Gas chromatographic and mass spectrometric investigations of organic residues from Roman glass unguentaria

Page 1

A

Guttidombywu©

K

1

pprt[iha

0d

Available online at www.sciencedirect.com

Journal of Chromatography A, 1183 (2008) 158–169

Gas chromatographic and mass spectrometric investigationsof organic residues from Roman glass unguentaria

Erika Ribechini a,∗, Francesca Modugno a,Maria Perla Colombini a, Richard P. Evershed b

a Dipartimento di Chimica e Chimica Industriale, University of Pisa, Via Risorgimento 35, 56126 Pisa, Italyb Organic Geochemistry Unit, Bristol Biogeochemistry Research Centre, School of Chemistry,

University of Bristol, Cantock’s Close, Bristol BS8 1TS, UK

Received 24 October 2007; received in revised form 10 December 2007; accepted 21 December 2007Available online 16 January 2008

bstract

A combination of gas chromatographic (GC) and mass spectrometric (MS) techniques, including direct exposure-MS (DE-MS), high-temperatureC–MS (HTGC–MS) and GC–MS of neutral and acid fractions, was employed to study the composition and recognise origin of the organic materialssed to manufacture balm residues surviving in a series of glass unguentaria recovered from excavations of a Roman villa (Villa B) in the ancientown of Oplontis (Naples, Italy). DE-MS provided comprehensive ‘fingerprint’ information on the solvent soluble components of the contents ofhe unguentaria, while GC–MS analyses provided detailed molecular compositions, highlighting the presence of a wide range of compound classesncluding mid- and long-chain fatty acids, long-chain hydroxy-acids, n-alkanols, alkandiols, n-alkanes, long-chain monoesters, phytosterols anditerpenoid acids. Characteristic biomarkers and their distributions indicate the presence of beeswax, Pinaceae resin and another wax, as the mainrganic constituents of all of the preparations examined. In particular, the occurrence of phytosterols and long-chain monoesters, in which the acyloiety was not exclusively palmitic acid, suggested the presence of a second waxy-lipid constituent of plant origin. The results are consistent with

eeswax being used in the preparation of the cosmetics preserved in the unguentaria, while the other lipids are most likely the residue of some as

et unidentified plant extract(s), possibly deriving from the cuticular waxes of flowers and/or leaves. The composition of the extracts are consistentith the ancient practices of maceration and/or “enfleurage”, in which lipid-based materials, such as beeswax, animal fat or vegetables oils, weresed to extract aromatic and fragrant substances from resin, flowers, spices and scented wood, in order to produce unguents and balms. 2008 Elsevier B.V. All rights reserved.

y-lip

itatti

m

eywords: DE-MS; HTGC–MS; GC–MS; Beeswax; Pinaceae resin; Plant wax

. Introduction

Ancient societies were fastidious, just like today, aboutersonal hygiene and their appearance. Consequently, balms,erfumes, cosmetic and remedial preparations had a prominentole both in terms of their aesthetic, magical and ritual func-ions, and because of their therapeutic and medicinal properties1]. The use of scented unguents, ointments, creams and cosmet-

cs with aesthetic and pharmaceutical properties is attested byistorical documents from ancient Egypt, Mesopotamia, Cyprusnd Crete, not only in religious ceremonies and rituals but also

∗ Corresponding author. Tel.: +39 0052219312.E-mail address: [email protected] (E. Ribechini).

tieriOn

021-9673/$ – see front matter © 2008 Elsevier B.V. All rights reserved.oi:10.1016/j.chroma.2007.12.090

id; Archaeological glass unguentaria

n everyday use. The science of cosmetics already existed inhe Mediterranean area during the old Kingdom in Egypt [1–3]nd continued into the Greco-Roman period [4]. Interestingly,he spread of cosmetics throughout the Roman Empire and theechnical improvements in their fabrication coincided with thenvention of blown glass in the 1st century BC.

On the basis of ancient documents [3,4] and of modern cos-etology, it can be presumed that since Egyptian and Roman

imes, extracts of plants and animals have been fundamentalngredients of balms, creams and ointments. In spite of the recov-ry of numerous archaeological vessels, such as unguentaria,

eeds and canisters, very little information is available concern-ng the chemical compositions and origins of such substances.ccasionally, such vessels contain residues of the origi-al preparations, offering opportunities to chemically analyse

Page 2

atog

tpusartppocmimipttictiuWucvguocsmp[

tgeTobPtfn(wufscooorc

tiMctabpohptrvcihttttdant

2

2

2

gaIN1(

2

t7fua7tna flattened fold on top.

