An Escherichia coli System Expressing Human Deoxyribonucleoside Salvage Enzymes for Evaluation of Potential Antiproliferative Nucleoside Analogs

Abstract

Deoxyribonucleoside salvage in animal cells is mainly dependent on two cytosolic enzymes, thymidine kinase (TK1) and deoxycytidine kinase (dCK), while Escherichia coli expresses only one type of deoxynucleoside kinase, i.e., TK. A bacterial whole-cell system based on genetically modified E. coli was developed in which the relevant bacterial deoxypyrimidine metabolic enzymes were mutated, and the cDNA for human dCK or TK1 under the control of the lac promoter was introduced. The TK level in extract from induced bacteria with cDNA for human TK1 was found to be 20,000-fold higher than that in the parental strain, and for the strain with human dCK, the enzyme activity was 160-fold higher. The in vivo incorporation of deoxythymidine (Thd) and deoxycytidine (dCyd) into bacterial DNA by the two recombinant strains was 20 and 40 times higher, respectively, than that of the parental cells. A number of nucleoside analogs, including cytosine arabinoside, 5-fluoro-dCyd, difluoro-dCyd, and several 5-halogenated deoxyuridine analogs, were tested with the bacterial system, as well as with human T-lymphoblast CEM cells. The results showed a close correlation between the inhibitory effects of several important cytostatic and antiviral analogs on the recombinant bacteria and the cellular system. Thus, E. coli expressing human salvage kinases is a rapid and convenient model system which may complement other screening methods in drug discovery projects.

In mammalian proliferating cells, the deoxyribonucleoside salvage pathway is initiated by the cytosolic enzymes TK1 and dCK. The physiological roles of these two enzymes are to phosphorylate Thd and dCyd to the corresponding monophosphates, which can be further phosphorylated by other enzymes to di- and triphosphates for incorporation into DNA. In addition, these two enzymes can phosphorylate many pharmacologically important nucleoside analogs. The expression of TK1 is highly cell cycle dependent, while dCK is expressed in a tissue-specific fashion, and this leads to large variations among the capacities of different cells to phosphorylate deoxynucleosides and their analogs. Furthermore, there are considerable differences in the substrate specificity of dCKs from different mammalian species, making rodent cells a poor model for the development of nucleoside analogs intended for use in humans (1).

A number of animal viruses, e.g., the herpesvirus family, encode deoxynucleoside kinases which can accept a much broader spectrum of substrates than human enzymes, and this is the basis for the efficiency and selectivity of several of the most important antiviral drugs used today (4, 9, 14).

Deoxynucleoside salvage in prokaryotes shows large variations, especially with regard to activation of deoxyadenosine (dAdo), dGuo, and dCyd (30). Lactobacillus acidophilus contains three distinct deoxynucleoside kinases: a TK, a dCyd/dAdo kinase (dCK/dAK) and a dGuo/dAdo kinase (dGK/dAK), and the structure of the operon for the latter two enzymes has recently been elucidated (22, 23). Bacillus subtilis expresses a TK, a dCK/dAK, and a dGK and has a salvage pathway similar to that in animal cells (27, 28). Escherichia coli has only one deoxynucleoside kinase, i.e., TK. Due to the lack of dCK, E. coli metabolizes dCyd by deamination catalyzed by the inducible cytidine/dCyd deaminase. The product of this reaction, deoxyuridine (dUrd), is further metabolized either by TK or by thymidine phosphorylase. This explains the inability of E. coli to incorporate dCyd, in contrast to Thd, into its DNA (19).

The most popular systems for screening pharmacologically interesting nucleoside analogs include in vitro cell culture systems and direct assays with pure or partially purified target enzymes such as kinases, catabolic enzymes, and DNA/RNA polymerases (1, 26). Although these are reasonable and successful approaches, the difference between the species and cell types mentioned above and the variability and cost involved in cell culture study cause considerable problems. The purification of nucleoside metabolic enzymes is also costly and complicated, and in vitro assays are always subject to criticism of their in vivo relevance.

