Conjugation of Methotrexate-Amino Derivatives to Macromolecules through Carboxylate Moieties Is Superior Over Conventional Linkage to Amino Residues: Chemical, Cell-Free and In Vitro Characterizations

In this study, we examined the possibility of introducing methotrexate (MTX) to the carboxylate rather than to the ε-amino side chains of proteins. We found that MTX—amino compounds covalently linked to the carboxylate moieties of macromolecules, undergo unusual peptide-bond cleavage, with the release of the MTX amino derivatives from the conjugates. This event takes place at an accelerated rate under acidic conditions, and at a slower rate at physiological pH values. The glutamate portion of MTX is responsible for this behavior, with little or no contribution of the p-aminobenzoate-pteridine ring that is linked to the α-amino side chain of the glutamate. Carboxylate-linked Fmoc-Glu-γ-CONH-(CH2)6-NH2 undergoes hydrolysis in a nearly indistinguishable fashion. A free α carboxylate moiety is essential for this effect. Carboxylate linked Fmoc-glutamic-amide-γ-CONH-(CH2)6-NH2 undergoes no hydrolysis under acidic conditions. Based on these findings, we engineered a cysteine specific MTX containing reagent. Its linkage to bovine serum albumin (BSA) yielded a conjugate with profound antiproliferative efficacy in a MTX-sensitive glioma cell line. In conclusion, carboxylate linked MTX-amino derivatives in particular, and carboxylate linked R-α-GLU-γ amino compounds in general are equipped with‘built-in chemical machinery’ that releases them under mild acidic conditions.


Introduction
In previous studies we have focused on developing long-acting prodrugs that undergo reactivation in body fluids with desirable pharmacokinetic properties (reviewed in [1,2]). Our efforts were undertaken predominantly in the fields of diabetes and obesity [3][4][5][6][7][8]. The "prodrug concept" was introduced in the late 1950's [9] and is relevant also to the field of cancer therapy, in particular antibody-drug conjugates which have emerged in recent years as promising anticancer agents [10]. Low molecular weight anticancer drugs covalently linked to proteins are ordinarily inactive conjugates that can be engineered to undergo reactivation, either enzymatically or chemically upon reaching their target tissues [11].
To date, linkage of low molecular-weight anticancer drugs was conducted through the cysteine moiety or the ε-amino side chains of proteins [12]. In this study we initially linked MTXamino derivatives to the carboxylate moieties, of a macromolecular model compound, composed of polyethylene glycol of 20-40kDa, containing a single carboxylate, and then to proteins. We report here that such carboxylate-based macromolecule-MTX conjugates have unique, unexpected chemical and pharmacological features. These are described here in detail.