In all the unguentaria, the organic remains were found as

E. Ribechini et al. / J. Chrom

heir contents in order to provide new insights into ancientreparation techniques. Such investigations have revealed these of minerals, natural pigments and also of wet-chemicallyynthesised inorganic salts, in Egyptian and Roman cosmeticsnd pharmaceutical preparations [5–8]. However, with someare exceptions, such as the cosmetic contained in a Romanin canister that was completely characterised [7], it is rarelyossible to confirm the origin of the organic materials com-rising the contents of such containers. The recovered residuesf archaeological cosmetic and medical preparations are oftenomplex mixtures of aged/degraded natural and sometimesan-made substances. Their preparation would certainly have

nvolved a range of physical and chemical treatments and theixing of a variety of natural ingredients before use, lead-

ng to marked changes in composition compared with thearent components. The complexity of such matrices makeshem extremely difficult to characterize completely, particularlyhe organic constituents, which certainly would have been anmportant part of balms, creams and unguents, as in modernosmetology. Although the use of minerals and pigments inhe preparation of make-up has been the focus of several stud-es [5–8,11], very little is known about the organic materialssed in the manufacture of beauty and pharmaceutical products.hat work has been undertaken [7,9,10], has highlighted the

se of natural resins and, animal and vegetable fats/oils. Furtherhemical studies of such rare archaeological finds will providealuable information concerning the technical, social and reli-ious contexts in which the preparations were formulated andsed. Such studies will also help to improve our knowledgef ancient cosmetic and pharmaceutical recipes; for example,hemical analyses of pigments in beauty products have demon-trated that the Egyptians mastered sophisticated practices foranipulating the chemistry of aqueous solutions, being able to

roduce synthetic salts for cosmetic and/or therapeutic purposes6,11].

In this paper we present the results of investigations ofhe chemical composition of organic residues from severallass ointment jars (1st century BC to 1st century AD) recov-red at the archaeological site of Oplontis (Naples, Italy).he site of Oplontis is in the middle of the modern townf Torre Annunziata. The excavations of Oplontis Villa By Sovrintendenza Archeologica (archaeological authority) ofompeii brought to light an exceptional series of glass unguen-

aria, still containing residues of their original content in theorm of solid or semisolid materials, found close together andow conserved at National Archaeological Museum of NaplesItaly). Ancient historical sources reveal that the Vesuvian areaas once considered important for the production of glassnguentaria, and for the manufacture of cosmetics and per-umes. The shape and size of the vessels found in Oplontisuggest that they were likely to have contained balms andosmetics. The contents of these vessels provide a uniquepportunity to study the formulation of cosmetic preparations

f the Romans. This investigation focussed of the contentsf seven glass unguentaria. Due to the wide range of mate-ials that might be present in such balms, and the expectedomplexity of the composition of the residues, the inves-

ablt

r. A 1183 (2008) 158–169 159

igations of the origins of the amorphous organic residuesnvolved the use of DE-MS, HTGC–MS and GC–MS. DE-

S was adopted as a rapid screening technique to provideomprehensive overviews of chemical composition withouthe need for sample pre-treatment. The spectra obtainedllow broad classification of the organic substances presentased on average mass spectra of volatilised/pyrolysed com-onents. The utility of this method to highlight the presencef resinous materials and beeswax in archaeological residuesas been demonstrated [12–14], although in the case of com-lex mixtures DE-mass spectra only provide information onhe most abundant components [14]. Guided by the DE-MSesults, GC–MS analyses were undertaken in order to pro-ide detailed compositional information on mixtures of organicompounds, such as those expected to be found in archaeolog-cal residues from cosmetic or medicinal preparations. GC–MSas been successfully applied for more than two decades inhe characterization of archaeological organic residues con-aining lipids, natural waxes and terpenic resins [2,15–17]. Inhis study high-temperature gas chromatography mass spec-rometry (HTGC–MS) was applied to total lipid extracts, aftererivatisation with BSTFA (N,O-bis(trimethyl)silyltrifluoro-cetamide), then GC–MS was used to investigate acidic andeutral fractions (after alkaline hydrolysis, separation andrimethylsilylation).

