1 OriginalPaper 2 LuisFermnCapitn-Vallvey...

J. Sep. Sci. 2008, 31, 3817 –3828 R. Checa-Moreno et al. 3817

Ram�n Checa-Moreno1

Eloisa Manzano2

Gloria Mir�n2

Luis Ferm�n Capit�n-Vallvey2

1Laboratorio Central de SanidadAnimal, Ministerio deAgricultura Pesca yAlimentaci�n, Santa Fe,Granada, Spain

2Solid Phase SpectrometryResearch Group, Department ofAnalytical Chemistry, CampusFuentenueva, Faculty ofSciences, University of Granada,Granada, Spain

Original Paper

Revisitation of the phenylisothiocyanate-derivatives procedure for amino acid determinationby HPLC-UV

A revisitation of the well known chromatographic procedure for the determinationof amino acids as phenylthiocarbamyl derivatives (PTC) is performed. The methodwas developed for a microbore column that it is more appropriate to our later aims,the characterization of proteinaceous binders present in microsamples comingfrom the Cultural Heritage field. Several variables relating to chromatographicaspects were studied such as the pH and temperature of the mobile phase, bufferand modifier (triethylamine) concentrations in mobile phase and the stability ofPTC-derivatives in solution. The calibration function was studied in depth. To pre-vent the heteroscedastic behaviour that it is observed, we used the weighed leastsquares fitting as the best strategy among other normalizing transformations, suchas square root and logarithmic functions. Finally, the proposed method showedresults similar to the traditional method in terms of efficiency, runtime, LODs andother characteristics, but with two additional advantages: a lower mobile phase con-sumption and the possibility of working with a lower sample volume. The useful-ness of proposed method is checked against easel painting samples of Pictorial Her-itage.

Keywords: Heteroscedastic behaviour / HPLC-UV / Phenylthiocarbamyl derivatives / PITC aminoacids determination / Pictorial Heritage /

Received: June 25, 2008; revised: September 9, 2008; September 9, 2008

DOI 10.1002/jssc.200800363

1 Introduction

In paintings, easel, polychrome sculpture, parchmentand the like, the characterization of organic binders,mostly proteinaceous in nature, used because of theirpigment fixative and dispersing properties, is veryimportant because it offers information both for recon-structing the working techniques used and for defining aprogramme for the restoration and conservation of thework of art itself. Different techniques have been usedfor the purpose of identifying the chemical nature andorigin of binders. Some works in the literature provide

methods for characterizing proteins [1, 2]. The most com-mon approach to the specific identification of proteinmedia and their degradation compounds is the use ofchromatographic techniques [3–10].

If we consider the tendency of organic materials toundergo degradation, transformation and oxidationprocesses in addition to the small amount of sampleavailable as well as the small percentage of binder, theneed for an analytical technique with good performancebecomes quite clear. With this goal in mind, in this paperwe study a procedure for amino acid analysis based onthe LC of amino acid derivatives.

The determination of amino acids has usually beendone using ion-exchange chromatography, followed bypost-column derivatization with ninhydrin, but the useof precolumn derivatization procedures and RP HPLCseparation of the derivatives has become widely accepted[11]. This amino acid analysis by HPLC requires a previousderivatization step due to its lack of spectral features. Ahuge number of reagents and procedures have beendescribed for rendering absorbing or fluorescent aminoacid derivatives [12]. The most popular include the use ofreagents such as ortho-phthalaldehyde (OPA) [13], l-dime-

Correspondence: Dr. Luis Ferm�n Capitan-Vallvey, Departmentof Analytical Chemistry, Campus Fuentenueva, Faculty of Scien-ces, University of Granada, E-18071 Granada, SpainE-mail: [email protected]: +34-958-243328

Abbreviations: Ala, alanine; Arg, arginine; Asp, aspartic acid;Glu, glutamic acid; Gly, glycine; HOpr, hydroxyproline; His, his-tidine; hydroxyproline; Ile, isoleucine; Leu, leucine; Lys, lysine;Met, methionine; NOR, norleucine; OPA, ortho-phthalaldehyde;Phe, phenylalanine; PITC, phenyl isothiocyanate; Pro, proline;PTC, phenylthiocarbamyl derivatives; Ser, serine; TEA, triethyl-amine; Thr, threonine; Tyr, tyrosine; Val, valine

i 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.jss-journal.com

3818 R. Checa-Moreno et al. J. Sep. Sci. 2008, 31, 3817 – 3828

thylaminonaphthalene-5-sulphonyl chloride (Dansyl)[14], 4-dimethylaminoazobenzene-49-sulphonyl chloride(Dabsyl) [15], 9-fluorenylmethyl chloroformate (FMOC)[16], 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate(AQC) and phenyl isothiocyanate (PITC) [17–19]. The useof OPA offers high sensitivity capabilities but suffersfrom disadvantages such as the lack of proline (Pro) reac-tion, instability of the fluorescent products such as theOPA-lysine (Lys) derivative and difficulties in quantita-tion due to the sensitivity to quenchers. Quantitation inthe case of Pro and hydroxyproline (HOpr) requires spe-cial oxidative procedures for opening the ring in theseamino acids [20]. Dansyl chloride reaction is slow espe-cially with Pro, so this method has not gained wideacceptance. Something similar occurs with Dabsyl deriv-atives. The FMOC method presents the disadvantage ofreactive interference. The use of PITC (or Edman'sreagent) for amino acid analysis was first described byTarr [21], and Heinrickson and Meredith [22] and thendeveloped commercially by Waters (Milford, MA) underthe name of Pico-Tagm [23], and includes a hydrolysis stepfor proteins. Derivatization is performed by reacting thefree amino acids, under basic conditions, with PITC toproduce first phenylthiocarbamyl (PTC) and later themost stable cycled form, PTH amino acid derivatives(PTH). This derivatization step is rapid and complete withall essential amino acids. PITC reacts equally well withprimary and secondary amines; so, it forms the samechromophore with Pro as it does with primary aminoacids. Consequently, PITC is used for precolumn derivati-zation of both primary and secondary amino acids. RPgradient elution LC is used to separate the PTH deriv-atives, which are then detected by UV absorption. ThePITC derivatization method has been reported to berapid, efficient and reproducible and to provide resultswith most amino acids. Dozens of papers and reviews [12,24 –26] have been published on this topic.