In this report, we describe a bacterial system based on genetically modified E. coli cells which may be used to determine the toxicity of nucleoside analogs to proliferating cells by monitoring the selective inhibition of bacterial growth caused by these analogs. Mutants of E. coli defective in the major deoxypyrimidine catabolic enzymes and, in one case, also in TK were constructed, and the cDNA for human dCK or TK1 under the control of the lac promoter/repressor was introduced. Degrees of growth inhibition of the bacteria caused by nucleoside analogs with and without the expression of the human enzymes were compared. A number of deoxynucleoside analogs were tested with this system, as well as with human T-lymphoblast CEM cells, which is a much-used model system in antiviral and antitumor research (33). Several cytostatic and antiviral analogs were shown to perform their inhibitory effects on the growth of engineered bacteria at levels similar to those at which human CEM cells were inhibited.

MATERIALS AND METHODS

Abbreviations.

AraC, 1-β-d-arabinofuranosylcytosine; AZT, 3′-azido-2′,3′-dideoxythymidine; CAFdA, 2-chloro-2′-fluoroarabinosyl adenine; CdA, 2-chloro-2′-deoxyadenosine; ddC, 2′,3′-dideoxycytidine; dFdC, 2′,2′-difluoro-2′-deoxycytidine; EC₅₀, compound concentration at which bacterial growth is inhibited by 50%; FIAU, 1-(2′-deoxy-2′-fluoro-β-d-arabinofuranosyl)-5-iodouracil; FLT, 3′-fluoro-2′-deoxythymidine; FMAU, 1-(2′-deoxy-2′-fluoro-β-d-arabinofuranosyl)-5-methyluracil; HIV, human immunodeficiency virus; IPTG, isopropyl-β-d-thiogalactopyranoside. TK, thymidine kinase; dCK, deoxycytidine kinase; dCyd, deoxycytidine; dUrd, deoxyuridine; Thd, deoxythymidine; dAK, deoxyadenosine kinase; 5-F-dCyd, 5-fluorodeoxycytidine; dGuo, deoxyguanosine; 5-F-dUrd, 5-fluorodeoxyuridine; 5-Br-dUrd, 5-bromodeoxyuridine; 5-I-dUrd, 5-iododeoxyuridine; 5-Cl-dUrd, 5-chlorodeoxyuridine.

Bacterial strains and growth media.

The bacterial strains used were all derivatives of E. coli K-12 and are listed in Table 1. Strains SØ5110, SØ5282, and SØ5286 were constructed by P1-mediated transduction as described by Miller (24). Luria broth was used as rich medium (24). The minimal medium was AB medium (11) supplemented with 0.2% glucose, 0.2% vitamin-free Casamino Acids, and 1-μg/ml thiamine. When required, tryptophan was added at 50 μg/ml and uridine was added at 20 μg/ml (82 μM). Antibiotics were used at the following final concentrations: ampicillin, 100 μg/ml; tetracycline, 10 μg/ml; kanamycin, 30 μg/ml.

TABLE 1.

Bacterial strains used in this study

Strain	Genotype	Source or reference
CSH1	trp thi lacZ rpsL	24
HO602	F−supF relA spoT rpsL lamB metB deoD udp::Tn5 gsk-3	B. Hove-Jensen
HO959	Hfr Δ(lac) thi Δ(deoCABD) zjj-202::Tn10 upp udp ton Φ80r	B. Hove-Jensen
KY895	F−tdk-1 ilv	16
MC1061	F− Δ(ara-leu)7696 galE15 galK16 Δ(lac)X74 rpsL hsdR2 mcrA mcrB1	25
SØ1452	cdd::Tn10	B. Mygind
SØ5110	MC1061 cdd::Tn10	By P1 (SØ1452)
SØ5218	SØ5110/pTrcHUMdCK	Present work
SØ5282	trp thi lacZ rpsL Δ(deoCABD) zjj-202::Tn10 udp::Tn5	From CSH1 by P1 (HO959) and P1 (HO602)
SØ5286	thi lacZ rpsL Δ(deoCABD) zjj-202::Tn10 udp::Tn5 tdk-1	From SØ5282 by P1 (KY895)
SØ5288	SØ5286/pTrcHUMTK1	Present work
SØ5292	SØ5286/pTrc99-A	Present work

Nucleoside analogs.