. Experimental

.1. Samples and materials

.1.1. ChemicalsAll solvents were Carlo Erba (Milan, Italy) pesticide analysis

rade. n-Tetratriacontane (internal standard, I.S.), n-tridecanoiccid (internal standard, I.S.1), n-hexadecane (internal standard,.S.2), hydrochloric acid (HCl), potassium hydroxide (KOH) and,O-bis(trimethyl)silyltrifluoro-acetamide (BSTFA) containing% trimethylchlorosilane, were purchased from Sigma–AldrichMilan, Italy).

.1.2. Archaeological samplesSamples were collected from seven glass unguentaria (inven-

ory numbers 73291, 74229, 74676, 74677, 74678, 74680 and4682) still containing part of their original contents in theorm of a brownish-blackish solid amorphous material. Thenguentaria are transparent light bluish-green coloured and havesimilar shape. As an example, a rendering of unguentarium

4682 is shown in Fig. 1. It is tubular, swelling out at the bot-om to form a conical body. It has a height of about 13 cm, aeck slight widening towards the top and a spreading rim with

morphous residues both adhering to the interior walls and toottom of the vessels. Samples (typically 10–50 mg) were col-ected with a scalpel, deposited in glass vessels and preservedhere until they were chemically analysed.

Page 3

160 E. Ribechini et al. / J. Chromat

2

2

ad[rvio1iMTwcg1wr

2

pp(u

Tcsaiiaeqiw2gcAtt3a

2

a(l(fehn5ulst6anbCP(itwou32T22c

Fig. 1. Rendering of unguentarium 74682.

.2. Methods and instrumentation

.2.1. DE-MSSub-samples (1–3 mg) of the vessel contents were analysed

fter dissolution in dichloromethane (1 mg ml−1) as previouslyescribed for the analysis of diterpenic [13] and triterpenic resins14]. One microlitre of the latter solution was deposited on thehenium filament of the direct exposure probe, and, after sol-ent evaporation, introduced into the MS. The probe was heatedn the current programmed mode, with the maximum currentf 1000 mA, which corresponds to filament temperature of ca.000 ◦C. The desorbed components were ionised by electronmpact (EI, 70 eV) in a Thermo Finnigan Polaris Q ion trap

S, and a total ion current acquired as a function of time.he ion source temperature was maintained at 230 ◦C. The MSas scanned over the range m/z 50–1000. An optimal total ion

urrent (TIC) curve as a function of time was achieved by pro-ramming the probe as follows: 0 mA for 20 s, from 0 mA to000 mA in 2 s, and 60 s at 1000 mA. Mass spectral ‘fingerprints’ere obtained by averaging the mass spectra in the desired time

ange.

.2.2. HTGC–MSTotal lipid extracts were obtained by modification of a

reviously published procedure [18,19]: to the powdered sam-le (5–10 mg) was added 5 �l of n-tetratriacontane in hexane1 mg ml−1) as internal standard and extraction performed byltrasonication with chloroform/methanol (2:1, v/v, 3× 3 ml).

ssw

ogr. A 1183 (2008) 158–169

he total lipid extracts (TLEs) were dried, re-dissolved inhloroform/methanol (2:1, v/v, 1 ml), and filtered through ailica gel 60 column. After solvent evaporation, BSTFA wasdded, and the sample heated to 70 ◦C for 30 min. The result-ng trimethylsilyl derivatives (TMS derivatives) were dissolvedn dichloromethane (50 �l), and 2 �l submitted to HTGC–MSnalysis on a Trace GC (Thermo Electron Corporation, USA)quipped with a PTV injector and connected to a Trace MSuadrupole MS (ThermoElectron Corporation, USA). The PTVnjector was operated in the splitless mode at 100 ◦C. The MSas operated in the EI mode (70 eV, ion source temperature90 ◦C) and set to scan in the range m/z 50–850. Chromato-raphic separation was performed on a fused silica capillaryolumn (25 m length × 0.32 mm I.D.; HT5, SGE, Melbourne,ustralia) coated with 5%-phenylpolycarboranesiloxane (film

hickness 0.1 �m). The GC conditions were as follows: ini-ial temperature 100 ◦C, 1 min isothermal, 10 ◦C min−1 up to50 ◦C, 30 min isothermal. Helium (purity 99.9995%) was useds carrier gas and maintained at a constant flow of 2 ml min−1.