The procedure required for amino acid analysis in theCultural Heritage field should be useful for samples thatweigh as little as tens of milligrams and that add up tofinal volumes of some 50 –100 lL and require injectionvolumes that are not higher than 25 lL. This compels usto change the usual column for amino acids analysissuch as PTH-derivatives (3.9–4.6 mm internal diametercolumns and several lengths) to a microbore column,which makes it possible to use lower injection volumeswithout any loss of sensitivity. In this paper we revisitthe PITC HPLC procedure for amino acid analysis in orderto check and adjust its characteristics to the analysis ofprotein binders in artistic samples.

The PITC derivatization method has been widelyapplied and different modifications have been proposed;nevertheless, the proposed modifications mainly con-cerned previous steps in the method such as hydrolysisor amino acid derivatization [26, 27], but few [28, 29] con-

cern scale reduction with an increase in analytical per-formance. In this work, we revisited this well knownchromatographic method in order to implement it in amicrobore column more appropriate to our objectives.We have considered several influential chromatographicfactors on the separation process such as pH and the tem-perature of the mobile phase and triethylamine (TEA)and buffer concentrations in mobile phase and perform-ed its sequential optimization. Additionally, statisticalfeatures of the analytical method such as range, reprodu-cibility and sensitivity were studied as well.

The proposed method was applied to chromatographicseparation, the identification and quantitative determi-nation of amino acids present in a commercially avail-able amino acid hydrolysed solution with good results.The method described here is actually being used for theidentification of the nature of protein binders present insamples coming from Cultural Heritage with goodresults [30].

2 Experimental

2.1 Apparatus and software

A Hewlett-Packard 1090 model liquid chromatographequipped with a DAD detector monitoring at 254 nm(Waldbronn, Germany) connected to a Pentium 200 PCfitted with 3D ChemStation HPLC (Win95) software fordata acquisition and signal processing was used for allseparation. The column used was a 2.1 mm6200 mmODS AminoQuant from Hewlett-Packard (Seville, Spain)situated in a column heater held at 408C. The mobilephase was ultrasonically degassed at vacuum. The sam-ples were introduced later with an automatic injector. APico Tag workstation from Waters (Milford, MA, USA) foramino acid derivatization was also used.

The Mettler AE 160 analytical balance (Mettler-ToledoAG, Greifensee, Switzerland) was regularly checked withcertified type E2 weights (5 mg, 100 mg and 100 g). Thefixed volume micropipettes (Biohit, Helsinki, Finland)were periodically controlled through gravimetry inorder to ensure the traceability of the results. All the cal-culations were carried out using ChemStation HPLC(Hewlett-Packard) software, Microsoft Excel Spread Sheet(Microsoft Office 97, Ver. 8.0 – 1997, Microsoft Ib�ricaSCR, Spain) and Statgraphics Plus 6.0 (StatisticalGraphics System Corporation, USA, 1992).

2.2 Reagents and standards

All reagents were of analytical-reagent grade unlessstated otherwise. PITC and TEA were purchased fromSigma (Deisenhofen, Germany), hydrochloric acid, ACN(HPLC quality) and acetic acid from Panreac (Montcada iReixac, Barcelona, Spain) and absolute ethanol from


J. Sep. Sci. 2008, 31, 3817 –3828 Liquid Chromatography 3819

Merck (Darmstadt, Germany). Reverse osmosis qualitywater was produced by a Milli-RO and Milli-Q 185 Pluspurification system (Millipore, Bedford, MA, USA).

The individual amino acids considered (Sigma) were:aspartic acid (Asp), glutamic acid (Glu), HOpr, serine(Ser), glycine (Gly), histidine (His), arginine (Arg), threo-nine (Thr), alanine (Ala), Pro, tyrosine (Tyr), valine (Val),methionine (Met), isoleucine (Ile), leucine (Leu), phenyla-lanine (Phe) and Lys. Norleucine (NOR) was used as stand-ard internal. Standard stock solutions of each amino acidwere prepared by adequate weighting and solution in0.1 M HCl. Chromatographic standards were then pre-pared by mixing these solutions appropriately and theuse of the correct amount of NOR as an internal standardfollowed by the dilution necessary to obtain the requiredconcentrations.

We prepared some natural binders to be used as refer-ence materials (egg albumin, milk casein and collagen-like substances). Protein standard preparation similar tobinders used by ancient artists was carried out accordingto old recipes [31] and in this way, we obtained standardsof collagen coming from skins, bones and cartilagesfrom rabbit, standards of albumin coming from chickenegg and casein coming from cow milk.

2.2.1 Preparation of standard and derivatizingsolutions

PTH-amino acid derivatives were prepared according tothe Waters PicoTagm procedure. Standard samples (25 lL)in small tubes (6 mm650 mm) were dried at vacuumand then redried by adding 20 lL of ethanol solution(ethanol/water/TEA, 1:1:1). PITC solution (20 lL; ethanol/water/TEA/PITC, 7:1:2:1) were added to the samples inorder to derivate the amino acids, settled for 10 min,dried again in a Waters PicoTag Workstation, and recon-stituted in 100 lL of sample diluent for analysis. The sam-ple diluent was 0.005 M sodium phosphate buffer,pH 7.42, and 6% ACN. Finally, 5 lL were injected for anal-ysis.

2.3 Chromatographic procedure

The value of the chromatographic conditions optimizedby us for PTH-amino acid separation were: flow rate0.5 mL/min; buffer A: 0.28 M sodium acetate, 0.05% TEA,and 5% ACN (pH 6.38 l 0.04); B: 60% ACN; mobile phaseisocratic at 0% B for 2 min; first linear gradient from 0 to43% B in 7 min; second linear gradient from 43 to 50% in4 min, finally changed to 100% B in 1 min. Washing stepat 100% B isocratic flow for 8 min; returning to initialconditions 0% B in 1 min, and maintenance at 0% B iso-cratic flow for 9 min to the column equilibration step.The total runtime was 32 min.