All of the analogs used in this study were purchased from Sigma, except FIAU and FMAU, which were synthesized and provided by J. Fox at the Memorial Sloan-Kettering Cancer Institute, and FLT, which was a gift from N. G. Johansson of Medivir AB, Huddinge, Sweden.

Plasmid constructions and expression.

Plasmid pTrc99-A (Pharmacia) was used throughout as a cloning and expression vector.

(i) pTrcHUMdCK.

The pET-3d expression vector containing the coding sequence of the human dCK cDNA was obtained from B. Mitchell at the Department of Pharmacology, University of North Carolina (10). It contained a unique NcoI site overlapping the dCK start codon and a unique BamHI site immediately 3′ of the stop codon. The entire dCK coding region was recloned as a 780-bp NcoI/BamHI fragment into the multiple cloning site of pTrc99-A, yielding pTrcHUMdCK. In this construct, the dCK cDNA was transcribed from the vector-borne IPTG-inducible trc promoter, as the lacI^q gene encoding the lac repressor was also expressed from pTrc99-A. Translation of dCK was initiated from the lacZ ribosomal binding site located 6 bp upstream of the NcoI cloning site. Plasmid pTrcHUMdCK was transformed into cytidine deaminase (cdd)-negative E. coli SØ5110. As E. coli is naturally devoid of dCK, the expression from the plasmid was the sole source of this enzyme in the recombinant cells.

(ii) pTrcHUMTK1.

The coding sequence of human TK1 cDNA was amplified by PCR using pTK11, obtained from Bradshaw and Deininger (3), as the template. The 5′ sense primer, 5′CGGAATTCAAGGAGGCGTAATGAGCTGC, contained an EcoRI site (bold) and a good E. coli ribosome binding site (underlined) upstream of the TK1 start codon (italics), and the 3′ reverse complement primer, 5′CGGGATCCTCAGTTGGCAGGGC, had a BamHI site (bold) immediately following the stop codon (italics) when read on the complementary sequence. Amplified DNA was digested with EcoRI and BamHI and cloned into the EcoRI/BamHI sites of pTrc99-A, yielding pTrcHUMTK1. As with pTrcHUMdCK, the transcription of the cloned cDNA was from the IPTG-inducible trc promoter in the vector, whereas translation was initiated from the ribosome binding site located within the 5′ primer. pTrcHUMTK1 was introduced into SØ5286, yielding SØ5288. SØ5286 is unable to catabolize Thd and dUrd due to mutational inactivation of Thd phosphorylase (deoA) and uridine phosphorylase (udp). In addition, SØ5286 carries the tdk-1 mutation inactivating the endogenous E. coli TK (16).

Incorporation of dCyd and Thd by recombinant E. coli.

Incorporation experiments were carried out as described by Karlström (19), with the following modifications. An overnight culture was diluted to an A₆₀₀ of 0.01 with fresh medium containing 1 mM IPTG, and the culture was grown with shaking at 37°C until an A₆₀₀ of 0.2 was reached. A final concentration of 5 μM radioactive dCyd or Thd was added, and the growth was continued for another 30 min. The incorporation was linear for up to 60 min, and 0.2 ml of the culture was withdrawn and mixed with ice-cold 5% trichloroacetic acid. After centrifugation, the pellets were resuspended in 1 M KOH and incubated at 37°C for 20 h. The samples were then transferred to glass fiber filters (Whatman) and washed, and the radioactivity was counted with Beckman liquid scintillation system LS3800 to determine the incorporation of labelled nucleoside into bacterial DNA.

Determination of dCK and TK activities in bacterial extracts.

