Glycine-modified polyamidoamine dendrimers: synthesis and structural characterization using nuclear magnetic resonance, ion-mobility mass spectrometry and capillary electrophoresis

Abstract

We present here the preparation and the structural characterization of first-generation ammonia-cored polyamidoamine (PAMAM) dendrimers modified with glycine residues. Chemical modification of the dendrimer was done to increase the biocompatibility of these compounds, known to be effective delivery agents for drugs or genes. Fully modified PAMAM [Gly₆G1(N)] on the one hand and partially modified [Gly_nG1(N), with n = 0 to 6)] on the other hand were obtained depending on the experimental conditions. The resulting modified PAMAM dendrimers have to be cautiously characterized to understand and interpret their physico-chemical and biochemical properties as well as to control their chemical design. The structural characterization was carried out using ion mobility spectrometry-mass spectrometry (IM-MS), multistage tandem mass spectrometry (MSⁿ), accurate mass measurements by high resolution–mass spectrometry (HR-MS), two dimensional nuclear magnetic resonance (NMR) and capillary electrophoresis (CE). Characteristic fragmentation patterns for these compounds were obtained from ESI/MSⁿ (with n = 2 to 4) experiments. IM-MS and CE analysis showed that a single component was mainly obtained for the complete grafting experimental conditions while a distribution of oligomers was produced for partially grafted products. The physical separation of Gly_nG1(N) oligomer ions was achieved in the gas phase (IM-MS) as well as in the condensed phase (CE). Besides, the collision cross sections (CCS) were estimated by IM-MS and compared to theoretical values. Then, the glycine grafting yield for Gly_nG1(N) PAMAM was determined by both NMR and IM-MS experiments.

---

Introduction

PAMAM dendrimers are a class of synthetic polycationic polymers showing a unique spatial structure of globular shape. They are composed of three architectural elements: a core, repeating units and terminal functional groups.¹ The reiteration of monomer addition yields dendrimers of higher generation. The size, shape and surface functionalities of these compounds are well-controlled and their synthesis yields narrow molecular weight distributions, if prepared carefully. It has been reported that high generations of PAMAM dendrimers can form complexes with DNA molecules offering the possibility to use such dendrimers as non-viral transfection systems for gene therapy.² However, the cationic surface of PAMAMs, appropriate for binding chemical entities, induces cytotoxicity limiting their biomedical applications. Therefore, various chemical modifications of the PAMAM surface have been carried out to both decrease the toxic effects and increase transfection efficiency as drug and gene carrier.^3,4 Thus, the grafting of various amino acid residues at PAMAM chain ends to form heterobifunctional dendrimers⁵ and reduced surface positive charge density was envisaged;⁶ phenylalanine- and arginine-grafted-PAMAM dendrimers showed lower cytotoxity and higher gene delivery activity than unmodified PAMAMs or the commercial transfection agent polyethylenimine (PEI).^6,7 The biological activities of these modified PAMAMs have been widely evaluated and some analytical studies using gel electrophoresis, gel permeation, NMR and HPLC analyses permitted to study distributions of modified PAMAM.^4,6–8 Thus, G5 PAMAM bioconjugates as well as dendrimer nanoparticule-ligands distributions could be quantitatively assessed.⁸ On the other hand, unmodified small generation PAMAM dendrimers have been extensively studied by infra-red, ultra-violet/visible, nuclear magnetic resonance (NMR) and mass spectrometry (MS) techniques.^9–13 In particular, MS appeared to be useful for the full-characterization of small generation PAMAM dendrimers using either electrospray (ESI) or matrix assisted laser desorption ionization (MALDI) techniques.^10–13

To understand the molecular mechanisms involved in DNA/modified-PAMAM complexes formation, the study of non covalent interactions implicated in such dendriplexes formation can be envisaged on small dendrimer models. Thus, even if small generations of PAMAM are not expected to be applied in biomedical applications, we intend to design amino acid-modified G1 PAMAM as a good starting point for optimization and monitoring of the chemical design as well as for analytical methodology implementation.

Thus, in the present study, we propose to chemically modify ammonia-cored PAMAM G1(N) dendrimers with glycine residues and to present the structural characterization of the resulting modified-PAMAMs. Glycine was chosen as a model amino acid to optimize and control the surface chemical modification. Moreover, the glycine can be considered as a potential spacer for further conjugation with functional groups such as drugs and targeting moieties. The amino acid grafting step was optimized to yield fully surface-modified PAMAM G1(N) [Gly₆G1(N)] in the one hand, and partially modified PAMAM G1(N) [Gly_nG1(N), n = 0 to 6] in the other hand.

A thorough structural characterization of our modified PAMAM dendrimers involving mono- and two-dimensional NMR analyses, multi-stage mass spectrometry, accurate mass measurements, ion-mobility coupled to mass spectrometry (IM-MS) and capillary electrophoresis (CE) is presented. CE and IM-MS were proposed as alternative techniques for cationic PAMAM separation, appending literature data.^13,14 Moreover, IM-MS is well-known to present the advantage to permit in a single run the separation and the identification of ionic species in the millisecond time-scale from direct infusion of samples (a prior LC separation being not necessarily required). It is the first time to our knowledge that such comprehensive structural analysis is reported.

Results and discussion

1. PAMAM synthesis, chemical modification and deprotection

The PAMAM G1(N) was synthesized according to Tomalia's divergent approach,¹⁵ based on combination of acrylate Michael addition and amidation chemistry. PAMAM G1(N) was then conjugated with tert-butoxycarbonyl protected glycine (N-Boc-glycine) in order to obtain, in a first step, N-BocGly₆G1(N) in the one hand and N-BocGly_nG1(N) in the other hand (step (a) of Scheme 1). The conjugation of the primary amine groups of PAMAM surface with N-Boc-glycine involved O-(benzotriazol-1-yl)-N,N,N′,N′- tetramethyluronium hexafluoro-phosphate (HBTU) and 1-hydroxybenzotriazole hydrate (HOBt hydrate) as coupling agents to activate the carboxylic acid groups of the protected glycine.

Scheme 1 Synthesis of glycine modified PAMAM Gly_nG1(N). Reagents: (a) HBTU, HOBt, DIPEA in DMF/H₂O (b) H₃PO₄ in TFE.

In the case of N-BocGly₆G1(N), a slight excess of N-Boc-glycine was necessary to obtain fully modified PAMAM.