Synthesis of MTX-anhydride
Methotrexate (45.4 mg, 100μmoles) was dissolved in 0.9 ml dimethyl sulfoxide (DMSO) and 95 μl from a solution of 1M DCC in DMF (95 μmoles) was then added. The reaction was carried out for 3 hrs at 25°C. Dicyclohexylurea was removed by filtration. The MTX-anhydride formed was kept at 4°C.
Preparation of PEG 40 -CONH-(CH 2 ) 6 -NH 2 Solid PEG 40 -OSu (280mg~7 μmoles) was added to 5.0 ml hexamethylenediamine-2HCl (1M in 0.1M Hepes buffer pH 7.4 precooled to 0°C). The reaction was carried out for 5 hrs at 0°C, with stirring. The derivative obtained was dialyzed against H 2 O with several changes of H 2 O over a period of 3 days and then lyophilized.
Preparation of PEG 40 -CONH-(CH 2 ) 5 -COOH This derivative was prepared by the same procedure applied for PEG 40 -CONH-(CH 2 ) 5 -NH 2 except that hexamethylenediamine was replaced by 6-aminocaproic acid. The product thus obtained was extensively dialyzed against H 2 O and then lyophilized.
Synthesis of cysteine-specific-MTX-containing reagents [MAL-(CH 2 ) 2 -CONH-(CH 2 ) 2,3,6 -NHCO-MTX] Solutions of MTX-CONH-(CH 2 ) 2 -NH 2 , MTX-CONH-(CH 2 ) 3 -NH 2 or MTX-CONH-(CH 2 ) 6 -NH 2 (100 μmoles of each in 1 ml DMSO) were neutralized with one equivalent of DIPEA (100μmoles, 17.4 μl), and each combined with 120μmoles of MAL-(CH 2 ) 2 -COOSu (32mg in 0.3ml DMSO). The reaction was carried out for 3 hrs at 25°C with stirring. The products were precipitated with ethylacetate, washed 5 times with this solvent, twice with ethylether and dried in Vacuo. The products obtained were characterized for their calculated molecular weights by mass spectroscopy and for having a 1:1 MTX to maleimide molar ratio. MTX was quantitated by its absorbance at 372 nm (ε 372 = 7200) and the alkylating capacity of MAL was determined by reacting the reagents with reduced glutathione (GSH) and determining the amount of GSH consumed with 5,5'-dithio-2 nitrobenzoic acid (DTNB). No significant decrease in the alkylating capacity of the maleimido group was detected during storage of those compounds, in their solid state, over a period of several months.
Derivatization of BSA and HSA with the cysteine-specific MTXcontaining reagents 70mg (1 μmole) of HSA and BSA were dissolved on 2.0 ml of 0.05M Hepes buffer (pH 7.3). this was followed by the addition of cysteine specific reagents (i.e. MAL-(CH 2 ) 2 -CONH-(CH 2 ) 2,3,6 -NHCO-MTX, 3μmoles, 30μl from a solution of 100μmoles/ml in DMSO). The reaction was carried out for 1 hr at 0°C with stirring. The protein derivatives (20mg of each) were loaded on Sephadex-G-50 columns (14x1.5cm) that were equilibrated and run with 0.01M NaHCO 3 . The fraction corresponding to the protein derivatives were collected and lyophilized. This procedure removed unreacted as well as non-covalently adsorbed MTX derivatives to the protein.

Synthesis of Fmoc-L-glutamic acid anhydride
Fmoc-L-glutamic acid (369mg, 100μmoles) dissolved in 0.9 ml DMSO, and 95μl from a solution of 1M DCC in DMF (95 μmoles) was then added. The reaction was carried out for 3 hrs at 25°C with stirring. Dicyclohexylurea was removed by filtration. The product formed was kept at 4°C until used.
Synthesis of Fmoc-L-glutamic acid amide-γ-hexamethyleneamine F-moc-glutamic-amide (37mg, 100μmoles) was dissolved in 0.1ml DMSO and mixed with 100μl of 110μmoles hexamethylene-diamine 2HCl, that was neutralized with DIPEA. DCC, 0.1 ml from a solution of 1M in DMF (100μmoles) was then added. The reaction was carried out for 4 hrs at 25°C. DCU was removed by precipitation and the product obtained was precipitated with ether, washed 3 times and desiccated. The product was obtained in 64% yield. It has Table 1. Principle macromolecules conjugates, prepared and investigated in this study.

Conjugate number
Abbreviated structure a molar extinction coefficient of ε 280 = 7200±100. The calculated molecular weight is 466 Da, found by ESMS for M+H 467.29.
Preparation of PEG 40 -CONH-(CH 2 ) 6 -NHCO-Glu-Fmoc (compound I, Table 1) PEG 40 -CONH-(CH 2 ) 6 -NH 2 (80mg,~2 μmoles) was dissolved in 2.0 ml of 0.02M Hepes buffer (pH 7.4), cooled to 0°C and then 6 aliquots of Fmoc-glutamic-acid-anhydride (20 μl each, 2 μmoles) from a solution of 100 μmole/ml in DMSO were added over a period of 1 hr. The reaction mixture was incubated for an additional 2 hrs at 0°C, and then dialyzed against H 2 O with several changes of H 2 O over a period of 2 days and lyophilized. The product thus obtained, contains 0.7 mole Fmoc-glu per mole conjugate (40 mg solid material) as judged by its absorbance at 280 nm (ε 280 = 7200) and by its content of glutamic acid following acid hydrolysis and quantitative amino-acid analysis.
Partial purification of dihydrofolate reductase (DHFR) from chicken liver This was carried out by the procedure of Kaufman and Kemerer [14] with some modifications: Fresh chicken liver (20 grams) was cut into pieces, homogenized in 0.1M MnCl 2 and centrifuged. The supernatant was brought to pH 5-6 with HCl, and ZnSO 4 was added gradually to obtain a final concentration of 20 mM. The pH was elevated to pH 7.5-8 by the addition of solid NaHCO 3 . The precipitate formed (about 87% of total protein) was removed by centrifugation. The proteins remaining in the supernatant were precipitated with 4.54 M ammonium sulfate. The precipitate was then dissolved in 50 mM Tris-HCl (pH 7.5) containing 0.1M KCl, dialyzed for two days against the same buffer and then frozen in aliquots at -80°C until used. Partially purified DHFR prepared by this procedure has a specific activity of about 0.05 units/mg.