2.4 Sample treatment

2.4.1 Analytical procedure

Samples containing protein (0.5–1 mg) were extractedin HPLC quality water according to a previously pub-lished extraction procedure [2]. Protein reference sam-ples were solved in water. Then they were hydrolysedand derivatized with PITC, and finally, analysed by HPLCaccording to the procedure described in Section 2.3, inorder to determine their amino acid profile.

2.4.2 Art work samples

Four samples were extracted from easel paintingslocated in the Cathedral Museum of Guadix (Granada,Spain). Samples EP-1 from Madonna with Child (13th C), EP-2 from Eucharist exaltation (18th C), EP-3 from Jesus Christ(16th –17th C) and EP-4 from Eucharist exaltation (18th C),were used in this study. The sampling involved theremoval of paint fragments with a scalpel under a stereo-microscope. A very small sample, typically between 2.5and 4.0 mg, was removed from each easel painting.Before the analyses, samples were stored in suitable con-ditions. Standards and samples coming from easel paint-ings were treated and analysed according to describetreatment procedure to determine the amino acid com-position of the proteinaceous material. In this way, itwas possible to obtain the amino acid profile for eachsample considered.

3 Results and discussion

Of the large number of derivatizing agents described foramino acid determination by HPLC, we selected PITCbecause of its good properties. It produces stable deriv-atives and only one product for each amino acid and thereaction kinetic is very fast and met by all amino acids.Cysteine and its dimer cystine were not considered inthis work due to the fact that their recoveries after pro-tein hydrolysis are poorer. Usually, they are determinedas cysteic acid prior to performic acid oxidation [32] by amodified method that is very time consuming and tedi-ous or using another derivatizing agent such as buthyl-isothiocyanate (BTC) [33]. Something similar occurs totryptophan which is destroyed during the prior hydroly-sis. Alkyl and imino-substituted amino acids were muchless affected as cystine. So, cysteine, cystine and trypto-phan were not considered for performing the character-ization in this work by the reason of its instability thatresults in more alteration by pigments and aging thanthe other amino acids [34]. Most of described applicationsuse regular analytical columns with different lengths(10–30 cm) and 4 mm or higher internal diameter. Theapplication in this study requires using a microbore col-umn with a low internal diameter that makes it possible



to work with low injection volumes and, thus, be able touse the sample in a reduced volume. In this way, the sam-ple reconstitution volume can be the lowest possible,consequently achieving higher sensitivity. From severalcolumns assayed, we selected an Aminoquant200 mm62.1 mm from Hewlett-Packard. Previous worksbased on microbore or narrow-bore columns [28, 29, 35,36] have shown poor results relating to the resolution ofseveral pairs of amino acids like Thr–Ala, Glu–Ser andIle–Leu even though they use longer columns than wehave considered in this work. Furthermore, the optimiza-tion of chromatographic separation has been performedin terms of amino acid derivative resolution aiming toobtain an appropriate value higher than 1.5.

3.1 Optimization procedure

We test different gradients for the column being consid-ered, including that of the original procedure and betterresults in terms of resolution were obtained with (0% B2 min; up to 43% B 7 min; up to 50% 4 min and 100% B1 min). Then, some characteristics of mobile phase andcolumn operation were adjusted and optimized using asequential and single factorial design, with the factorsstudied being: pH, acetate buffer and TEA concentrationsand column oven temperature. These factors stronglyinfluenced both the separation and elution order of theamino acid derivatives [18, 29]. The optimization of thechromatographic separation was carried out startingfrom conditions used with a previously describedmethod [37]. The response variable used in this study wasthe chromatographic resolution of the pairs Arg –Thr,Thr–Ala, Ala–Pro and Ile–Leu, the most difficult couplesto resolve. The value interval of the variables being stud-ied was delimited to prevent reversed elution order fromoccurring. Another factor considered here was the influ-ence of the column equilibration time in the reproduci-bility of the retention time of derivatives. We studied theretention time reproducibility at different equilibrationtimes (0, 4, 7, 10 and 13 min). To do this, three aminoacid standards were injected at each equilibration timeconsidered and the variability of the retention time ofamino acid derivatives was calculated as the variationcoefficient (CV%). A stability study of amino acid deriv-atives in buffer solution at room temperature was car-ried out by injecting the same 50 pmol standard mixturesolution throughout the time consecutively.

3.2 Influence of mobile phase pH on separation

We began using the experimental conditions describedby Bidlingmeyer et al. [23] and a bad chromatographicprofile was obtained with poor resolutions (typically lessthan 0.9), especially for amino acids pairs like Asp–Glu,Ser –Gly, Gly–His, Arg–Thr, Thr–Ala, Ala–Pro and Ile–

Leu. Moreover, several groups of PTH-amino acids weredistinguished in the chromatogram: (i) those near deadvolume with higher polarity such as Asp, Glu and HOpr;(ii) two groups composed of: Ser, Gly and His (the first)and Arg, Thr, Ala and Pro (the second); (iii) a group ofintermediate polarity PITC-amino acids like Tyr, Val,Met, Ile and Leu; and finally (iv) the group with the high-est polarity derivatives Phe and Lys, including NOR (inter-nal standard). Several aqueous mobile phases preparedaccording to the Section 2 were assayed by systematicmodification of their pH value. Variation of pH was per-formed using NaOH to adjust it as necessary, rising from6.04 to 7.03. The mobile phase for amino acid derivativesrequired very precise pH control due to the dependenceof the elution positions of several key amino acids onthis parameter in the above indicated pH range. It can benoted that pH exerts a high influence (Fig. 1A) on theIle–Leu pair showing a plateau at pH interval 6.35–6.60.Pairs Arg–Thr and Thr–Ala present a similar behaviour,decreasing the resolution at pH higher than 6.40. Ala–Pro resolution did not show any variation in the pH inter-val studied. The rest of the PTH-derivative pairs showedgood resolutions, typically higher than 1.5. Thus, weselected 6.38 l 0.02 as working pH, similar value (6.5)than that proposed by Guitart et al. [38] for the sameamino acids, but they employ longer runtimes andobtain poorer resolutions. Lottspeich [36] and Meuth andFox [39] studied the 4.5–5.5 interval and found that pHaffects only the retention of His and Arg whereas noeffect was observed on the rest of amino acids. On theother hand, Somack [18] considered the pH influencealong with the ionic strength of the aqueous mobilephase, concluding that the higher the ionic strength thelower the pH influence on amino acid retention; butnothing was indicated about how this affects the resolu-tion.

