fluxes are then given by expressions such as δ_he + ε_ht, which indicates the isotopic composition of the ammonium being transported by the AmtB channel. (A) Cells not using the AmtB channel to actively transport NH₄⁺. (B) Cells in which the AmtB channel is used.

Table 3.

Process-related isotope effects

Process	Related experiments	Result*
1. Assimilation of NH3 by GS	WT, c0 ≥ 0.89 µM, n = 7	95% Confidence intervals:
For g → 0	AmtB−, c0 ≥ 0.89 µM, n = 3	= −21.1 ± 1.3‰
εGS = εb − εh	AmtB∆C-term, c0 = 0.89 µM, n = 1	= −19.2 ± 1.7‰
	GDH−, c0 ≥ 0.89 µM, n = 7	εGS ∼ 0
2. Transmembrane transport of NH3	AmtB−, c0 < 0.2 µM, n = 2	= −30.1 ± 0.5‰
For g → 1		εat = −10.9 ± 0.7‰
εat = εb − εh
3. Transport of NH4+ by AmtB	WT, c0 = 0.089 µM, n = 1	= −5.5 ± 0.3‰
εht = εb + (1 − g)εat	GDH−, c0 = 0.089 µM, n = 1	For g = 0.2 ± 0.05, εht = −14.1 ± 0.8‰
4. Transport of NH4+ by altered AmtB	AmtB∆C-term, c0 = 0.089 µM, n = 2	= −17.6 ± 0.3‰
= εb + (1 − g)εat		For g = 0.5 ± 0.1, = −23.1 ± 1.2‰
5. Assimilation of NH3 by GDH	GOGAT−, AmtB−, c0 = 7 µM, n = 1	= −10.5 ± 0.2‰
εGDH = εb − εh	GOGAT−, c0 ≥ 7 µM, n = 3	εGDH = 8.8 ± 0.4‰

^* denotes weighted mean. Indicated uncertainties are SEs except for process 1, where 95%confidence intervals are specified.

An alternative interpretation, with g appreciably greater than zero, is not tenable. Specifically, 18 experiments yielding ε_b ≅ ε_h could then be explained only if (i) g, ε_GS, and ε_at happened in all cases to have values satisfying the relationship g = ε_GS/(ε_GS − ε_at) and (ii) g was independent of the external concentration of NH₃. Additionally, a well-documented experimental study of nitrogen and carbon isotope effects associated with glutamine synthetase from E. coli found a near-zero nitrogen isotope effect of −0.7 ± 0.6‰ (mean and SD, n = 7) (35).

It is unlikely that GS limits growth, because it is synthesized in excess when supplies of NH₃ are plentiful; also, its catalytic activity is regulated downward by covalent modification (36–38). It is more likely that the flux of N into glutamate, the most plentiful intermediate in central nitrogen metabolism, is limited by the capacity of GOGAT. No isotopic fractionation is associated with GOGAT, because practically all of the amide N in gln is transferred to 2-oxoglutarate to produce glu.

When external NH₃ concentrations are below 0.89 µM, E. coli K12 depends on the ammonium channel AmtB to maintain an optimal growth rate. Cells lacking this channel (ΔamtB) depend on uncatalyzed transport of NH₃ across the membrane. At c₀ ≤ 0.2 µM, growth of the ΔamtB strain is extremely slow, and the rate decreases as the external concentration of NH₃ declines (Fig. 2A and Table 2). The mass balance described by Eq. 1 applies, but the conditions differ from the conditions just discussed. Instead, g ≠ 0, and because GS imposed no fractionation (even when supplies of NH₃ were abundant), we know that δ_ai = δ_b and ε_GS = 0. Making these substitutions and simplifying, we obtain δ_ae = δ_b − gε_at. Substituting δ_ae = δ_e + ε_h and ε_b = δ_b − δ_e leads to ε_b = gε_at + ε_h. At the limit in which growth is limited by transport of NH₃ and g → 1, ε_at = ε_b − ε_h. If the slowest growing cultures (Table 2, experiments 1 and 27) represent that case, ε_at = −30.1 + 19.2 = −10.9 ± 0.7‰ (SE from combining 0.5 and 0.4 in quadrature). This relatively large value suggests that transport of NH₃ across the membrane is limited by some process other than simple diffusion. Polar interactions within the membrane may play a role. Rishavy and Cleland (39) commented that the isotope effect in a related case “could easily be 2%” (i.e., 20‰; almost two times the value estimated here) (39).