In the other hand, it appeared necessary to adjust and optimize the HBTU/HOBt coupling method to obtain a distribution of partially modified N-Boc-Gly_nG1(N) (with n = 0 to 6) since the sole diminution of N-Boc-glycine/PAMAM ratio only afforded a mixture of both unmodified and fully modified PAMAM (data not shown). The new grafting strategy involved the prior activation of N-Boc-glycine with HBTU/HOBt coupling agents followed by the slow dropwise introduction of activated N-Boc-glycine to a solution of PAMAM G1(N) in order to avoid high concentration of the activated amino acid in the reaction medium. Under such conditions, a distribution of differently modified PAMAM could be obtained. Note that, in the case of N-BocGly₁G1(N), N-BocGly₅G1(N) and N-BocGly₆G1(N) PAMAM dendrimers, a single regioisomeric structure can be formed while the other partially modified samples (Gly₂G1(N) to Gly₄G1(N)) can yield two different isomers depending on the position of the grafted glycine residues (onto the branches of the same arm or on different arms) (Scheme 1S of the ESI†).

The second and final step of modified PAMAM preparation consisted in N-Boc groups deprotection (Step (b) of Scheme 1).

The classical approach uses trifluoroacetic acid (TFA, pKa 0.3). However, such approach induced the hydrolysis of the amide linkage of our modified PAMAM dendrimers, even when mild conditions were used, with consecutive losses of grafted glycine residues. Thus, we developed a new strategy of deprotection inspired by the approach of Li et al.^16,17 These authors have reported the use of aqueous phosphoric acid (85% Wt, pKa₁ 2.15) as an alternative reagent for N-Boc groups deprotection. In our case, the use of H₃PO₄ in 2,2,2-trifluoroethanol (TFE) afforded the Gly₆G1(N) and Gly_nG1(N) PAMAM dendrimers in quantitative yields.

The resulting partially and totally modified PAMAM were then structurally characterized using NMR, ESI-MSⁿ, IM-MS and CE techniques after dialysis.

2. Structural characterization of glycine modified-PAMAM dendrimers

Nuclear magnetic resonance (NMR). The ¹H NMR spectra of the unmodified, partially and fully modified PAMAM G1(N) (Fig. 1A–C, respectively) show multiple peaks between 2.35 and 3.50 ppm which correspond to the methylene protons of PAMAM branching units (the ¹H NMR spectrum of the unmodified PAMAM G1(N) exclusively displayed these peaks). These signals were assigned by a standard procedure including two dimensional ¹H–¹H homonuclear scalar correlation (COSY)¹⁸ and ¹H–¹³C heteronuclear correlation (HMQC and HMBC).¹⁹

Fig. 1 ¹H NMR spectra of (A) PAMAM G1(N) (pH = 10.84) (B) Gly_nG1(N) (pH = 7.87) and (C) Gly₆G1(N) (pH = 5.84). Inset: partial model structure of (A) PAMAM G1(N) (B) Gly_nG1(N) and (C) Gly₆G1(N). *residual EDA (ethylenediamine) from the second step procedure of PAMAM synthesis used as repeating unit.

Characteristic spin systems were first identified in each structure from ¹H–¹H COSY and ¹H–¹³C HMQC NMR spectra. Then, they were assigned to specific locations in the structure by the observation of multiple bond ¹H–¹³C heteronuclear correlation (HMBC) between resonances of sequentially adjacent spin systems (Fig. 2B). This methodology permitted to obtain the complete ¹H NMR signal assignments for Gly₆G1(N) (Fig. 1C).

Fig. 2 (A) 600 MHz HMQC and (B) HMBC maps of Gly₆G1(N).

As previously mentioned, the proton signals between 2.50 and 3.50 ppm were assigned to the different methylene groups of the PAMAM G1(N) moiety while the signal at 3.73 ppm corresponded to the methylene (H-17) protons of the grafted glycine. Indeed, this signal was unambiguously identified by the HMBC experiment which showed the ¹H–¹³C correlation between glycine and dendrimer moieties providing evidence for the grafting of the amino acid residues onto the PAMAM surface; the carbonyl carbon C-16 of the glycine substituent shows three bonds ¹H–¹³C correlation with the methylene protons H-14 of the dendrimer moiety in addition to the expected ¹H–¹³C two bonds correlation with the methylene protons H-17 of the glycine residue (dotted lines in Fig. 2B).

In addition, the ¹H NMR data allowed the determination of the average number of glycine residues attached on the PAMAM G1(N) surface. The integration ratio of the H-17 protons at 3.73 ppm to that of the summed integral for the protons in the 2.35–3.50 ppm region indicated 100% conversion of amino groups meaning that six molecules of glycine were grafted to the six amino groups of the G1(N) molecule, as expected for Gly₆G1(N).

The structure of Gly_nG1(N) was elucidated from the combination of COSY, HMQC and HMBC experiments (data not shown), as afore mentioned, and by comparison with NMR data previously acquired for both the unmodified and fully modified PAMAM G1(N) samples. Thus, the methylene protons (H-17) of the attached glycine were observed at 3.56 ppm while the cluster of signals between 2.35 and 3.50 ppm correspond to the methylene protons of PAMAM branching units (Fig. 1B). Comparison of the Fig. 1B with both Fig. 1A and C provide evidence that the partially modified dendrimers contained a mixture of unreacted and modified branch termini. Indeed, the different signals corresponding to protons of both the unmodified and modified branches were detected in Fig. 1B, the H-13 and H-14 methylene protons of the modified branches appeared at δ 3.29 and 3.32 ppm, respectively, whilst the methylene protons of the unmodified end branches H-13′ and H-14′ respectively were at δ 3.46 and 3.10 ppm. Chemical shift differences were also observed for the methylene protons H-9, H-9′ and H-10, H-10′.

It is important to note that by comparing the three ¹H NMR spectra, strong changes in the signals characteristic for each system were observed. Indeed, from Fig. 1A to C, one can notice an increase of the downfield shifts and a broadening of the peaks for the whole spin systems. These changes were due to pH variations, as we established by studying the chemical shifts of unmodified PAMAM with the decrease of sample pH (Figure 1S of the ESI†). Such ¹H NMR chemical shift changes had also been reported for model bases during titration with acids.²⁰ The pH decrease between the unmodified, partially and totally modified PAMAM samples could easily be explained by the use of phosphoric acid during the deprotection step.

The average number of grafted glycine was estimated from ¹H NMR data as previously described; 25% conversion of the amino groups was deduced from the integration ratio of the protons at 3.56 ppm to that of the summed protons within the 2.35–3.50 ppm region (Figure S2 of the ESI†).