3.3 Influence of buffer concentration

Another factor considered was the analytical concentra-tion of acetic acid in the aqueous mobile phase. Severalacetic acid/acetate buffer solutions ranging from 0.15and 0.44 M were assayed (Fig. 1B), selecting 0.28 M as theoptimal buffer concentration. Lottspeich [36] andSomack [18] noted that only the elution of His and espe-cially Arg were affected by the buffer concentration ofthe mobile phase but at a lower concentration rangethan considered here (10–100 mM).

3.4 Influence of column temperature

We considered the oven temperature effect rising fromroom temperature to 458C, and the higher the tempera-ture of the column oven, the higher the resolutionobtained for all pairs of PTH-derivatives studied (Fig. 1C).



The Ala–Pro and Arg–Thr resolution showed depend-ence that seems to go beyond the interval studied,whereas Thr–Ala and Ile–Leu showed constant behav-iour at temperatures higher than 408C. We selected 408Cas the working condition for the oven column becauseabove this temperature, the separation of Pro and urea-derivative is poor. The working temperature selectedhere is lower than values of 47 [40], 52 [17] and 608C [36]previously proposed, which permits a larger column life-time.

3.5 Influence of TEA concentration in mobilephase

The aqueous mobile phase was assayed at different TEAconcentrations whereas the rest of the factors were keptconstant (Fig. 1D). The resolution of PTH-derivative pairsis better in the TEA concentration interval 0.050–0.225%v/v, except for the Ile–Leu pair which is constant. More-over, it was observed that the higher the TEA concentra-tion, the lower the ammonium-Pro derivative resolutionis. At 0.15% TEA, ammonium and Pro derivatives coelute.Better conditions were found in the 0.075 –0.100% inter-val, selecting 0.075% as working condition. The effect ofTEA is explained by an ion-pairing mechanism with sev-eral amino acid derivatives, in addition to protectingfree silanol groups in the stationary phase. Here, wefound that the presence of TEA in mobile phase is veryvaluable in the resolution of Arg–Thr and Thr–Ala deriv-atives.

The optimized conditions for the gradient, pH(6.38 l 0.02), analytical concentration of acetic acid

(0.28 M), temperature of column oven (408C) and TEAconcentration (0.075% v/v) make it possible to obtain avery good resolution (F1.5) for all the pairs of amino-derivatives and, consequently, good and adequate quan-tification is achieved. The chromatographic separationprofile obtained is shown in Fig. 2A.

3.6 Column conditioning

Amino acid derivatives elute at the beginning of thechromatogram followed by other substances like matrixconstituents, derivation by-products, unreacted chemi-cals, reagents and so on; thus, column washing and equi-librating steps are necessary. Column washing makes itpossible to elute substances that have been retained inthe head of the column. These substances can lead togradual poisoning of the column and consequently a lossof column efficiency. Moreover, a column equilibrationstep with an initial mobile phase makes it possible toobtain best reproducibility for the retention times of theamino acid derivatives.

The column washing step after elution was performedby assaying several intervals (4, 8, 12 and 16 min) of 100%B (the stronger elution phase). Eight minutes wasselected as the best equilibration time, which is from 14to 22 min in the run. In this way, the chromatographicselectivity was improved.

Gradient chromatographic separation of amino acidderivatives was carried out in the presence of TEA. Thissupposes that the column must be equilibrated with theinitial mobile phase prior to each injection. If the col-umn is not properly equilibrated, the retention times of


Figure 1. Effect of several mobilephase properties on the resolutionof the worst resolved four pairs ofPTC-derivatives. (A) pH of mobilephase A; (B) analytical concentra-tion of AcH/Ac – buffer in mobilephase A; (C) temperature of col-umn oven; (D) concentration ofTEA in mobile phase A. (a) Ala–Pro; (b) Thr–Ala; (c) Arg–Thr and(d) Ile–Leu.


the amino acid derivatives can rise between injectionsand the retention time variability is higher as the equili-bration time is lower (Fig. 3). Thus, coefficients with a var-iation from 8–21% at no equilibration step (t = 0 min)range to lower than 0.64% when the equilibration time is10 min or higher. These results are in agreement withthose reported by Guitart et al. [38] for several aminoacid derivatives, but Asp and Glu derivatives show thebest results here, 0.34 and 0.58%, respectively, againstsome 1–3% in previously published data [18, 24, 28, 41].The equilibration time proposed here is much lowerthan the 60 min used by Beaver [42]. In brief, the bestchromatographic separation is achieved when the wash-ing step is 8 min at 100% B, and the equilibration step10 min at the initial conditions of the mobile phase (0%B). This good reproducibility in retention times increasesthe methods ability to identify amino acid derivatives inview of its UV-spectrum similarity.

3.7 PTH-amino acid derivative stability in solution

One of the advantages of the analysis of amino acids suchas PTH-derivatives with respect to other derivatizing

agents is its stability. Samples can be derivatized andkept during some time prior to analysis. But this stabilityis only relative. Many researchers have considered thestability of dried PTH-amino acids in refrigeration andfreezing conditions, reporting different lifetimes fromone to several months kept freeze–dried [23, 27, 43]. Inbuffer solution lifetime data reported are very different:6 [23], 8 [41], 10 [33], 12 [44], 24 [41] and 32 h [45]. In viewof this, we decided to perform a stability study of thereconstituted derivatives in buffer solution at room tem-perature. The instrumental signal (relative to IS) showeda linear tendency over time with a negative slope foreach PTH-derivative. Several amino acids like HOpr, Gly,Met and Leu do not show almost any variation in theinterval considered, whereas Asp, Glu, Tyr, Phe andmainly Val and Ile exhibit a higher negative slope. OnlyPro presents the opposite behaviour with a positiveslope, which could be due to the coelution of PTH-ammo-nium coming from the decomposition of other PTH-amino acids. The absolute signal of NOR shows a similarperformance for all the derivatives. The signal decayrelating to NOR was adjusted to a linear model for all theamino acids, and thus the ordinate and slope wereobtained (Table 1).