Cell-free enzymatic assay for DHFR
The assay was based on DHFR dependent reduction of DHF to tetrahydrofolate. In this process NADPH is oxidized to NADP, and the extent of the oxidation is monitored by the decrease in absorbance at 340 nm. Tubes containing 1.0 ml each of 0.05M Tris-HCl buffer (pH 7.5), 0.5M KCl, 0.137 mM NADPH and 0.187 mM DHF were incubated for 40 min at 25°C in the absence and the presence of partially purified DHFR (35μg/ml), and increasing concentrations of MTX or its derivatives. The absorbance at 340 nm was then monitored. In a typical assay, absorbance at 340 nm amounted to 1.7±0.06 and 0.65±0.03 in the absence and the presence of DHFR. MTX inhibits DHFR with an IC 50 value of 0.08 μM (80 picomoles/ml). A methotrexate derivative yielding an IC 50 value of 0.8 μM was considered as having 10% the inhibitory potency of the native folic acid antagonist.

Growth inhibition effects of BSA-MTX analogues
The MTX-sensitive glioma cell line CNS-1 was grown in 96 well plates in DMEM containing 10% fetal calf serum; 2mM L-glutamine, penicillin (100 units/ml) and streptomycin (0.1 mg/ ml) under humidified atmosphere containing 5% CO 2 . Cells were seeded at 1,000cells/well. 24hrs later, MTX or BSA-MTX analogues were added to each plate to give a final MTX or albumin-MTX analogue concentration as indicated in the text. Cell viability was measured after 72 h using a standard MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) assay as described before [15]. Experiments were repeated at least 3 times in quadruplicate. IC 50 values were calculated from dose response curves using a median-effect plot.

Statistical analysis
Statistical analyses were performed using the Prism 4 software. Data are presented as the means ± standard error of the mean (SEM) or standard deviation (STDEV). Differences between two groups were assessed by an unpaired t-test and among three or more groups by a one-way analysis of variance followed by Tukey's Multiple Comparison Test. A p-value of less than 0.05 was considered to be statistically significant.

Results
Covalently-linked MTX-hexamethylenamine to PEG 40 -------COOH, undergoes cleavage under acidic conditions In Table 1 we have listed the abbreviated structures of all macromolecular conjugates prepared and studied. In the first two, MTX was linked either to an amino side-chain (conjugate A) or a carboxylate moiety (conjugate B), located on a polyethylene glycol chain of 40kDa. The two conjugates (0.2μmol/ml) were dialyzed against one liter of 1 mM HCl (pH 3.0) and the absorbance at 305nm in the dialysis tubes was recorded as a function of time. In contrast to the amino-linked PEG 40 -MTX analogue (conjugate A), in conjugate B the absorbance at 305nm was substantially decreased within minutes and about 80% was dialyzed within a period of two hours (Fig 1). This finding suggested to us that the peptide bond connecting MTX-hexamethylenamine to this carboxylate-containing macromolecule, undergoes cleavage under acidic conditions. Fig 2 illustrates the structures of MTX and Fmoc-L-glutamic acid used in this study.