Enzyme levels in crude bacterial extracts were determined as follows. A 1.5-ml sample withdrawn from the culture for the in vivo incorporation experiments described above was centrifuged, and the cells were resuspended in 0.5 ml of buffer A (50 mM Tris [pH 7.6], 1 mM EDTA, 50 mM NaCl, 0.5 mM phenylmethylsulfonyl fluoride, 0.5% Nonidet P-40, 2 mM dithiothreitol, 1-mg/ml lysozyme). The samples were incubated at room temperature for 30 min, and after centrifugation (15,000 × g for 20 min), the supernatant was used for enzyme assays. The assays were performed as described previously (12).

Growth inhibition by nucleoside analogs.

A fresh overnight culture, prepared by inoculating a single colony from a petri plate into 10 ml of minimal medium containing antibiotics, was diluted into fresh minimal medium supplemented with antibiotics and 1 mM IPTG to achieve an initial A₆₀₀ of 0.01. This culture was subsequently divided into tubes, 1 ml in each, and 20 μl of analog solution was added to each tube. The tubes were cultured at 37°C with shaking for 3 to 4 h, until the control tube, to which no analog was added, reached an A₆₀₀ of about 0.6. The A₆₀₀s of all of the tubes were measured, and the relative growth of each sample was calculated by comparing its A₆₀₀ to that of the control tubes, which was set as 100%.

RESULTS AND DISCUSSION

TK and dCK activities in extracts of engineered bacteria and in vivo incorporation of Thd and dCyd into bacterial DNA.

To demonstrate that human dCK and TK1 were successfully expressed in E. coli, we determined the levels of dCK and TK in the crude extracts of the recombinant strains. As shown in Table 2, strain SØ5218, which contains cDNA for human dCK, exhibited significant dCK activity in the presence of IPTG, while parental strain SØ5110 had no detectable dCK activity. The dCK level in the induced bacteria is close to that in human CEM cell extracts (33). Both SØ5110 and SØ5218 demonstrated a low level of TK activity due to the endogenous enzyme, while SØ5288, harboring the cDNA for human TK1, expressed a 200-fold higher TK level when IPTG was added. As expected, there was no detectable TK activity in TK-deficient E. coli SØ5292, the parental bacterium of SØ5288 (Table 2). The large difference between the levels of recombinant dCK and TK1 in induced bacteria may be due partially to the higher specific activity of human TK1 protein than dCK (1). Furthermore, there might be a difference in the regulation of human TK1 and dCK expression in E. coli, since the specific activity of dCK in noninduced SØ5218 was approximately 40% of that in induced bacteria, whereas the presence of IPTG resulted in a 34-fold increased TK level in SØ5288 (results not shown).

TABLE 2.

Rate of dCyd and Thd incorporation into E. coli DNA and enzyme levels in extracts from IPTG-induced bacteria^a

Strain	Incorporation rateb		Sp actc
	dCyd	Thd	dCK	TK
SØ5110	19	288	<1	94
SØ5218	850	215	160	23
SØ5292	NDd	25	ND	<1
SØ5288	ND	540	ND	22,660

^aValues are means of triplicate determinations. 

^bValues are in picomoles per minute per unit of cell mass. 

^cValues are in picomoles per minute per milligram of protein. 

^dND, not determined. 

The in vivo rate of dCyd and Thd incorporation into DNA by the recombinant strains was determined, and the results showed a 20- to 40-fold difference between the expressing and control strains (Table 2), confirming that the human enzymes were expressed in E. coli. The rates of Thd incorporation into all of the TK-containing bacteria (SØ5110, SØ5218, and SØ5288) were similar to each other and were also close to that reported earlier for E. coli B (19). The very high level of TK in SØ5288 did not lead to significantly higher Thd incorporation compared to SØ5110 and SØ5218, indicating that the formation of dTMP from exogenous Thd is not the rate-limiting step in this pathway in E. coli.

Growth inhibition of SØ5218 by dCyd analogs.