The estimation of PTH-derivative lifetime in buffer wascarried out taking into account the reproducibility of themethod. Any variation due to random errors would beinside the confidence interval: Signal (t0) l 2Srepet

(P = 95%), where Srepet is the reproducibility expressed asmethod repeatability. Then, the lifetime of each PTH-derivative can be obtained as a cut-off between its linearrelative response in the time with upper or lower lines ofthe confidence interval, as shown in Fig. 4 for Val. Table 1shows the stability values (tmax) obtained in this way foreach derivative amino acid. Lifetimes vary between sev-eral thousand hours and 13 h for Val. Consequently, thesamples were dissolved for no longer than 13 h beforeinjection.


Figure 2. (A) Amino acid profile obtained, 100 pmol of eachamino acid derivative injected. Experimental conditions: flowrate 0.5 mL/min; buffer A: 0.28 M sodium acetate, 0.05%TEA, and 5% ACN (pH 6.38 l 0.04); B: 60% ACN; mobilephase isocratic at 0% B for 2 min; first linear gradient from 0to 43% B in 7 min; second linear gradient from 43 to 50% in4 min, finally change to 100% B in 1 min. (B): Amino acidprofile obtained for EC-1 artwork sample. Experimental con-ditions as previously.

Figure 3. Variability of retention time as CV (%) with equili-bration time in initial mobile phase.


3.8 Study of calibration function

Finally, the validation of the method was performedthrough calibration. For this, different analytical param-eters such as linear range, slope, ordinate at origin, preci-sion and LOD were determined. To establish the functionrelating the chromatographic parameter to the concen-tration we devised and performed experiments consist-ing of the derivatization and subsequent chromato-graphic analysis of mixtures of the 17 selected aminoacids at twelve concentrations ranging from 5 to500 pmol (on column) with five replicates at each con-centration in a random arrangement. Concentrationshigher than 500 pmol were not assayed since the aminoacids from the proteins to be investigated are not presenttogether in proportions higher than two orders of mag-nitude.

3.9 Calibration function

The dataset obtained at each concentration, area andheight, was tested with the Q-Dixon statistical test inorder to reject the anomalous data. Then, we estimatedthe SD at each concentration for all amino acids. Wenote that the precision data used here refer to repeatabil-ity, because the analysis was performed under constantconditions, during a short interval of time, in one labora-tory and by only one operator always using the same

equipment. The chromatographic data evaluated in thiswork are peak heights rather than areas.

3.9.1 Heteroscedasticity

The data obtained in the above experiment exhibited alinear relationship between peak height and concentra-tion, but the variance was not constant for any of theamino acids studied, changing regularly with the con-centration. This feature does not seem to have beendescribed before. Since the variance is not constant, theheteroscedasticity condition was then investigated byplotting the residuals from the regression line obtainedfrom an ordinary least-squares treatment against theconcentration, with all the amino acids studied showingthe typical funnel shape. The heteroscedasticity condi-tion has been demonstrated for all analytes using Bar-tlett's test for the equality of the variances [46].

Heteroscedasticity implies that the direct applicationof the ordinary least squares technique produce devia-tions in precision estimation. Therefore, from severalpossible alternatives allowing the use of more sophisti-cated regression techniques than ordinary least squares,we tried the weighted regression technique and differentscale transformations [47, 48]. The more usual normaliza-tion transformations are logarithmic or potential func-tions. The weighted regression assigns less weight to datawith a large variance than to data with smaller one.Weighting factors, wi, are inversely proportional to thevariance (S2

i ) at each concentration. The weighed leastsquares technique does not substantially alter the slopeestimate, b, but it has a large effect on the precision esti-mated, especially at lower concentrations, where moreprecise results are required, as in our case.


Table 1. Parameters for estimating PTH-derivative lifetime inbuffer solution: Srepet, linear decay function, Ao l 2Srepet (inter-cept of linear decay function and confidence level band atrepeatability conditions) and lifetime (hours)

Aminoacid deriv-ative

Srepet Decay linear functionof relative signal

Ao l2Srepet

tmax

(h)

O.O. Slope

Asp 0.101 0.914 –0.0051 0.712 40Glu 0.149 0.907 –0.0061 0.608 49HOpr 0.184 1.097 0.0027 1.466 961Ser 0.110 0.991 –0.0029 0.771 76Gly 0.121 1.024 0.0009 1.267 2449His 0.156 0.893 –0.0034 0.580 91Arg 0.162 1.123 –0.0028 0.798 116Thr 0.116 1.024 –0.0046 0.791 50Ala 0.137 1.018 –0.0040 0.744 68Pro 0.201 1.006 0.0090 1.409 269Tyr 0.099 1.055 –0.0063 0.854 32Val 0.091 1.028 –0.0137 0.844 13Met 0.129 1.049 0.0003 1.308 7558Ile 0.090 0.948 –0.0098 0.767 19Leu 0.119 1.014 0.0029 1.252 778Phe 0.100 0.890 –0.0062 0.688 32Lys 0.253 1.748 –0.0064 1.233 81

Srepet: SD expressed as repeatability. O.O.: ordinate at origin.

Figure 4. Relative signal decay for Val-derivative with time.Dotted lines: confidence intervals of relative signal at time0 h.


3.9.2 The variance modelling

To obtain the SD at each concentration value, the SDswere fitted to a first and a second-order function by ordi-nary least squares. To select the best fit, we used the ten-dency of the residuals plot and an ANOVA [49]. Table 2shows the function selected for each analysed aminoacid after prior rejection of data with a test of outliers[50]. Here, the model was empirically obtained using SDrather than variance because the best results wereobtained with the former.

3.9.3 Weighed calibration

From the different forms to define the weighting factorswi we use [47] wi ¼ k½siðxiÞ�2 where k is a normalization fac-tor defined as k ¼ nP

sðxiÞ�2 and subject to the conditionP

wi ¼ n. However, the use of the inverse of SD or var-iance as a weighting factor leads to worse results. In the

weighed least squares method, the sum to be minimizedisP½wiðyi � yyiÞ�2 ¼ minimum, where ðyi � yyiÞ are the

residuals.In order to apply weighed least squares, the modelled

or experimental variances must be used to calculate theweighting factors. To check the linear fit, we used a lack-of-fit test valid for the heteroscedastic case [51]. Table 2indicates the results obtained using a noniterativeweighed least squares analysis [30] with the modelledweighting factor.