For cells with a normal AmtB channel (WT or ΔgdhA strain) and c₀ = 0.089 µM, ε_b was observed as low as −5.4 ± 0.3‰, much lower than the value of a strain lacking the channel (ΔamtB). The corresponding mass balance is shown schematically in Fig. 3B and expressed mathematically in Eq. 2:

Here, ε_ht is the isotope effect associated with transport of NH₄⁺ by the AmtB channel, and substitutions introduced above have been adopted where appropriate (δ_he = δ_e, δ_ai = δ_b). Simplifying gives ε_b = ε_ht − (1 − g)ε_at. A recent quantitative study of AmtB function (11) indicates that g is ∼ 0.2. For ε_b = −5.4‰, adopting ε_at ∼ −10.9‰ (see above), we find ε_ht = −14.1‰. If an uncertainty of 0.05 is assigned to g, the estimated SE of ε_ht is 0.7‰. Notably, the reduced fractionation at low values of c₀, a condition that may be encountered in nature, derives from not only fractionations associated with the AmtB channel but also, an interplay between fractionations associated with ε_ht and ε_at.

When the WT strain was grown with a slightly higher c₀ = 0.18 µM, the observed fractionation increased to −8.1‰. The AmtB channel also functions at this NH₃ concentration, because an AmtB⁻ strain continues to grow suboptimally. Hence, Eq. 2 applies. Using ε_ht = −14.1 ± 0.7‰ and solving for g, we find g = 0.45 ± 0.1. When AmtB functions, the internal ammonium concentration is held constant, and hence, there is less leakage of NH₃ at higher external NH₃ concentrations (11).

For cells with an altered AmtB channel lacking the cytoplasmic C-terminal extensions (16), ε_b = −17.6 ± 0.3‰ (weighted mean and SE) at c₀ = 0.089 μM. The fraction of N assimilated by GS is 0.5 (11), and solving as above yields (where a prime is used to denote the altered channel). If the uncertainty assigned to g is doubled (to 0.1), the SE increases only to ±1.2‰. Hence, the isotope effect for the mutant AmtB channel is significantly larger than that for the WT channel. The mutant channel is known to lack coordination between the function of its individual monomers and have other unusual properties (40).

To make E. coli dependent on the low-affinity pathway for assimilation of ammonium, we inactivated GOGAT. This inactivation eliminates the GOGAT cycle and makes the organism dependent on biosynthetic GDH. At an external NH₃ concentration of 10 μM, GDH activity already limits the growth rate of the GOGAT⁻ strain, and the strain does not grow at all at c₀ = 1 μM. The fractionation observed for the GOGAT⁻ strains with c₀ ≥ 7 µM is −10.5‰. Assuming that NH₃ inside and outside the cell was in equilibrium, such as for the WT, ΔgdhA, and ΔamtB strains at the same concentrations, it follows that internal NH₃ was depleted in ¹⁵N by 19.2‰ relative to external dissolved inorganic N and therefore, that the observation of ε_b = −10.5‰ requires inverse fractionation of ¹⁵N by GDH (i.e., enrichment of the product relative to the reactant) with ε_GDH = 8.7 ± 0.4‰. An inverse isotope effect has also been reported for bovine liver GDH (41).

Conclusion.

Our studies of E. coli K12 have yielded in vivo isotope effects as summarized in Table 3. To our knowledge, the isotope effect for transport of NH₃ is the first for a biological membrane. A previous measurement was made in vitro with a membrane filter (14).