Characterizing the released MTX-hexamethylenamine
To verify this point, the conjugate (2μmoles in 1ml) was dialyzed against 15 ml of 1.0 mM HCl over a period of 24 hrs. The dialyzed fraction was collected, lyophilized and characterized. Table 2 summarizes several features of the collected fraction. It absorbs at 305 and 372 nm with similar extinction coefficients characterizing MTX or MTX-amino derivatives. It has 1.0±0.1 mol/mol glutamic acid, the corrected mass of MTX-hexamethylenamine (552.3±0.04 Da) and it inhibits DHFR with equipotent efficacy to that of MTX-hexamethylenamine (IC 50 = 190±7 picomoles/ml, Table 2, Fig 3).  The conjugate lost its absorbance in the visible region (at 305nm) when dialyzed at pH 3.0 with a t 1/2 value of 15±1 min, while at pH 7.4, t 1/2 value was 80±7 min (Fig 4). Thus, the rate of hydrolysis of such carboxylate-based conjugate is strongly acid-pH dependent. It exceeds by 5.3 fold the rate obtained at physiological pH.  Designing a cysteine-specific-MTX-containing reagent that releases the MTX-amino compound under acidic conditions Based on the above findings, we hypothesized that a cysteine-specific reagent consisting of maleimide (MAL) on one hand, MTX on the other, and the type of peptide bond which undergoes MTX-dependent hydrolysis, can be engineered. Such a reagent was prepared, linked to PEG 20 -SH (conjugate C, Table 1) and analyzed for its rate of hydrolysis at pH 3.0. As found for conjugate B, it was hydrolyzed with a t 1/2 value of 17±2 min and in a nearly quantitative fashion within a period of 2 hrs (Fig 5). We subsequently linked this reagent to the single cysteinyl moiety of HSA (conjugate D, Table 1). This conjugate was hydrolyzed as well at pH 3.0 ( Fig 5) and at pH 7.4 (not shown) in a nearly indistinguishable manner to that obtained with conjugate B.  Structural alterations in the cysteine-specific MTX-containing reagent: Effect on the rates of hydrolysis

Rates of hydrolysis of PEG
In order to obtain more structural data on this acid dependent chemical hydrolysis, we compared the hydrolysis rates of conjugates D, E, F, and G (Table 1) at pH 3.0. Those differ from each other in the type of the spacer arm connecting MTX to the cleavable peptide bond (Table 3). Conjugate E, having the shorter version of spacer arm (two ethylene groups), hydrolyzed at the fastest rate (t 1/2 = 6±0.7 min) exceeding in rate nearly 3 times that of conjugate D (six ethylene groups, t 1/2 = 17±2 min, Table 3). Conjugate F (3 ethylene groups) hydrolyzed at the slowest rate (t 1/2 = 58±3 min, Table 3). The lysine containing analogue G hydrolyzed at a  fast rate as well. In this analogue, the spacer arm differs from the others in containing additional carboxylate moiety, one bond away from the cleavable peptide bond (Tables 1 and 3).
Role of the non-glutamate portion of MTX and the contribution of its αcarboxylate moiety to the acid-dependent hydrolysis MTX is composed of L-glutamic acid linked through its α-amino side chain to p-aminobenzoate-pteridine ring (Table 1). In order to determine whether this non-glutamate portion of MTX is required for the observed acid-dependent hydrolysis, we have prepared PEG 30 --------COOH linked to Fmoc-L-glutamate-CONH-(CH 2 ) 6 -NH 2 (compound H, Table 1). As shown in Fig 5A and 5B, this analogue was hydrolyzed both at pH 3.0 (t 1/2 = 27±2 min) and at pH = 7.4 (t 1/2 = 65±3 min) in a nearly indistinguishable pattern to that obtained with the equivalent MTX-containing analogues (Figs 3 and 4). Thus, the non-glutamate portion of MTX has no role in facilitating hydrolysis and essentially can be replaced by any other R-group-located at the N-terminal side chain of glutamic acid. Finally, we have prepared PEG 30 ------COOH linked to Fmoc-glutamic acid-amide, an analogue in which the α-carboxylate moiety of glutamic acid is amidated (compound J, Table 1) and found that this analogue undergoes no hydrolysis under acidic conditions (Fig 6C). Thus, the free, non-derivatized α-carboxylate moiety of glutamate is essential for this chemical based hydrolysis.
Enhanced antiproliferative efficacy by hydrolyzable BSA-MTX conjugate in a CNS-1 glioma cell line  Table 1). This was compared to an analogue in which MTX was covalently linked to the lysine moieties of BSA (conjugate L, Table 1). This study was carried out in a rat-gioma cell line selected to be highly sensitive to the modulating action of MTX (Fig 7A). Both analogues are antiproliferatively active, however conjugate K shows IC 50 value of 140±17 nM, as opposed to an IC 50 value of 381±56 nm found for conjugate L (Fig 6B  and 6C and summarized in Table 4). Non-covalently linked MTX-CONH-(CH 2 ) 6 -NH 2 facilitates its antiproliferative effect with an IC 50 value of 67±4nM (Fig 7D, Table 4), indicating that conjugate K (IC 50 = 140nM) preserves a substantial amount (48%) of its antiproliferative efficacy following the covalent introduction of the chemically-releasable MTX-amino compound to this carrier protein.