3.9.4 Scale transformations

As an alternative to weighted calibrations we tried scaletransformations, namely logarithmic (ln (100 Y) =a + b ln X) and square root (Y = a + bX ). We use ln (100 Y)that means ln (100) + ln Y rather than ln Y to avoid theuse of negative values. Y is the response variable and X


Table 2. Comparison between weighed (WC) and ordinary least square (ORL) calibrations for PTH-derivatives

Amino acid Calibrationfunction

SD functiona) b r Vres D.L. Precision at two levels ofconcentration (pmol)

5 pmol 500 pmol

Asp WC Linear 0.0123 0.998 0.00003 a1 1 20OLS n/a 0.0117 0.996 0.02992 14 20 26

Glu WC Linear 0.0112 0.998 0.00007 1 1 19OLS n/a 0.0105 0.996 0.02596 15 21 27

HOpr WC Quadratic 0.0115 0.999 0.00001 a1 a1 14OLS n/a 0.0115 0.999 0.00427 5 8 10

Ser WC Linear 0.0108 0.998 0.00129 3 4 18OLS n/a 0.0101 0.998 0.01400 11 16 21

Gly WC Linear 0.0119 0.999 0.04109 2 3 15OLS n/a 0.0113 0.998 0.01232 9 14 17

His WC Linear 0.0132 0.999 0.00093 2 3 12OLS n/a 0.0129 0.999 0.00964 7 10 14

Arg WC Quadratic 0.0132 0.999 0.00089 2 3 12OLS n/a 0.0129 0.999 0.00964 7 10 14

Thr WC Linear 0.0151 0.998 0.00052 1 2 19OLS n/a 0.0146 0.999 0.01351 8 11 14

Ala WC Quadratic 0.0151 0.999 0.00173 3 4 14OLS n/a 0.0100 0.998 0.10237 14 10 a1

Pro WC Quadratic 0.0170 0.999 0.00084 2 2 14OLS n/a 0.0176 0.999 0.01128 6 8 11

Tyr WC Linear 0.0206 0.999 0.00093 1 2 17OLS n/a 0.0202 0.998 0.04623 10 15 19

Val WC Quadratic 0.0174 1.000 0.00071 1 2 9OLS n/a 0.0170 0.999 0.01363 7 10 12

Met WC Linear 0.0169 1.000 0.00044 1 2 9OLS n/a 0.0163 0.999 0.01612 7 11 14

Ile WC Quadratic 0.0160 0.999 0.00149 2 3 13OLS n/a 0.0153 0.999 0.00893 6 9 11

Leu WC Quadratic 0.0163 0.999 0.00061 1 2 11OLS n/a 0.0152 0.999 0.01546 8 11 15

Phe WC Linear 0.0150 0.999 0.00044 1 2 15OLS n/a 0.0151 1.000 0.00420 4 6 8

Lys WC Quadratic 0.0220 0.998 0.00332 3 4 20OLS n/a 0.0224 0.998 0.04991 11 14 19

b: calibration slope; r: correlation coefficient; Vres: residual variance and D.L.: LOD (IUPAC criterion); n/a: not applicable.a) Type of fitting for SD used in weighed calibration.


the independent variable (amino acid concentration). Inboth cases, the variance is stabilized and the heterosce-dastic tendency disappears with a logarithmic transfor-mation as shown in the plot of residuals versus concentra-tion. The disappearance of heteroscedasticity now makesit possible to use ordinary least squares to obtain the cali-bration function. Table 3 shows the results for logarith-mic and square root transformations.

3.9.5 Comparison of results

The reported values of the LODs for the PITC methodfound up to now in the literature show few pmol for allamino acids based on an S/N and a precision, expressedas RSD, of 1.5% as an average. However, the LODsobtained by us using unweighed calibrations as thereported above give values more like 7–15 pmol ratherthan 1 pmol (see Table 2).

The range of the method does not vary with the type ofcalibration in the studied interval. The upper limit of thecalibration function was tested using linearity and lack-of-fit tests, proving that in all cases the (up to at least500 pmol) calibration function was linear.

The precision increases through the use of transforma-tions such as the logarithm and the square root. Ofcourse it is necessary to back-transform the logarithmicor square root data used to be able to compare the data,as we have done in Table 3. The use of a weighted calibra-tion improves the precision as well, but to a lesser extentthan the functional transformations (see Table 1). Notethat the reproducibility is worse for the amino acids Aspand Glu, probably because they elute first on the chroma-tograms and their peaks are more asymmetric.

In short, the use of a suitable functional transformationsuch as logarithmic and/or weighted calibrations im-


Table 3. Comparison between logarithmic (LnT) and square root (SRT) transformation on calibration for PTH-derivatives

Amino acid Transformation r Vres D.L. Precisiona) (CV%)