As mentioned above, E. coli does not express dCK but has the ability to deaminate dCyd through the activity of cytidine deaminase (cdd). To increase the capacity of the bacteria to salvage dCyd, as well as its analogs, cdd-deficient strain SØ5110 was used as the host for the plasmid carrying cDNA for human dCK. A number of dCyd analogs were tested with this system. All growth experiments were carried out with IPTG-induced bacteria so that maximal expression of recombinant human dCK was achieved. It was shown that the introduction of human dCK cDNA made the cells sensitive to compounds such as dFdC (Fig. 1) and AraC (Fig. 2), both of which analogs are known to be toxic to animal cells (32, 35). Several other pyrimidine analogs have also been tested with this system, and the results showed that dFdC (C-2), a much-used anti-tumor drug (32, 33a), had the most potent inhibitory effect (Table 3). 5-Azacytidine (C-4), which also is a known cytostatic ribocytidine analog (2), showed an inhibitory effect on both SØ5218 and the control strain, most likely as a result of the activity of the bacterial cytidine-uridine kinase. Only minor inhibition of growth was observed with ddC (C-1), which is an effective anti-HIV compound (8, 26).

FIG. 1.

Sensitivity to dFdC of E. coli expressing human dCK. Symbols: •, growth of SØ5110 in the presence of 1 mM IPTG; ■, growth of SØ5218 in the presence of 1 mM IPTG.

FIG. 2.

Sensitivity to AraC of E. coli expressing human dCK. Symbols: •, growth of SØ5110 in the presence of 1 mM IPTG; ■, growth of SØ5218 in the presence of 1 mM IPTG.

TABLE 3.

Effects of antiviral and cytostatic dCyd analogs on the growth of E. coli expressing human dCK^a

Compound code	Analogb	Relative growth (%) of:
		SØ5218	SØ5110
C-1	ddC	85	98
C-2	dFdC	1	103
C-3	2′-Azido-dCyd	15	99
C-4	5-Azacytidine	5	6
C-5	5-methyl-dCyd	92	98
C-6	5-F-dCyd	27	102
C-7	5-Br-dCyd	86	91
C-8	5-I-dCyd	100	98
C-9	AraC	48	102
C-10	5-Fluoro-AraC	97	102
C-11	5-Chloro-AraC	100	101
C-12	2,2′-Anhydro-AraC	57	93
C-13	CdA	72	95
C-14	CAFdA	27	73

^aValues are means of duplicate determinations. 

^bAll data were obtained with each analog at 20 μM, except for CdA and CAFdA, whose concentration was 5 μM. 

The EC₅₀s of the active cytosine analogs were determined with the bacterial system, as well as with a human T-lymphoblast CEM cell culture. The EC₅₀s of dFdC (C-2) and 5-F-dCyd (C-6) for bacterial and CEM cells were very similar (Table 4). dFdC has excellent activity against several forms of solid tumors, and the incorporation of difluorodeoxy-CTP into DNA leads to chain termination and DNA repair failure (33a).

TABLE 4.

EC₅₀s of dCyd analogs for E. coli expressing human dCK and for cultured human CEM cells

Compound code	Analog	EC50 for:
		Bacteria	CEM cells
C-1	ddC	600 μM	10 μM
C-2	dFdC	15 nM	10 nM
C-3	2′-Azido-dCyd	100 nM	10 μM
C-6	5-F-dCyd	15 nM	20 nM
C-9	AraC	6 μM	20 nM
C-12	2,2′-Anhydro-AraC	50 μM	20 nM

The capacity of AraC (C-9) and 2,2′-anhydro-AraC (C-12) to inhibit the dCK-expressing bacteria was almost 2 orders of magnitude lower than that of dFdC (C-2), while with CEM cells, all three compounds exhibited similar levels of toxicity (Table 4). AraC (C-9), an effective agent in the treatment of acute leukemia, is also utilized by dCK but with lower efficiency than 2′-deoxynucleosides. A limiting factor in mammalian cells was proven to be the membrane uptake system (35). One likely explanation for the discrepancy between the sensitivity of bacterial cells and that of human CEM cells is that the transport of arbinosylcytosine into E. coli is quite inefficient. It has been reported that the dCyd transport system in E. coli cannot be blocked by addition of excessive AraC (20, 35). Furthermore, preliminary experiments measuring AraC uptake into E. coli indicated that it was less than 5% compared to dCyd uptake and was not affected by the addition of nonradioactive cytidine (29). Thus, the low level of AraC uptake in E. coli could most likely explain the reduced inhibitory capacity of this compound and other arabinosyl analogs in the bacterial model system.