The equality of ε_b for the ΔghdA and WT strains at all external concentrations of NH₃ confirms the finding of Yuan et al. (24, 25) that the GOGAT cycle is the major means for assimilation of NH₃ by E. coli K12 at not only low but also, high concentrations. Although the GOGAT cycle is widespread in bacteria and archaea (8, 42, 43), whereas the occurrence of biosynthetic GDH seems to be more restricted (44), important examples of organisms naturally lacking GOGAT have recently come to light (45). Determining ε_b for the abundant ocean archaean Nitrosopumilis maritima will be particularly interesting, because it not only lacks GOGAT and depends on GDH for ammonium assimilation but also, oxidizes ammonium extracellularly as its primary energy source.

We hope that the present results will help future workers to correlate environmental genomic data with isotopic variations observed in nature.

Comparison with Earlier Work.

Studying the γ-proteobacterium Vibrio harveyi, a close evolutionary relative of E. coli K12, Hoch et al. (46) also found that isotopic fractionation between external ammonium and cell material varied with external ammonium availability. Now, by using AmtB⁻ strains, we have determined that the decrease in ε_b from ∼ −20‰ to −4‰ that Hoch et al. (46) observed as external ammonium was dropped from an initial concentration of ∼500 to ∼25 μM (pH 7.4), and it did, as Hoch et al. (46) proposed, depend on the activity of an active ammonium channel (46). Specifically, ε_b ∼ −5‰ results from (i) acquisition of NH₄⁺ rather than NH₃ by the growing cells, (ii) an isotope effect associated with transport of NH₄⁺ by the AmtB channel (ε_ht ∼ −14‰), (iii) equilibration of NH₄⁺ and NH₃ inside the cell, (iv) an absence of isotopic fractionation during assimilation of NH₃ by GS, and (v) leakage of ¹⁵N-depleted NH₃ from the cells.

Given the very large ε_b characteristic of E. coli AmtB⁻ strains at low external NH₃ (−30‰), which seems to result from rate-limiting diffusion of NH₃ across the cytoplasmic membrane, it is tempting to speculate that the fractionation observed in V. harveyi as the external concentration of ammonium decreased from 182 to 107 μM in a single experiment [ε_b = −26.5‰ (46)] may indicate the precise range of external ammonium concentrations at which unmediated diffusion of NH₃ becomes limiting in WT V. harveyi just before activation of its AmtB channel. In E. coli, expression of AmtB occurs in response to a decrease in the internal free-pool concentration of glutamine, whereas activation requires, in addition, an increase in the pool concentration of its precursor metabolite 2-oxoglutarate (11). The latter occurs at lower external NH₃ concentrations than the former. Activation requires release of the inhibitory gating protein GlnK (20–22).

Finally, we think that the decrease in ε_b from −21‰ to −14‰, which Hoch et al. (46) observed above 5 mM external ammonium, is an artifact of growth inhibition (doubling time increased from the optimal of 84 min to 138 min at high ammonium). Whatever the explanation, the ε_b of −14‰ cannot be characteristic of GDH as Hoch et al. (46) proposed, because the NADH-dependent GDH that they characterized is a catabolic enzyme (46). The genome sequence of V. harveyi is now available and it apparently lacks a biosynthetic, NADPH-dependent GDH. Moreover, we find that the biosynthetic, NADPH-dependent GDH of E. coli contributes very little to ammonium assimilation even at 20 mM ammonium, their highest concentration and our highest concentration.

Materials and Methods

Bacterial Strains and Cultures.

NCM3722 (18) was the parental strain for all genetic mutant strains used in this work (Table 1). Additional details of strain construction are in SI Materials and Methods. For growth experiments, bacterial cultures were grown on the minimal medium by Neidhardt et al. (47) in MOPS buffered medium, pH 7.4, with 0.1% glucose as sole carbon source and NH₄Cl as sole nitrogen source. For experiments at pH 5.5, cultures were additionally adapted to low pH in minimal medium buffered with MES at pH 5.5. Growth and doubling time were determined by measuring OD at 420 nm.

Ammonia Assay.