Mass Spectrometry (MS). The positive ion ESI mass spectra of fully and partially modified PAMAM G1(N) are depicted in Fig. 3A and B, respectively. Under our experimental conditions, doubly- and triply-charged sodiated ions were the major ionic species detected. Other doubly- and triply-charged ions, ([M + pH + nNa]^(p+n)+), were also detected as minor species. Accurate mass measurements and elemental composition of the doubly-charged sodiated ions from the different dendrimers are reported in Table 1. Then, ESI/MSⁿ (n = 2 to 4) experiments selecting [M + 2Na]²⁺ as precursor ions were performed for all the modified-PAMAMs. Fig. 4A and B show MS² experiments for the fully Gly₆G1(N) and the partially-modified (Gly₃G1(N)) PAMAM, respectively.

Fig. 3 Positive ions ESI-MS mass spectra of (A) Gly₆G1(N) and (B) Gly_nG1(N) (with n = 0 to 6). *Glycine modified-PAMAMs with structural defects correspond to Gly₄G1(N) presenting intermolecular cyclization with a lack of EDA (60 Da). ✚ [M + 3H]³⁺, ▲ [M + Na + 2H] ³⁺, ● [M + 2Na + H]³⁺, ♦ [M + Na + H]²⁺and ■ [M + 2H]²⁺ of Gly_nG1(N) or of defective structures (★). ESI mass spectra acquired using the SYNAPT G2 mass spectrometer.

Table 1 Elemental compositions of G1(N) and Gly_nG1(N) modified PAMAM (n = 0 to 6) determined by accurate mass measurements

Compound	Elemental compositions^a	m/z exp.	m/z calcd^b	Error (ppm)	Ion
a Elemental compositions of the neutral species. b Monoisotopic values calculated with the Waters MassLynx software.
G1(N)	C45H93N19O9	544.8614	544.8594	3.6	[M + 2Na]^2+
Gly1G1(N)	C47H96N20O10	573.3701	573.3707	−1.0	[M + 2Na]^2+
Gly2G1(N)	C49H99N21O11	601.8814	601.8814	0.0	[M + 2Na]^2+
Gly3G1(N)	C51H102N22O12	630.3920	630.3922	−0.3	[M + 2Na]^2+
Gly4G1(N)	C53H105N23O13	658.9033	658.9029	0.6	[M + 2Na]^2+
Gly5G1(N)	C55H108N24O14	687.4142	687.4136	0.9	[M + 2Na]^2+
Gly6G1(N)	C57H111N25O15	715.9253	715.9243	1.4	[M + 2Na]^2+

---

Fig. 4 ESI-MS² mass spectra of [M + 2Na]²⁺ ion of (A) Gly₆G1(N) (m/z 715.92) and (B) Gly₃G1(N) (m/z 630.15). These data were acquired using the SYNAPT G2 mass spectrometer.

Abundant [M + 2Na]²⁺ and [M + 3Na]³⁺ multiply charged ions of fully modified PAMAM Gly₆G1(N) were respectively observed at m/z 715.92 and m/z 484.94 in the ESI mass spectrum (Fig. 3A). Their assignments were confirmed by accurate mass measurements (Table 1). Additional low-intensity ions were also detected at m/z 628.86 corresponding to the [M + 2Na]²⁺ ion of a defective modified structure since the Tomalia synthesis is known to yield structural defects.¹ Thus, the m/z 628.86 corresponded to the cyclic defect with one missing ethylenediamine molecule (60 Da) and four grafted glycine residues (C₅₁H₁₉₇N₂₁O₁₃, m/z calculated: 628.8685; found: 628.8690 for [M + 2Na]²⁺ion). Multi-stage mass spectrometry carried out on the doubly-charged precursor ion [M + 2Na]²⁺ of Gly₆G1(N) yielded product ions arising from neutral losses of modified branches (171 Da) and of a fully modified arm (456 Da) (Fig. 4A for MS² data), in accordance with the systematic fragmentation pattern previously described for the unmodified PAMAM.^11–13

The well-established fragmentation pathway involved retro-Michael rearrangements leading to successive losses of branches and arm(s). Such dissociation processes yielded the singly-charged product ions at m/z 1237.74, m/z 1066.64, m/z 952.60, m/z 781.47 and m/z 479.26 as well as doubly-charged product ions at m/z 630.37 and m/z 544.82 (Fig. 4A). Note that the ion at m/z 479.26 corresponds to the sodiated fully modified arm. In a similar way, the product ions at m/z 194.08 and m/z 308.17 correspond to sodiated modified branches containing one and two branch unit(s), respectively. These fragmentation pathways were confirmed by ESI-MS³ and MS⁴ experiments selecting the different product ions detected in Fig. 4A (data not shown).

The ESI mass spectrum of the partially modified PAMAMs showed a main distribution corresponding to the doubly-charged ions [M + 2Na]²⁺ of Gly_nG1(N) with n = 0 to 6 (Fig. 3B). The accurate mass measurements of these ions were consistent with the elemental composition of the various partially substituted Gly_nG1(N) derivatives (Table 1). Thus, the ESI mass spectrum displayed the different individual modified species and showed that partial surface-functionalisation had been achieved, the most abundant species corresponding to the grafting of one, two and three glycine residues. Minor distributions were also detected and corresponded to either [M + Na + H]²⁺ (♦) and [M + 2H]²⁺ (■) doubly-charged ions of Gly_nG1(N) or to defective structures [M + 2H]²⁺ ions (★) (Fig. 3B) for which elemental compositions were also determined by accurate mass measurements (data not shown).

ESI/MS² experiments were performed selecting the doubly charged ion [M + 2Na]²⁺ of each partially modified PAMAM as the precursor ion. The fragmentation processes were analogous to those previously mentioned for the doubly charged ion of Gly₆G1(N) although, in the case of partially modified PAMAM, neutral losses of either modified or unmodified branch(es) or arm(s) could be possible depending of the number of glycine residues grafted onto the PAMAM surface. Thus, the [M + 2Na]²⁺ ion (m/z 630.15) of partially modified PAMAM Gly₃G1(N) dissociated via a retro-Michael rearrangement with the neutral losses of a modified branch (171 Da), an unmodified branch (114 Da), an unmodified arm (342 Da) and/or modified-arm(s) with one or two grafted glycine units (399 or 456 Da) (Fig. 4B).