The inhibitory effect of 2′-azido-dCyd (C-3) and 2,2′-anhydro-AraC (C-12) on the recombinant E. coli is also different from that observed with CEM cells (Table 4). The latter compound is most likely converted to AraC during the assay, and therefore its effect is similar to that of AraC (C-9) (17). 2′-Azido-dCyd (C-3) requires activation by dCK to form the monophosphate, and its diphosphate has been shown to be toxic to mammalian cells due to the inhibition of ribonucleotide reductase, resulting in decreased DNA precursor levels (18). Although this compound might act in a similar way in the recombinant E. coli, the difference in the EC₅₀s for bacterial and CEM cells implies that bacterial ribonucleotide reductase is more sensitive to 2′-azido-dCyd than is the mammalian enzyme. However, further studies on the metabolism of 2′-azido-dCyd and its effects on target enzymes are required to prove this hypothesis. The anti-HIV compound ddC is not very toxic to CEM cells and is even less toxic to bacterial cells (Table 4). In the latter case, differences in transport may likewise explain the discrepancy in toxicity between the bacterial system and human CEM cells.

Several purine analogs known to be substrates for dCK were also tested in the bacterial system. With CdA (C-13) and CAFdA (C-14), which are both efficient antileukemic nucleoside analogs (5, 6), it was found that the growth of both SØ5218 and control strain SØ5110 was inhibited at high drug concentrations (Table 3). However, with CAFdA, which is known to be more resistant to nucleoside bond cleavage, dCK-dependent inhibition at lower analog concentrations was observed (Table 3 and unpublished results). The results indicate that catabolism of purine analogs may be involved in the dCK-independent toxicity of certain nucleosides, and further study to clarify the mechanism is in progress.

Growth inhibition of SØ5288 by Thd analogs.

Wild-type E. coli metabolizes Thd and dUrd either anabolically through the action of TK (tdk) or catabolically through phosphorolytic cleavage catalyzed by Thd phosphorylase (deoA) and, less efficiently, by uridine phosphorylase (udp) (31). Thus, SØ5286, in which all three of the enzymes mentioned above were eliminated by mutagenesis, was used as a host for human TK1 cDNA. It was expected that SØ5286 or its derivative SØ5292, which contains the vector pTrc99-A with no insertion, would be resistant to 5-F-dUrd, but this was not the case (Fig. 3a). A possible reason for this is that uridine kinase may phosphorylate 5-F-dUrd sufficiently to cause the toxicity. Accordingly, addition of uridine to the culture made the bacteria resistant to this analog, presumably by competing with 5-F-dUrd for uridine kinase. When the cDNA for human TK1 was introduced into SØ5286, on the other hand, the resulting strain, SØ5288, was sensitive to 5-F-dUrd, irrespective of the presence of uridine (Fig. 3b). This shows that human TK was expressed in SØ5288 and that the enzyme was able to activate 5-F-dUrd to the toxic form 5-fluorodeoxy-UMP (15).

FIG. 3.

Sensitivity of E. coli strains to 5-F-dUrd in the presence of 1 mM IPTG. a, growth of SØ5292; b, growth of SØ5288. Symbols: □, with 5-F-dUrd only; ■, with 0 to 500 nM 5-F-dUrd and 82 μM uridine.

The effects of several antiviral and cytostatic dUrd and Thd analogs have been tested with SØ5288 and SØ5292. Besides 5-F-dUrd (U-1), most of the 5-halogenated dUrd analogs, as well as the antiviral compounds AZT (U-5), FLT (U-9), and FMAU (U-12), showed selective inhibition of bacterial growth (Table 5). The inhibition curves were determined for several analogs with the bacterial system, as illustrated for AZT in Fig. 4, and the EC₅₀s for six analogs are presented and compared with the values for human CEM cells in Table 6. Except for AZT (U-5), the other five analogs were found to have similar effects on both the recombinant E. coli and CEM cells. AZT is a much-used anti-HIV compound (26), and it is known to cause side effects mainly by inhibiting bone marrow-derived stem cells. The inhibition of SØ5288 by AZT occurred at considerably lower concentrations than those observed with many different human cell culture lines, including CEM cells (26, 33).