.2.3. GC–MSThe analytical procedure [20] used to obtain the acidic

nd neutral fractions is summarised as follows: a sub-sample1–3 mg) of the unguentaria contents was subjected to alka-ine hydrolysis by adding 1 ml of methanolic KOH [KOHCH3OH10% weight)/KOHH2O (10% weight), 2:3], and heating at 60 ◦Cor 3 h. After hydrolysis, neutral organic components werextracted with n-hexane (3× 500 �l) and, after acidification withydrochloric acid (10 M; to pH 2), the acidic organic compo-ents were extracted from the hydrolysate with diethyl ether (3×00 �l). Aliquots of both fractions were evaporated to drynessnder a gentle stream of nitrogen and subjected to trimethylsi-ylation. This was achieved by mixing the dried aliquots with aolution of internal standard (5 �l of n-tridecanoic acid solu-ion, 140 �g g−1) and derivatising with 20 �l of BSTFA (at0 ◦C, 30 min), using 150 �l iso-octane as the solvent. Afterdding 10 �l of n-hexadecane solution (80 �g g−l) as an inter-al standard for the injection, 2 �l of the solution were analysedy GC–MS. The GC–MS was a Trace GC (ThermoElectronorporation) equipped with a PTV injection port, linked to aolaris Q (ThermoElectron Corporation) ion trap-MS detectorEI 70 eV, ion source temperature 230 ◦C, scanning m/z 50–650,nterface temperature 280 ◦C). The PTV injector was operated inhe “constant temperature splitless with purge” mode at 280 ◦Cith a purge pressure of 100 kPa. GC separation was performedn an HP-5MS chemically bonded fused silica capillary col-mn (Hewlett Packard; 5% phenyl 95% methylpolysiloxane,0 m × 0.25 mm I.D., 0.25 �m film thickness, connected to am deactivated fused silica capillary pre-column, I.D. 0.32 mm).he GC conditions were as follows: initial temperature 80 ◦C,min isothermal, 10 ◦C min−l up to 200 ◦C, 6 ◦C min−1 up to80 ◦C, 35 min isothermal. Carrier gas: He (purity 99.9995%),onstant flow 1.2 ml min−1.

Peak assignments were performed on the basis of the analy-is of available standard compounds, by interpretation of masspectra, comparison with mass spectral libraries (NIST 2.0) andith published mass spectra and chromatograms.

Page 4

E. Ribechini et al. / J. Chromatogr. A 1183 (2008) 158–169 161

of th

3

3

srtwpto

fMa[mlbap

T3Cawdtecrsm

bTbaf

Fig. 2. Mass spectra obtained by DE-MS of sub-samples

. Results and discussion

.1. DE-MS

Fig. 2 shows the DE-mass spectra obtained from the analy-is of two of the unguentaria residues, 73291 and 74677. Theesidue from unguentarium 73291 reveals a complex mass spec-rum (Fig. 2a) featuring a high abundance of m/z 253, togetherith m/z 239; these ions are the base peaks in the spectra ofure 7-oxo-dehydroabietic and dehydroabietic acids, respec-ively, thereby suggesting the presence of a diterpenoid resinf Pinaceae origin [13].

A series of six regularly spaced ions separated by 28 amuorms a cluster in the range m/z 592–732 and are attributable to

+ of long-chain fatty acyl esters (cerides or wax esters), whichre present in high abundance (about 35%, w/w) in beeswax12]. The presence of beeswax is reinforced by fragment ions at/z 256 and 257, which derive from the characteristic beeswax

ong-chain palmitate wax esters by fission of the alkyl-oxygenond [12,21]. Further pairs of fragment ions present at m/z 284nd 285, 312 and 313, 340 and 341, and 368 and 369, indicate theresence of long-chain monoesters, other than palmitate esters.

aosm

e contents of glass unguentaria 73291 (a) and 74677 (b).