The loss of 285 Da corresponding to a moiety containing two branches with only one grafted glycine (two branches + one glycine) could also occur consecutively to the elimination of one branch, modified or not. As previously observed in the case of Gly₆G1(N), some product ions corresponding to the sodiated modified branch (two branches and one glycine) moiety and modified arms with one or two Gly were detected at m/z 194.08, 308.17, 422.24 and 479.26, respectively. The sodiated unmodified two-branches and an unmodified arm were detected at m/z 251.14 and 365.22, respectively. The presence of product ions at m/z 479.26 and 365.22 respectively indicated that two glycine residues were grafted onto the same arm while one arm remained unmodified. Such result is consistent with the presence of a Gly₃G1(N) PAMAM isomer, which presents a totally modified arm, an unmodified arm and a third half-modified arm since, as previously mentioned in the chemical modification section, the grafting of three glycine residues could yield two different isomers for Gly₃G1(N) PAMAM (Scheme S1 of the ESI†). The assumption of this structural isomer presenting two grafted glycine residues onto one arm and one glycine onto another arm was subsequently verified through MS³ and MS⁴ experiments which permitted determination of the relationship between the different product ions. In fact, the whole fragmentation pathways reported in the first generation product ion spectrum presented in Fig. 4B were deduced from several multi-stage mass spectrometry experiments. Therefore, the presence of m/z 781 and 895 in the MS³ spectrum “630.1 → 1237.8 → 2^nd generation product ions” is consistent with the supposed structure because m/z 781.5 and 895.6 arose from the m/z 1237.8 intermediate ion with neutral losses of a totally modified arm (456 Da) and an unmodified arm (342 Da), respectively (Scheme S1-A of the ESI†). Moreover, the MS⁴ spectrum “630.1 → 1237.8 → 781.5 → 3^rd generation product ions” showed m/z 439.2 and 382.2 that arose from m/z 781.5 with elimination of a half-modified arm (399 Da) and an unmodified arm (342 Da), respectively (Scheme S1-A of the ESI†). Therefore, unmodified, half-modified and totally modified arms constituted this isomer.

The presence of a second isomer of Gly₃G1(N), formed during glycine grafting and carrying three half-modified arms (Scheme S1–B of the ESI†) was also evidenced because the “630.1 → 1237.8 → 2^nd generation product ions” spectrum also showed an ion at m/z 838.5 ion which further dissociated into the m/z 439 ion as confirmed by the MS⁴ experiment “630.1 → 1237.8 → 838.5 → 3^rd generation product ions” as well as the MS³ experiment “630.1 → 838.5 → 2^nd generation product ions”. Both m/z 838 and m/z 439 arose from elimination of a half-modified arm indicating that three half-modified arms constituted this second isomer of Gly₃G1(N) formed during glycine grafting (Figure S3 of the ESI†).

In the same way, the presence of two isomers for Gly₂- and Gly₄- G1(N), according to the distribution of the glycine residues onto the arms, was demonstrated through specific and appropriate MS³ and MS⁴ experiments.

Investigation of modified PAMAM distribution by capillary electrophoresis (CE) and ion mobility spectrometry coupled to mass spectrometry (IM-MS). Distribution of the partially modified PAMAM Gly_nG1(N) was studied using two different separation approaches suitable for charged species: in condensate phase (CE) in the one hand (Fig. 5A); and in the gas phase (IM-MS) in the other hand (Fig. 5C). The objective was to differentiate individual species to access semi-quantitation and to estimate the average number of grafted glycine units, a value which could be compared with that previously determined using ¹H NMR. Electropherogram (Fig. 5B) and ion mobility spectrum (Fig. 5D) of completely modified Gly₆G1(N) showed that a single component was mainly presented in our sample.

Fig. 5 (A) Electropherogram and (C) Ion mobility spectrum of Gly_nG1(N). Peak identification: (a) G1(N) (b) Gly₁G1(N) (c) Gly₂G1(N) (d) Gly₃G1(N) (e) Gly₄G1(N) (f) Gly₅G1(N) (g) Gly₆G1(N). The insert of (C) shows an expansion of the Gly_nG1(N) doubly-charged ions distribution. (B) Electropherogram and (D) Ion mobility spectrum of completely modified Gly₆G1(N) (g).

Capillary electrophoresis. CE was performed because this separation method, valuable for ionized compounds, can offer high efficiencies and consequently high resolving power. The separation of polycationic polymers, particularly dendrimers, has been previously achieved using acidic electrolytic buffers.¹³ As the amino groups of either the PAMAM branches or the grafted glycine are highly basic (pKa > 9), all the PAMAM, modified or not, were positively charged in such electrophoretic conditions (phosphate buffer 20 mM, pH = 2.7).

Therefore, the charge density was expected to decrease when the number (n) of grafted glycine units increased, and consequently the electromigration time should increase.

The electropherogram acquired after injection of the partially modified PAMAM G1(N) showed the distribution of the differently grafted Gly_nG1(N) (Fig. 5A). Another injection of this sample spiked with unmodified PAMAM G1(N) showed an increase of the first peak (noted a in Fig. 5) therefore assigned to unmodified PAMAM G1(N). In consequence, the peaks (b) to (g) could be assigned to Gly_nG1(N) with n = 1 to 6, respectively, in agreement with the expected electromigration order. However, under our conditions, the CE separation was insufficient to resolve individual species impeding peak integration and the estimation of the level of amino group conversion.

Ion mobility spectrometry coupled to mass spectrometry. An alternative method for investigating the distribution of Gly_nG1(N) by ion separation involved IM-MS using travelling wave ion mobility (TWIM) cell which is known to allow separation of ions in the gas phase according to their size-to-charge ratio (IM) and also their m/z ratio (MS).

The two-dimensional ion mobility vs. m/z correlation data of the partially modified PAMAMs showed three groups of multi-charged ions distributed on two main diagonals (Fig. 6). Doubly-charged ions ([M + 2Na]²⁺, [M + H + Na]²⁺ and [M + 2H]²⁺ ions) were aligned on a first diagonal whilst triply- and quadruply-charged ions were fitted with a second diagonal. Thus, doubly-charged species were well-separated from other multiply-charged ions of Gly_nG1(N) as also evidenced on the ion mobility spectrum (Fig. 5C). Focusing on the drift time distributions of doubly-charged ions which were selectively extracted, seven peaks could be differentiated (inset of Fig. 5C): these peaks labeled (a) to (g) could be respectively assigned to G1(N), Gly₁G1(N), Gly₂G1(N), Gly₃G1(N), Gly₄G1(N), Gly₅G1(N) and Gly₆G1(N) thanks to the orthogonal information of the m/z ratio. Then, comparing IM to CE data, identical migrating orders were obtained using both gas and condensate phase electromigration.