5 pmol 500 pmol

Asp LnT 0.9993 0.0032 1.1 0.1 0.1SRT 0.9981 0.0021 0.3 0.7 0.7

Glu LnT 0.9993 0.0032 1.1 0.1 0.1SRT 0.9979 0.0010 0.4 0.7 0.7

HOpr LnT 0.9997 0.0013 1.1 0.1 0.1SRT 0.9996 0.0007 0.1 0.3 0.3

Ser LnT 0.9992 0.0032 1.1 0.1 0.1SRT 0.9989 0.0012 0.2 0.5 0.5

Gly LnT 0.9996 0.0018 1.1 0.1 0.1SRT 0.9992 0.0008 0.1 0.4 0.4

His LnT 0.9997 0.0014 1.1 0.1 0.1SRT 0.9995 0.0005 0.1 0.3 0.4

Arg LnT 0.9997 0.0014 1.1 0.1 0.1SRT 0.9979 0.0004 0.4 0.7 0.7

Thr LnT 0.9993 0.0032 1.1 0.1 0.1SRT 0.9991 0.0005 0.2 0.4 0.5

Ala LnT 0.9995 0.0020 1.1 0.1 0.1SRT 0.9994 0.0001 0.1 0.4 0.4

Pro LnT 0.9997 0.0014 1.1 0.1 0.1SRT 0.9995 0.0001 0.1 0.3 0.4

Tyr LnT 0.9994 0.0025 1.1 0.1 0.1SRT 0.9989 0.0001 0.2 0.5 0.5

Val LnT 0.9995 0.0020 1.1 0.1 0.1SRT 0.9996 0.0001 0.1 0.3 0.3

Met LnT 0.9998 0.0011 1.1 0.1 0.1SRT 0.9995 0.0001 0.1 0.3 0.4

Ile LnT 0.9996 0.0016 1.1 0.1 0.1SRT 0.9997 0.0001 0.1 0.3 0.3

Leu LnT 0.9994 0.0025 1.1 0.1 0.1SRT 0.9994 0.0001 0.1 0.4 0.4

Phe LnT 0.9989 0.0045 1.2 0.1 0.1SRT 0.9997 0.0001 0.1 0.3 0.3

Lys LnT 0.9996 0.0014 1.1 0.1 0.1SRT 0.9990 0.0001 0.2 0.5 0.5

b: calibration slope; r: correlation coefficient; Vres: residual variance and D.L.: LOD (IUPAC criterion).a) Precision estimated as repeatability at two concentrations.


proves the sensitivity and precision of amino acid analysissuch as PTH-derivatives. This is very interesting for thecharacterization of proteinaceous binding media

through amino acid profiles where we normally have towork with a very small amount of sample (microgram orsub-microgram). In these cases, the availability of precise


Figure 5. Relative amino acid percentages and amino acid ratio flow chart versus data of percentage content of amino acidsfrom collagen, casein and albumin protein standards.


and sensitive analytical methods is greatly recom-mended.

3.10 Applications

The proposed method was applied to chromatographicseparation, identification and quantitative determina-tion of amino acids present in a commercially availableamino acid hydrolysed solution with good results. Theresults achieved have shown that chromatographic sep-aration is acceptable with resolutions higher than 1.5 forall PITC-amino acid derivatives. Furthermore, from aquantitative point of view, the recovery found in a com-mercially available hydrolysed solution was 107% onaverage. The method described here is currently beingused with good results (e.g. Fig. 2B) for the identificationof protein binder present at samples coming from Cul-tural Heritage [30].

Additionally, we tested the proposed method for theidentification of proteinaceous binders in easel paint-ings located in the Cathedral Museum of Guadix (Gran-ada, Spain). To identify the proteinaceous binder presentin the painting samples the relative amino acid percen-tages and amino acid ratio flow chart was used togetherwith the data of percentage content of 15 amino acids(Asp, Glu, HOpr, Ser, Gly, His, Arg, Thr, Ala, Pro, Tyr, Val,Met, Ile, Leu, Phe, and Lys) obtained with protein binderstandards (collagen, casein, albumin) prepared using tra-ditional recipes [31] and the samples coming from paint-ing under study (Fig. 5).

All the samples show high Gly content and the pres-ence of HOpr, characteristic of collagen (animal glue)presence. The content of the most labile amino acids,such as Met and Lys, was lower than expected, thus high-lighting a degradation of the protein due to both naturalageing and to chemical processes derived from the par-ticular environmental conservations conditions [30]. Theamino acid profile of the most stable amino acids, evenwhen pigments are present in the sample [34], as Ala, Val,Ile, Leu, Gly, Pro and HOpr and values of some ratio stud-ied, supports that samples EP-1, EP-2, EP-3 and EP-4closely resembled that of our reference sample of colla-gen.

4 Concluding remarks

To improve sensitivity and to achieve better reproducibil-ity and quantitation for routine analysis, we used amicro LC system for the separation of PTC-amino acids.We revisited a well known and long-used chromato-graphic method for the determination of amino acids asPTC-derivatives. The classical method was adapted to amicrobore column which makes it possible to work witha lower sample volume (5–10 lL). Other advantages

include chromatographic efficiency that is good with res-olutions higher than at least 1.5 at the same runtimes asthe previously described methods. Similar chromato-graphic sensitivity as previously described methods isachieved using a lower sample injection volume whichmakes is possible to work with very small sampleamounts as is usual in Cultural Heritage protein binderidentification. The weighed calibration proposed herehelps us to know the amino acid content in a samplewith higher precision. Furthermore, the revisitedmethod lets us save the mobile phase because the chro-matographic flow is lower than that used in higher inter-nal diameter columns. It has been calculated that the sav-ings can vary around 50 –70%, as this method uses amobile phase flow of 0.5 mL/min in comparison with0.8–1.5 mL/min used with 4.6 mm id columns. Thisaspect is important not only economically, but also eco-logically from the point of view of the waste residues pro-duced. The procedure here studied offer good results inthe identification of proteinaceous binders in easelpaintings.

We acknowledge financial support from the Ministerio de Educa-ci�n y Cultura, Direcci�n General de Ense�anza Superior (Spain)from Projects BHA2003-08671 and HUM2006-09262, coordinatedby Dr. L. R. Rodr�guez-Sim�n.

The authors declared no conflict of interest.

5 References[1] Masschelein-Kleiner, L., in: Brommelle, N.,Smith, P. (Eds.), Conser-

vation and Restoration of Pictorial Art, Butterworths, London 1976,pp. 105 – 183.

[2] Halpine, S. M., Stud. Conserv. 1992, 37, 22 – 38.

[3] Colombini, M. P., Modugno, F., J. Sep. Sci. 2004, 27, 147 – 160.

[4] Yang, Y., Kondo, T., Kennedy, T. J., J. Chromatogr. Sci. 2005, 43,518 – 521.

[5] Peris-Vicente, J., Gimeno Adelantado, J. V., Carbo, M. T. D., Cas-tro, R. M., Reig, F. B., Talanta 2006, 68, 1648 – 1654.

[6] Nevin, A., Cather, S., Anglos, D., Fotakis, C., Anal. Chim. Acta 2006,573+574, 341 – 346.

[7] Pitthard, V., Griesser, M., Stanek, S., Ann. Chim. 2006, 96, 561 –573.