Residual ammonium in cell-free supernatants was assayed in a GDH catalyzed reaction (AA0100 kit; Sigma). In the assay, 2-oxoglutarate is reduced to l-glutamate by GDH using ammonium as substrate and NADPH as the cofactor providing reducing equivalents. Oxidation of NADPH is measured by a change in absorbance at 340 nm.

Sample Preparation and Isotopic Analyses.

Bacterial cell samples were taken at various points during growth and removed from the supernatant by high-speed centrifugation. The cell-free supernatant was frozen at −80 °C for later measurement of residual ammonium, glucose, and final pH. The cells were washed two times in medium without additional glucose or ammonium and dried in air overnight. Uniform amounts of 2 mg dry weight, yielding 0.8 mg carbon and 0.2 mg nitrogen, were transferred into preweighed tin capsules (part number 240–053-00; Costech Analytical Technologies, Inc.). Capsules containing only the reactant glucose and ammonium chloride used in the media were also prepared. All samples were analyzed at the University of California at Berkeley Center for Stable Isotope Biogeochemistry. δ¹⁵N and %N were determined by using a Europa 2020 mass spectrometer coupled to an automatic elemental analyzer (48). The isotopic abundance parameters are defined as follows:

where ¹⁵R ≡ ¹⁵N/¹⁴N and the isotopic standard for nitrogen is N₂ in air, for which ¹⁵R = 0.0036765 (49). Values of δ¹⁵N express the relative difference between the isotope ratio in the sample and the standard, expressed in parts per thousand (‰). A value of δ¹⁵N = −12.2‰, for example, indicates that ¹⁵R_sample is 0.0036316. The precision of the analyses, expressed as the SD of a single observation and based on five pairs of duplicates and four sets of triplicates collected during the analyses (thus, 13 degrees of freedom), is 0.06‰.

Calculations.

The objective of the isotopic analyses is to determine ε_b, the overall isotope effect associated with the assimilation of N. The isotope effect is most simply expressed by the isotopic difference between the starting pool of inorganic N in the medium and the first increment of biomass formed after inoculation. In mathematical terms, the isotopic difference is expressed as a ratio of isotope ratios. Because the isotope ratios are very similar, a notation is used that expresses the difference in terms of parts per thousand:

where ¹⁵R_e0 is the ratio of ¹⁵N to ¹⁴N in the initial medium and ¹⁵R_b0 is the same ratio in the first increment of biomass. As growth proceeds, the measured isotopic compositions of both the medium and the biomass change as a result of preferential transfer of either ¹⁵N or ¹⁴N (depending on the sign of the isotope effect) from the medium to the biomass. Measurements of ε_b must take this change into account.

Here, we use the regression of δ_b on [f/(1 − f)] · lnf (1), thus fitting the observations to a linear equation of the form:

where δ_b is the measured δ¹⁵N of the biomass, δ₀ is the measured δ¹⁵N of the medium at t = 0, and f is the fraction of ammonium unused. If, for example, ε_b = −18.8‰, it indicates that ¹⁵N is assimilated and used to produce biomass 18.8 parts per thousand more slowly than ¹⁴N.

Values of δ₀ vary between experiments depending on the batch of NH₄Cl that was used. Specific values are, for experiments 1–5, 3.25 ± 0.19‰ (mean and SD, n = 8); for experiments 6–15, 1.43 ± 0.06‰ (n = 3); for experiments 16 and 17, 1.12 ± 0.19‰ (n = 2); for experiments 18 and 19, 0.96 ± 0.19‰ (n = 2); for experiments 20 and 21, 1.15 ± 0.21‰ (n = 2); for experiments 22–24, 1.36 ± 0.09‰ (n = 3); and for experiments 25–29, 0.91 ± 0.08‰ (n = 5).

Uncertainties in ε_b, calculated from the variance about the regression and expressed as SEs of the slope, are reported in Table 2 and range from 0.1‰ to 2.9‰. Where weighted means are reported, the weighting factor is the inverse variance. The reported SEs of weighted means are conventional or dispersion-corrected, whichever is greater. Uncertainties reported for calculated isotope effects are derived by conventional propagation of errors.