Fig. 6 Two-dimensional ion mobility (drift time) vs. m/z data of Gly_nG1(N).

However, the IM analysis was in the millisecond time scale while the CE separation was performed in the minute time scale. Moreover, since our CE apparatus was not coupled to mass spectrometry, the separation of individual modified dendrimers was only achieved using the IM-MS approach involving the combination of orthogonal IM and MS data. It was therefore possible to estimate the average number of grafted glycine units by integrating the peaks extracted from the ion mobility spectrum of Fig. 5C; the ion mobility peak areas corresponding to multi-charged ions were determined. The peak areas of the different multi-charged ions were summed after application of correction factors because discrepancies were evidenced in the response factors of the different modified PAMAMs depending on the number of grafted glycine residues; indeed, the ion mobility spectrum of an equimolar solution of unmodified G1(N) and totally modified Gly₆G1(N) showed that Gly₆G1(N) had higher peak intensities (by a factor of 5). After application of the correction factors calculated according to the method outlined in the experimental section, the percentage of each modified PAMAM could be determined from these integration data (they are related to the “% of areas” column of the Table S1 of the ESI†). Thus, the percentages from major species to minor species are: Gly₁G1(N) (31%), native G1(N) (28%), Gly₂G1(N) (15%), Gly₃G1(N) (11%), Gly₄G1(N) (8%), Gly₅G1(N) (5%) and Gly₆G1(N) (3%). Such quantitative information regarding the distribution of partially grafted PAMAM could therefore be provided by the integration of ion mobility peaks from IM-MS data. Moreover, the average number of grafted glycine residues was consecutively estimated to be near 28% (Table S1 of the ESI†). This result was in the same order of magnitude to that estimated by ¹H NMR. The difference between MS and NMR values could be explained by possible drawbacks in each analytical technique, i.e. discrepancies in MS response factors between the different modified-PAMAMs and possible bias in the ¹H NMR integrations because of the presence of defective structures. The IM-MS analysis could also afford information concerning the three-dimensional structure of the PAMAM ions, in particular their dimensions in the gas phase¹⁴ that could help the understanding of the bioactive conformation.²¹ Indeed, collision cross sections (CCS) and consequently the radii (according to the spherical ion model) could be estimated from calibration curves obtained using reference ions having well-described collision cross-sections, such as the protonated molecules of the polyalanine peptides. Thus, the experimental CCS of doubly-charged modified PAMAMs ([M + 2H]²⁺ in the particular case of IMS analysis) were determined and compared to theoretical CCS calculated using trajectory method (TM) model²² as well as CCS estimated from molecular weight (g mol⁻¹) using the spherical ion model (SIM) according to Ruotolo et al.²³ (Table 2).

Table 2 Theoretical and experimental collision cross-sections of PAMAM G1(N) and Gly_nG1(N)

	G1(N)	Gly1G1(N)	Gly3G1(N)	Gly6G1(N)
a Theoretical collision cross-sections (Ω) calculated with the TM model.^21 b Collision cross-section determined using the equation Ω = 2453 × M^2/3 for GlynG1(N) according to Ruotolo et al.^22 c Experimental collision cross-section measured in triplicate for doubly charged ions of PAMAM dendrimers.
Calculated Ωaverage (Å^2) using the TM model^a	328	335	343	373
Ω (Å^2) (SIM)^b (Ω = 2453 × M^2/3)	252	261	288	337
Ω experimental (Å^2) (TWIM)^c	240	245	261	282

---

As expected the calculated as well as the measured CCS increased with the number of grafted glycine units. Moreover, a linear relationship could be observed between the experimental CCS of the Gly_nG1(N) PAMAMs and the number of grafted glycine residues (Fig. 7). The measured CCS were however smaller than both the TM and the SIM values. That could be explained by the formation of more compact structures in the gas phase than in solutions.^24,25 These results showed that PAMAMs exhibited, as a general trend, compact structures from unmodified to fully modified forms which is consistent with the spherical nature of these compounds known to present limited extensibility.²⁶ However, to obtain more accurate cross section values, the use of calibration standards that most closely resemble our analytes should be envisaged, as previously reported by Salbo et al.²⁷

Fig. 7 Plot of the experimental CCS of Gly_nG1(N) PAMAMs vs the number of grafted glycine residues.

Conclusions

The synthesis of a series of fully- as well as partially-substituted PAMAM G1(N) dendrimers functionalized with glycine residues was achieved using different and separate optimized grafting reactions. Indeed, two different experimental conditions were necessary for the coupling step. The N-Boc deprotection step was also adjusted for the particular case of these modified dendrimers because the usual procedure induced hydrolysis of the grafted PAMAM. The full structural characterization of the series of surface-modified PAMAMs was performed using the combination of MS and NMR techniques. The glycine grafting yield as determined by one-dimensional ¹H NMR was approximately 25% in the case of partially modified PAMAM G1(N) while it was found to be 100% in the case of Gly₆G1(N). Effective grafting of glycine onto the dendrimer surface was verified by both two-dimensional NMR and MS experiments. The distribution of partially modified PAMAMs was evidenced by MS, IM-MS and CE analyses. Elemental composition of each synthesized dendrimer was determined by accurate mass measurements. Mass spectrometry permitted the detection of individual ionic species of each modified Gly_nG1(N) which were submitted to multi-stage MSⁿ experiments. Thus, appropriate MS², MS³ and MS⁴ experiments allowed the structural identification of each Gly_nG1(N) dendrimers, the differentiation and the identification of two isomeric structures for Gly₂-,Gly₃- and Gly₄G1(N) and the presence of only one isomer in the case of Gly₁G1(N), Gly₅G1(N) and Gly₆G1(N).

Both CE and IM-MS displayed separation of the differently modified dendrimers and consequently showed the distribution in a complementary way to the sole MS approach. In the case of IM-MS, the glycine grafting yield was estimated to be nearly 28%.

This work constitutes a model for the synthesis and the structural characterization of glycine-modified PAMAMs and represents a new approach for the construction of a library of amino acid-modified PAMAM dendrimers with ammonia cores and subsequently the extension of the data to EDA-cored surface-modified PAMAMs. Moreover, the glycine attached to the dendrimer surface can be further used as a spacer to conjugate functional groups such as drugs and targeting moieties.