[8] Prikryl, P., Havlickova, L., Pacakova, V., Hradilova, J., Stulik, K.,Hofta, P., J. Sep. Sci. 2006, 29, 2653 – 2663.

[9] Gautier, G., Colombini, M. P., Talanta 2007, 73, 95 – 102.

[10] Nevin, A., Cather, S., Burnstock, A., Anglos, D., Appl. Spectrosc.2008, 62, 481 – 489.

[11] Rekha, S. D., Maryanoff, C. A., in: Swadesh, J. K. (Ed.), HPLC Practi-cal and Industrial Applications, CRC Press, Boca Raton, FL 1997,p.111.

[12] Fountoulakis, M., Lahm, H. W., J. Chromatogr. A 1998, 826, 109 –134.

[13] Hill, D. W., Walters, F. H., Wilson, T. D., Stuart, J. D., Anal. Chem.1979, 51, 1338 – 1341.

[14] Hsu, K. T., Currie, B. L., J. Chromatogr. 1978, 166, 555 – 561.

[15] Stocchi, V., Piccoli, G., Magnani, M., Palma, F., Biagiarelli, B.,Cucchiarini, L., Anal. Biochem. 1989, 178, 107 – 117.



[16] Einarsson, S., Folestad, S., Josefsson, B., Lagerkvist, S., Anal. Chem.1986, 58, 1638 – 1643.

[17] Moser, P. W., Rickli, E. E., J. Chromatogr. 1979, 176, 451 – 455.

[18] Somack, R., Anal. Biochem. 1980, 104, 464 – 468.

[19] Sottrup-Jensen, L., Petersen, T. E., Magnusson, S., Anal. Biochem.1980, 107, 456 – 460.

[20] Bohlen, P., Mellet, M., Anal. Biochem. 1979, 94, 313 – 321.

[21] Tarr, G. E., Meth. Enzymol. 1977, 47, 335 – 357.

[22] Heinrikson, R. L., Meredith, S. C., Anal. Biochem. 1984, 136, 65 – 74.

[23] Bidlingmeyer, B. A., Cohen, S. A., Tarvin, T. L., J. Chromatogr. 1984,336, 93 – 104.

[24] Cohen, S. A., Strydom, D. J., Anal. Biochem. 1988, 174, 1 – 16.

[25] Molnar-Perl, I., J. Chromatogr. A 1994, 661, 43 – 50.

[26] Woo, K. L., Ahan, Y. K., J. Chromatogr. A 1996, 740, 41 – 50.

[27] Janssen, P. S. L., Van Nispen, J. W., Melgers, P. A. T. A., Van denBogaart, H. W. M., Hamelinck, R. L. A. E., Goverde, B. C., Chroma-tographia 1986, 22, 345 – 350.

[28] Cunico, R. L., Simpson, R., Correia, L., Wehr, C. T., J. Chromatogr.1984, 336, 105 – 113.

[29] Lu, H. S., Lai, P. H., J. Chromatogr. A 1986, 368, 215 – 231.

[30] Checa-Moreno, R., Manzano, E., Miron, G., Capitan-Vallvey, L. F.,Talanta 2008, 75, 697 – 704.

[31] Cenini, C., The Craftsman's Handbook. Il Libro dell'Arte, Dover Publi-cations, Inc., New York 1960.

[32] Tarr, G. E., in: Shively, J. E. (Ed.), Methods of Protein Microcharacteri-zation, Humana Press, Clifton NJ 1986, pp. 155 – 194.

[33] Woo, K. L., Hwang, Q. C., Kim, H. S., J. Chromatogr. A 1996, 740,31 – 40.

[34] Schilling, M. R., Khanjian, H. P., J. Am. Inst. Conserv. 1996, 35, 123 –144.

[35] Lu, H. S., Klein, M. L., Lai, P. H., J. Chromatogr. 1988, 447, 351 – 364.

[36] Lottspeich, F., J. Chromatogr. A 1985, 326, 321 – 327.

[37] Cohen, S. A., Tarvin, T. L., Bidlingmeyer, B. A., Am. Lab. 1984, 48 –59.

[38] Guitart, A., Hernandez Orte, P., Cacho, J., Analyst 1991, 116, 399 –403.

[39] Meuth, J. L., Fox, J. L., Anal. Biochem. 1986, 154, 478 – 484.

[40] Hayakawa, K., Hirano, M., Yoshikawa, K., Katsumata, N., Tanaka,T., J. Chromatogr. A 1999, 846, 73 – 82.

[41] Guitart, A., Hernandez-Orte, P., Cacho, J., Quim. Anal. 1996, 15,217 – 223.

[42] Beaver, R. W., Wilson, D. M., Jones, H. M., Haydon, K. D., J. AOAC1987, 70, 425 – 428.

[43] Woo, K. L., Ahan, Y. K., J. Chromatogr. A 1996, 740, 41 – 50.

[44] Inglis, A. S., Bartone, N. A., Finlayson, J. R., J. Biochem. Biophys.Methods 1988, 15, 249 – 254.

[45] Feste, A. S., J. Chromatogr. 1992, 574, 23 – 34.

[46] Meier, P. C., Zund, R. E., Statistical Methods in Analytical Chemistry, J.Wiley, New York 1993.

[47] Oppenheimer, L., Capizzi, T. P., Weppelman, R. M., Mehta, H.,Anal. Chem. 1983, 55, 638 – 643.

[48] Cela, R., Avances en Quimiometr�a Pr�ctica, Ed. Pub. Univ. Santiagode Compostela, Santiago de Compostela (Spain) 1994.

[49] Mart�n Andr�s, A., Luna del Castillo, J. D., Bioestad�stica para lasCiencias de la Salud, Ed. Norma, Madrid 1994.

[50] Analytical Methods Committee, Analyst 1994, 119, 2363 – 2366.

[51] Watters, R. L., Jr., Carroll, R. J., Spiegelman, C. H., Anal. Chem.1987, 59, 1639 – 1643.


1 OriginalPaper 2 LuisFermnCapitn-Vallvey...

Documents

Transcript of 1 OriginalPaper 2 LuisFermnCapitn-Vallvey...