Discussion
In general, the covalent conjugation of low molecular-weight drugs to proteins yields inactive products [16,17]. This phenomenon has profound clinical advantages if the conjugate releases the drug in its active form upon reaching its target tissues [18]. With regard to protein-methotrexate conjugates, this folate antagonist was linked through a variety types of linkers, either to a cysteinyl moiety or in cysteinyl-deficient proteins, to the ε-amino side chains of lysines [19,20]. Reactivation has been achieved, either by intracellular (lysozomal) proteolysis of the conjugate, yielding short MTX-containing active fragments, or by including a cleavable site in the spacer connecting MTX to the protein to permit cleavage by proteases present at the sites of inflammation [21].
In this study, we examined whether the covalent introduction of MTX, to the carboxylate moieties, rather than to the ε-amino side chains of proteins would be of value and offer potential clinical benefit. As a model for macromolecule-MTX conjugates, we first prepared polyethylene-glycol chains of 40 kDa containing either a single amino side chain or a carboxylate moiety. Methotrexate was covalently linked to each of those to obtain either amino-linked or carboxylate linked PEG 40 -MTX conjugates having the abbreviated structures summarized in Table 1.
Surprisingly, overnight dialysis of conjugate B (Table 1) against 1 mM HCl (pH 3.0) led to a total loss in the absorbance of this conjugate in the visible region (i.e. at 305 and 372 nm). This was found to be unique to this analogue, and was not shared with the amino-linked PEG 40 -MTX conjugate (Fig 1). We hypothesized that conjugate B (Table 1) uniquely undergoes acidbased chemical catalysis with the concomitant hydrolysis of the peptide-bond connecting PEG 40 ---------COOH to MTX-hexamethylenamine. This point was subsequently verified. The dialyzed fraction was collected and found to be identical to free MTX-hexamethylenamine by four different criteria, including an equipotent dose-dependent efficacy of DHFR inhibition (Fig 3, summarized in Table 2).
Based on these findings, we subsequently designed a cysteine-specific-MTX-containing reagent. It contains maleimide on one side, the MTX on the other side, and the peptide bond in the correct orientation to generate MTX-dependent-acidic dependent peptide bond cleavage. Its linkage to PEG 20 -SH or to the single-cysteinyl moiety of HSA, yielded conjugates that undergo acid-dependent hydrolysis in a similar quantitative and qualitative fashion, to that obtained with conjugate B (Table 1) respectively (Fig 4 and Table 3). Thus, the high reactivity of maleimide toward sulfhydryl moieties [22] enabled generation of specifically labeled conjugates that release MTX-amino compounds under acidic conditions. We next aimed to gain mechanistic insight into this peculiar chemical hydrolysis. For that reason, we first prepared a repertoire of HSA-MTX derivatives which differ in the length of the diamino spacer-arm bridging between MTX and the cleavable peptide bond. All were found to facilitate hydrolysis, some at a more accelerated rate (i.e. spacer arm H 2 N-(CH 2 ) 2 -NH 2 ) or at a slower rate (i.e. spacer arm H 2 N-(CH 2 ) 3 -NH 2 , Table 3). We next found that the non L-glutamic acid portion of MTX is not required to manifest this chemical hydrolysis. The P-aminobenzoate-pteridine ring can be replaced by other R-groups (Fmoc in this study) covalently linked to the α-amino side chain of glutamic acid (Fig 6). This observation extends this finding to other drug delivery systems as well as toward developing new versions for protracting the life-time of peptides and proteins in vivo (in preparation).