Experimental section

Materials

Methyl acrylate, ethylenediamine (EDA), N-(tert-butoxycarbonyl)glycine (Boc-gly-OH), O-(benzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium hexafluorophosphate (HBTU), 1-hydroxybenzotriazole hydrate (HOBt hydrate), N,N-dimethylformamide (DMF), N,N-diisopropylethylamine (DIPEA), phosphoric acid (H₃PO₄), toluene, dichloromethane (DMC), and HPLC grade methanol were purchased from sigma-aldrich (Saint-Quentin-Fallavier, France). Acetonitrile (ACN), 2,2,2-trifluoroethanol (TFE), acide phosphorique (H₃PO₄), Float-A-Lyzer G2 dialyzer (0.5 à 1kDa), ammonium acetate and LC-MS grade acetonitrile (ACN) were purchased from fisher scientific (Illkirch, France). Deionised water (18 MΩ) was obtained from a Milli-Q apparatus (Millipore, Bedford, MA, USA).

Synthesis of ammonia core PAMAM G1(N)

The PAMAM G1(N) was prepared according to Tomalia's divergent synthesis.¹⁴ A solution of methyl acrylate (2.14 equiv. per dendrimer amine end group) in methanol was added dropwise to a stirred solution of ammonia in methanol. The reaction mixture was stirred at room temperature in dark for three days then concentrated in vacuo to yield the half-generation PAMAM G-0.5. In a second step, to a solution of ethylene diamine (EDA) (10 equiv. per dendrimer ester end group) in methanol was directly added a solution of half-generation PAMAM G-0.5 in methanol then stirred for six days at room temperature giving the full generation PAMAM G0. Preceding two-steps were repeated using the full generation PAMAM G0 followed by a repetition azeotropic distillation [toluene 100 mL was added to the product dissolved in methanol (10 mL)]. For generations 0.5, the crude product was purified by column chromatography (R_f: 0.40, 12% MeOH in DCM, silica 60). The reiteration of this two-steps procedure gives the ammonia-cored PAMAM G1(N) (yield 97%). ESI-HRMS: m/z calculated and measured for C₄₅H₉₃N₁₉O₉ ([M + H]⁺) were 1044.7482 and: 1044.7487, respectively. ¹H NMR (D₂O). δ = 2.40 (18H, m, H-3 and H10′), 2.60 (6H, t, H7), 2.69 (12H, t, H9′), 2.76 (6H, t, H-2), 2.79 (12H, t, H-14′), 3.21 (12H, t, H-13′), 3.26 (6H, t, H-6).

Preparation of Gly₆G1(N)

The fully modified PAMAM-G1(N) was prepared in DMF/water (5 : 2 v/v) by mixing 0.95 equiv. of PAMAM G1(N), 6 equiv. of N-Boc-Gly-OH, 6.6 equiv. of HBTU, HOBt·H₂O and 24 equiv. of DIPEA for one night at room temperature. The product was dialyzed (Mw cutoff = 1000 Da) for 6 days against deionized water. The dialysed sample was then lyophilized to afford viscous orange oil. The N-Boc protecting group was removed in TFE by adding 17 equiv. of aqueous H₃PO₄ (85% Wt). The reaction mixture was stirred at room temperature in dark for 12 min (a precipitate was formed almost instantaneously). The reaction mixture was then diluted in H₂O and the pH was adjusted to ≈7 with ammonium acetate (1 M). The product was dialyzed for 6 days against deionized water and lyophilized to give a yellow solid (yield 70%). The ESI-HRMS data are reported in Table 1 of the Results and Discussion section. ¹H NMR (D₂O). δ = 2.59 (12H, t, H-10), 2.64 (6H, br. t, H-3), 3.03 (6H, br. t, H-7), 3.17 (18H, br. t, H-2 and H-9), 3.28 (12H, m, H-13), 3.31 (12H, m, H-14), 3.46 (6H, br. t, H-6), 3.73(12H, s, H-17). ¹³C NMR. δ = 30.6 (C-3 and C-10), 35.1 (C-6), 38.4 (C-13 and C-14), 40.7 (C-17), 49.4(C-2 and C-9), 52 (C-7),167.5 (C-16), 173.2 (C-4 and C-11).

Preparation of Gly_nG1(N)

A solution containing 3 equiv. of N-Boc-Gly-OH, 3.3 equiv. of HBTU, HOBt·H₂O and 12 equiv. of DIPEA in 5 mL of DMF was prepared (the flask was immersed in an ice and water bath) and stirred for 40 minutes. This solution was added dropwise over 30 minutes to a solution of PAMAM G1(N) (0.95 equiv.) in DMF/water (5 : 2 v/v) at 40 °C. The whole was then allowed to stir at 40 °C for 4 hours. Then, purification and deprotection of the N-Boc protecting group (as above) to give yellow solid of partially modified PAMAM G1 (N) with a distribution of attached glycine from 0 up to 6 (yield 68%). The ESI-HRMS data are reported in Table 1 of the Results and Discussion section. ¹H NMR (D₂O). δ = 2.41 (10H, m, H-10′), 2.46 (8H, m, H-3 and H-10), 2.63 (6H, m, H-7), 2.82 (18H, m, H-2, H-9 and H9′), 3.10 (8H, t, H-14′), 3.29 (6H, m, H-6 and H-13), 3.32 (8H, m, H-14), 3.46 (8H, t, H-13′), 2.94 (3.5H, Br.s, H-17). ¹³C NMR. δ = 32.38 (C-3, C-10 and C-10′), 36.73 (C-6 and C-13), 38.73 (C-14), 38.98 (C-14′), 41.79 (C-17), 48.84 (C-2, C-9 and C-9′), 51.39 (C-7), 171.2.5 (C-16), 175.0 (C-4 and C-11).

The average number of grafted glycine was measured by ¹H-NMR using the ¹H proton integration method. The ¹H-NMR spectrum of Gly_nG1(N) showed the summed protons within the 2.35–3.50 ppm region (72 H) corresponded to methylene of PAMAM G1(N) dendrimer. The proton signal at 3.56 ppm (2.94H) corresponded to the methylene of glycine residue. The integration ratio of the methylene protons in glycine residue and all the methylene protons of G1(N) moiety was used to calculate the grafting ratio; an average of 1.5 grafted glycine was therefore estimated (Figure S2 of the ESI†).

Nuclear magnetic resonance spectroscopy (NMR)

¹H, ¹³C, COSY, HMBC and HMQC NMR spectra of PAMAM G1(N), Gly_nG1(N) and Gly₆G1(N) were obtained in D₂O using a Bruker Avance III 600 MHz spectrometer equipped with a 5 mm CPTXi cryoprobe.