TABLE 5.

Effects of antiviral and cytostatic dUrd and Thd analogs on the growth of E. coli expressing human TK1^a

Compound code	Analog	Concn (μM)	Relative growth (%) of:
			SØ5288	SØ5292
U-1	5-F-dUrd	1	15	40
U-2	5-Cl-dUrd	50	69	103
U-3	5-Br-dUrd	20	30	85
U-4	5-I-dUrd	20	45	77
U-5	AZT	20	21	89
U-6	ddTb	50	98	98
U-8	3′-F-dUrd	20	87	92
U-9	FLT	50	76	104
U-10	AraTc	10	100	101
U-11	FIAU	50	90	104
U-12	FMAU	50	72	104

^aValues are means of duplicate determinations. 

^bddT, 2′,3′-dideoxythymidine. 

^cAraT, 1-β-d-arabinofuranosylthymine. 

FIG. 4.

Sensitivity to AZT of human TK1- and bacterial TK-expressing E. coli strains in the presence of 1 mM IPTG. Symbols: •, growth of SØ5292; ■, growth of SØ5288; ○, growth of SØ5110; □, growth of SØ5218.

TABLE 6.

EC₅₀s of dUrd and Thd analogs for E. coli expressing human TK1 and for cultured human CEM cells

Compound code	Analog	EC50 for:
		Bacteria	CEM cells
U-1	5-F-dUrd	8 nM	10 nM
U-3	5-Br-dUrd	5 μM	10 μM
U-4	5-I-dUrd	20 μM	8 μM
U-5	AZT	500 nM	150 μM
U-8	FLT	60 μM	7 μM

It has been shown that AZT is toxic to many members of the family Enterobacteriaceae, including E. coli (7, 13), as was also observed in this study (Fig. 4). A positive correlation between the TK1 levels in CEM cells and their sensitivity to AZT has been observed (33). On the other hand, we observed here that the sensitivity to AZT of E. coli SØ5110 and SØ5218, both of which express the endogenous TK, is very close to that of TK-deficient E. coli expressing human TK1 (SØ5288), although the TK level in the latter is 200-fold higher (Fig. 4). This observation might be explained by the fact that the rate-limiting step in the anabolism of AZT is usually the one catalyzed by thymidylate kinase (21). Since the TK level in E. coli is lower than that in mammalian cells, it is likely that in this case, thymidylate kinase in E. coli is likewise rate limiting. It is also interesting that FIAU (U-11), a good anti-hepatitis B analog, does not show inhibition of bacterial growth, as does its derivative FMAU (U-12), although both are equally good substrates for human TK1 (34).

There are several important differences between E. coli and mammalian cells in the overall biosynthesis, transport, and catabolism of nucleosides and nucleotides. Of particular interest in the present study is the biosynthesis pathway for dUMP, the ultimate thymidine nucleotide precursor. In E. coli, the predominant route is through deamination of dCTP, whereas in mammalian cells, it occurs via deamination of dCMP. Since the pyrimidine deoxyribonucleotide metabolism of B. subtilis resembles that of animal cells to a larger extent (27), a similar system but with B. subtilis as the host for the human enzymes may be a better model system. Recently, an approach similar to the one described here was undertaken by introducing several of the herpesvirus TKs into TK⁻ E. coli hosts (7). The bacteria expressing viral TKs became highly sensitive to several pyrimidine nucleoside analogs, and large differences in the sensitivities of the engineered bacteria were found to be correlated to the properties of the viral enzymes. The availability of an E. coli system expressing the cellular TK described here, as well as those expressing viral TKs, should enhance future drug discovery projects and could be the basis for selection of mutant forms of the recombinant enzymes.