We a priory assumed that the non-protonated α-carboxylate moiety of the MTX-amino derivative is essential to catalyze this peptide-bond cleavage. This carboxylate moiety has a pK 1 of about 2.1 and it coincides well with the 5 fold accelerated rate of hydrolysis at pH 3.0 as opposed to pH 7.4 (Fig 4). Our preliminary attempts however to quantitatively derivatize the α side chain of MTX-CONH-(CH 2 ) 6 -NH 2 with glycinamide failed. This might be attributed to the low reactivity and/or steric hindrance of this moiety in the MTX-amino derivative. The MTX-CONH-(CH 2 ) 6 -NH 2 67±9 (1) Conjugate K in Table 1. (2) Conjugate L in Table 1. ** p<0.01 vs. BSA---NH-CO-MTX (student t-test).
doi:10.1371/journal.pone.0158352.t004 finding that carboxylate linked Fmoc-glutamate-amino derivatives also undergo hydrolysis (Fig 5A and 5B), and the availability of Fmoc glutamic acid-amide, permitted us to demonstrate that carboxylate-linked Fmoc-glutamic-acid-amide-CO-γ-NH-(CH 2 ) 6 -NH 2 undergoes no hydrolysis (Fig 6C). The mechanism for this unprecedented hydrolysis is not clear. We propose however a possible mechanism which is brought in Fig 8. At both pH values, a state of pseudo ring structure is initially obtained. At acidic conditions hydrogen bond is formed between the α-carboxyl of glutamic acid and the amino group of the cleavable bond (Fig 8B) whereas at neutral pH, the α-carboxylate anion facilitates a nucleophilic attack on the carbonyl residue of the cleavable peptide bond (Fig 8C). Both pathways eventually lead to this amidebond cleavage either at a faster (acidic) or a slower (neutral) fashion.
Interestingly enough, as observed here, all conjugates having an even number of ethylene groups (two or six) in the diamino spacer arm bridging MTX to the cleavable peptide bond (conjugates B, C, D, and E) hydrolyze at faster rates (t ½ values of 6-17 min at pH 3.0, Table 3). The exception was conjugate F which has an uneven number (three ethylene groups) in the spacer connecting MTX to the cleavable peptide bond and hydrolyze at significantly lower rate (t ½ = 58 ±3 min at pH 3.0, Table 3). This finding suggests that the pseudo ring structure formed in the case of conjugate F is considerably more stable, thus undergoing hydrolysis at a considerably lower rate. From practical stand point, the result implies that introducing MTX to albumin through a diamino spacer arm having an uneven number of ethylene groups is preferable if one wishes to obtain slower rates of hydrolysis.
Finally, we have studied in a MTX-sensitive CNS-1 glioma cell line, the antiproliferative efficacy of a cysteine specific BSA-MTX derivative, as opposed to the nonhydrolyzable (lysinelinked) BSA-MTX, and found that the former has 2.7 times higher antiproliferative efficacy and preserves nearly 50% the antiproliferative potency of the free, non-covalently linked MTXhexamethylenamine (Fig 7, Table 4). These findings strongly suggest that this MTX-amino compound is released effectively from the conjugate, following internalization. It is of interest to determine if the released MTX-amino compound is 'trapped' inside the cell and/or the chemically cleavable bond is susceptible to intracellular enzymatic hydrolysis as well, two intracellular events that can explain this enhanced level of cytotoxicity (under study).
With regards to conjugate L, the amide bond between MTX and the ε-amino side chains of lysines was documented to be chemically and enzymatically stable [23]. We therefore hypothesize that the reactivation of such conjugates are likely to take place only following internalization into cells and massive lysosomal proteolysis.
In conclusion, we show here using cell-free as well as cell-based assays that carboxylate linked MTX-amino derivatives in particular, and carboxylate linked R-α-GLU-γ amino compounds in general are equipped with a 'built-in chemical machinery' that releases them under mild acidic conditions and suggest that this non-orthodox conjugation might have beneficial therapeutic advantages in the field of protein-drug conjugates.