Mass spectrometry (MS) and ion mobility spectrometry (IM)

Sample preparation. Gly₆G1(N) and Gly_nG1(N) samples were prepared in methanol and H₂O mixture (50 : 50 v/v) with a concentration of 2 ng μL⁻¹. Polyalanine peptide mixture was prepared to a concentration of 10 ng μL⁻¹ in 1% formic acid, acetonitrile and H₂O mixture (50 : 50 v/v).

ESI-IT-MSⁿ. ESI-MSⁿ (n = 1, 2, 3 and 4) analyses were performed using a Bruker HCT Ultra ETD II quadrupole ion-trap (QIT) mass spectrometer equipped with an ESI source and the Esquire control 6.2 and Data Analysis 4.0 softwares (Bruker Daltonics, Bremen, Germany).

For the ESI parameters the capillary and end plate voltages respectively set to −4.0 kV and −3.5 kV in positive ion mode. The skimmer voltage was set to 40 V and the injection low mass cut off (LMCO, corresponding to the “trap drive” parameter) value was m/z 69. The nebulizer gas (N₂) pressure, drying gas (N₂) flow rate and drying gas temperature were 10 psi, 5.0 L min⁻¹ and 300 °C, respectively. Helium pressure in the ion trap was 9.6 × 10⁻⁶ mbar. Spectra were acquired in the m/z 200–1300 range, using the ‘Standard-Enhanced’ mode (scan rate of 8000 m/z per second). The number of ions entering the ion trap was automatically adjusted by controlling the accumulation time with the ion charge control (ICC) mode (target 100 000) with a maximum accumulation time of 50 ms. The values of spectra averages and rolling average were 5 and 2 respectively. ESI-MSⁿ experiments were carried out by collision induced dissociation (CID) using a resonant excitation frequency with amplitudes from 0.29 to 1.0 V_p−p, helium as the collision gas, isolation width of 1 m/z unit for the precursor ions and of 2 m/z unit for the intermediate ions. Sample solutions were infused into the source at a flow-rate of 3 μL min⁻¹ by means of a syringe pump (Cole-Palmer, Vernon Hills, Illinois, USA).

ESI-Q-TOF-MS² and ESI-Q-IM-TOF-MS. These experiments were performed using a Waters SYNAPT G2 hybrid quadrupole/HDMS instrument equipped with an ESI LockSpray™ source, the MassLynx 4.1 and the DriftScope 2.2 softwares (Waters, Manchester, UK). The hybrid geometry of the MS system and the IMS module ‘Triwave’ consisting of three travelling wave-enabled stacked ring ion guides that have been described elsewhere.^28,29 The SYNAPT HDMS system was calibrated using sodium formate cluster ions (2 mg mL⁻¹) and operated in ‘V’ resolution mode (resolution 20000 FWHM). Leucine enkephalin (2 ng μL⁻¹) was used as the lock mass and was infused (3 μL min⁻¹) using an independent reference spray via the LockSpray™ interface which was operated at a reference scan frequency, lock spray capillary and collision energy of 10 s, 3 kV and 4 eV, respectively. The optimized conditions for the sample included capillary voltage; 3 kV, sample cone voltage; 40 V, source temperature; 90 °C, desolvation temperature, 250 °C; desolvation gas flow (N₂), 1200 L h⁻¹. For MS and MS–MS experiments, data were acquired (50–2000 m/z range with 1s scan time and 0.02s interscan delay). Sample solutions were infused into the source at a flow-rate of 3 μL min⁻¹ by means of a syringe pump (Cole-Palmer, Vernon Hills, Illinois, USA).

Q-TOF-MS² experiments involved (i) selection of the [M + 2Na]²⁺ precursor ion with the quadrupole mass analyzer within a 1 m/z unit window (ii) fragmentation in the trap collision cell (first ion guide of the Triwave), collision energy; 38 eV for Gly₃G1(N) and 40 eV for Gly₆G1(N).

Elemental compositions of protonated and sodiated multicharged ions (MS mode) were proposed from the accurate mass measurements and considering the isotope predictive filtering value (i-FIT™) given for each possible elemental composition.

The IMS conditions were optimized as followed: gas flow (N₂); 120 mL min⁻¹, wave height; 37.3 V and wave velocities 800 m s⁻¹. IM-MS analyses were performed in triplicate. For the three measurements, strictly identical drift times were observed for doubly charged ions of PAMAM dendrimers. That could be explained by the well-described performance of IMS technique which gives highly reproducible data.

The peak integration was performed from extracted ion mobility spectrum of each multi-charged ion. Before summing the peak areas, correction factors were estimated for each multi-charged ion from the analysis of an equimolar solution of G1(N) and Gly₆G1(N). The extracted ion mobility spectra of the G1(N)/Gly₆G1(N) mixture showed that doubly-, triply- and quadruply-charged ions of Gly₆G1(N) had higher peak intensities than the corresponding multi-charged ions of G1(N). A decrease of about 80% was observed from Gly₆G1(N) to G1(N) signals. Thus, we estimated that the diminution of response factor (X) per ungrafted end group could be calculated according to the formulae (X)⁶ ≈ 20%. Thus, X was estimated to be about 75–80%. The correction factors were systematically applied on the peak areas. Then, the corrected peak areas were summed for each modified PAMAM and the corresponding area percentage were calculated. The percentage of grafted glycine was determined by multiplying the area percentage with the ratio of grafted glycine which is equal to n/6 with n = number of grafted glycine. Finally, the total percentage of grafted glycine was calculated by summing all the individual percentage of grafted glycine (Table S1 of the ESI†).

Collision cross-section measurements. The Collision cross-sections were measured using polyalanine for IMS cell calibration. The instrument parameters used were those previously optimized for the PAMAM samples.

Collision cross-section estimation. Collision cross-section of each glycine-modified PAMAM [Gly_nG1(N), n = 0–6] was calculated according to Ruotolo et al.²³

Computational modeling. Structures of glycine-modified PAMAM [GlynG1(N), n = 0 to 6] were generated and geometry-optimized by minimizing its energy using the DISCOVER simulation package (Biosym Technologies, San Diego, CA, USA) and the consistent-valence force field (CVFF).³⁰ For each compound, 50 structures were generated and each theoretical collision cross-section was calculated using the trajectory method (TM) of MOBCAL software.²²

Capillary electrophoresis (CE). The CE separations were performed using a CE instrument (spectraphoresis 100, TSP Thermo Separation Products) consisting of a modular injector, a high-voltage power supply unit and an UV detector. Model 600 data software was use to process data. An uncoated fused-silica capillary (Supelco, Bellafonte, PA, USA) with an inner diameter of 50 μm, a total length of 68 cm and an effective separation length of 40.2 cm was used. The settled capillary was conditioned by successive rinsing (30 min each) with methanol, water, 1 M NaOH, 0.1 M NaOH, water and finally with an electrolyte buffer composed of 75 mM phosphate buffer (pH = 2.7). Injections (1s) at the anodic end of the capillary were performed using hydrodynamic injection mode. The electrophoretic separations were achieved at a constant voltage of 15 kV (17 μA). UV detection was set at 210 nm. The temperature of the capillary was 24 °C. The run time was set to 30 min. The samples were prepared by dissolving 3 mg of Gly_nG1(N) in 1 mL of electrolyte buffer and were run immediately after dissolving.

Acknowledgements

This work has been partially supported by INSA Rouen, Rouen University, CNRS, EFRD (N°31708), the IS:CE-Chem project and Interreg IV A France -(Channel)- England Programme and Labex SynOrg (ANR-11-LABX-0029). They are gratefully thanked for their financial support.

Notes and references

1. J. M. J. Fréchet and D. A. Tomalia, Dendrimers and Other Dendritic Polymers, Wiley Series in Polymer Science, John Wiley & Sons, Ltd, Chichester, UK, 2001, p. 1.
2. K. Fant, E. K. Esbjörner, P. Lincoln and B. Nordén, Biochemistry, 2008, 47, 1732–1740.
3. R. Jevprasesphant, J. Penny, R. Jalal, D. Attwood, N. B. McKeown and A. D'Emanuele, Int. J. pharm., 2003, 252, 263–266.
4. K. Fant, E. K. Esbjörner, A. Jenkins, M. C. Grossel, P. Lincoln and B. Nordén, Mol. Pharmaceutics, 2010, 7, 1734–1746.
5. R. S. Navath, A. R. Menjoge, B. Wang, R. Romero, S. Kannan and R. M. Kannan, Biomacromolecules, 2010, 11, 1544–1563.
6. X. Wang, Y. He, J. Wu, C. Gao and Y. Xu, Biomacromolecules, 2010, 11, 245–251.
7. H. Y. Nam, H. J. Hahn, K. Nam, W. H. Choi, Y. Jeong, D.-E. Kim and J. S. Park, Int. J. pharm., 2008, 363, 199–205.
8. D. Mullen, M. Fang, K. Desai, J. R. Baker, J. B. G. Orr and M. M. B. Holl, ACS Nano, 2010, 4, 657–670.
9. A.-M. Caminade, R. Laurent and J. P. Majoral, Adv. Drug Delivery Rev., 2005, 57, 2130–2146.
10. J. Peterson, V. Allikmaa, J. Subbi, T. Pehk and M. Lopp, Eur. Polym. J., 2003, 39, 33–42.
11. M. He and S. A. McLuckey, Rapid Commun. Mass Spectrom., 2004, 18, 960–972.
12. T. J. Vincent, R. Dole and C. M. Lange, Rapid Commun. Mass Spectrom., 2008, 22, 363–372.
13. E.-D. Leriche, F. Maire, M. C. Grossel, C. M. Lange and C. Loutelier-Bourhis, Rapid Commun. Mass Spectrom., 2012, 26, 1718–1724.
14. F. Maire, G. Coadou, L. Cravello and C. M. Lange, J. Am. Soc. Mass Spectrom., 2013, 24, 238–248.
15. D. A. Tomalia, H. Baker, J. Dewald and M. Hall, Polym. J., 1985, 17, 117–132.
16. B. Li, M. Berliner, R. Buzon, C. K. Chiu, S. T. Colgan, T. Kaneko, N. Keene, W. Kissel, T. Le, K. R. Leeman, B. Marquez, R. Morris, L. Newell, S. Wunderwald, M. Witt, J. Weaver and Z. Zhang, J. Org. Chem., 2006, 71, 9045–9050.
17. B. Li, R. Bemish, R. A. Buzon, C. K. Chiu, S. T. Colgan, W. Kissel, T. Le, K. R. Leeman, L. Newell and J. Roth, Tetrahedron Lett., 2003, 44, 8113–8115.
18. K. Nagayama, A. Kumar, K. Wuthrich and R. R. Ernst, J. Magn. Reson., 1980, 40, 321–334.
19. A. Bax and M. F. Summers, J. Am. Chem. Soc., 1986, 108, 2093–2094.
20. H. Kim, C. R. Babu and D. J. Burgess, Int. J. Pharm., 2013, 448, 123–131.
21. C. Lapthorn, F. Pullen and B. Z. Chowdhry, Mass Spectrom. Rev., 2013, 32, 43–71.
22. M. F. Mesleh, J. M. Hunter, A. A. Shvartsburg, G. C. Schatz and M. F. Jarrold, J. Phys. Chem., 1996, 100, 16082–16086.
23. B. T. Ruotolo, J. L. P. Benesch, A. M. Sandercock, S.-J. Hyung and C. V. Robinson, Nat. Protoc., 2008, 3, 1139–1152.
24. C. A. Scarff, K. Thalassinos, G. R. Hilton and J. H. Scrivens, Rapid Commun. Mass Spectrom., 2008, 22, 3297–3304.
25. T. W. Knapman, J. T. Berryman, I. Campuzano, S. A. Harris and A. E. Ashcroft, Int. J. Mass Spectrom., 2010, 298, 17–23.
26. E. N. Govorun, K. B. Zeldovich and A. R. Khokhlov, Macromol. Theory Simul., 2003, 12, 705–713.
27. R. Salbo, M. F. Bush, H. Naver, I. Campuzano, C. V. Robinson, I. Peterson, T. Jorgensen and K. F. Haselmann, Rapid Commun. Mass Spectrom., 2012, 26, 1181–1193.
28. K. Giles, S. D. Pringle, K. R. Worthington, D. Little, J. L. Wildgoose and R. H. Bateman, Rapid Commun. Mass Spectrom., 2004, 18, 2401–2414.
29. S. E. Slade, K. Thalassinos, R. H. Bateman, M. T. Bowers and J. H. Scrivens, Int. J. Mass Spectrom., 2007, 261, 1–12.
30. A. T. Hagler, The Peptides: Analysis, Synthesis, Biology, ed. S. Udenfriend, J. Meienhofer and V. J. Hruby, Academic Press, Orlando, FL, 1985, vol. 7, p. 213.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c3ra43939a

---