he aforementioned pairs of ions at m/z 284 and 285, 312 and13, 340 and 341, and 368 and 369, suggest the presence of C18,20, C22 and C24 fatty acyl moieties, respectively. The latterre not present in beeswax long-chain monoester fraction [12],hich comprises exclusively the C16:0 acyl moiety, and were notetected in DE-MS analyses of fresh beeswax analysed duringhis investigation (data not shown). The presence of the lattersters appears to suggest the presence of another waxy-lipidomposed of aliphatic long-chain wax esters with a chain lengthanging between C40 and C54 carbon atoms, coincidentally theame carbon number ranging as beeswax, but whose fatty acyloieties contain between 18 and 22 carbon atoms.Residues 74229, 74678 and 74682 showed a similar

ehaviour under DE-MS to residue 73291, discussed above.he mass spectrometric analyses demonstrate the occurrence ofeeswax, together with pine resin and another wax(es) in vari-ble proportions. The variation in the relative abundances of theragment ions m/z 239 and 253, deriving from dehydroabietic

cid and 7-oxo-dehydroabietic acid, suggests differing degreesf oxidative damage to the residues [13]. The averaged masspectrum of residue 74677 (Fig. 2b) showed several ions in com-on with the previously discussed samples. The predominance

Page 5

162 E. Ribechini et al. / J. Chromatogr. A 1183 (2008) 158–169

Fig. 3. TIC of the trimethylsilylated TLEs of the contents of glass unguentarium 73291 obtained by HTGC–MS. Ax:y are fatty acids of chain length x and degree ofu lengta DA isa

o6lo3arfdms

aecioroiHTop

3

iH7a

Caea

mdenvmpoceo[ciwwsmohtmq

nsaturation y; Cx are n-alkanes of chain length x; OHx are n-alkanols of chaincid; DA is dehydroabietic acid; 7OHDA is 7-hydroxy-dehydroabietic acid; 7Ocid; I.S. is the n-tetratriacontane internal standard.

f ions at m/z 256 and 257 together with those at m/z 592, 620,48, 676, 704 and 732, deriving from wax esters, clearly high-ights the presence of beeswax [12] as a significant componentf this residue. The pairs of ions at m/z 284 and 285, 312 and13, 340 and 341, and 368 and 369, discussed above are present,lthough at lower abundances than seen in the spectra of the otheresidues considered above, thereby providing further evidenceor the presence of an additional waxy-lipid. Ions diagnostic ofiterpenoids were not evident in this residue. The averaged DEass spectra obtained for residues 74676 and 74680 were quite

imilar to that obtained for 74677.On the basis of all the averaged DE mass spectra obtained, it

ppears that beeswax is a major component of all the balmsxamined, along with lower proportions of other lipids andoniferous resin. However, since DE-MS is only suitable fordentifying the major components of such mixtures [14], theccurrence of other important organic components cannot beuled out at this stage. For this reason, and due to the uniquenessf the contents of the unguentaria under study, their chem-cal composition at the molecular level was investigated byTGC–MS and GC–MS, the latter following saponification.he primary aim was to gain a more comprehensive assessmentf the biomolecular composition of the residues and obtain arofile for the ester-bound fatty acids and neutral components.

.2. HTGC–MS

The investigations carried out by HTGC–MS gave qual-

tatively similar results for all seven samples. Fig. 3 showsTGC–MS total ion chromatogram (TIC) obtained for sample3291. A range of even carbon numbered saturated n-alkanoiccids (containing 16–34 carbon atoms) were identified, with the

tp3f

h x; Ex are long-chain monoesters of chain length x; DDA is didehydroabietic7-oxo-dehydroabietic acid; 7OH15ODA is 7-hydroxy-15-oxo-dehydroabietic

24–C30 homologues being the most abundant. Oleic acid waslso present together with long-chain primary alcohols with anven number of carbon atoms in the range 24–34, maximisingt 30.

Moreover, a series of n-alkanes (ranging from C25 and C33,aximising at C27) with an odd-over-even carbon number pre-

ominance were observed, along with odd and even long-chainsters in the C40–C54 carbon number range. The odd carbonumber esters are present at very low abundance and are notery widespread in nature. As exemplificative of long-chainonoesters, the mass spectra observed for the chromatographic

eaks attributed to 50 and 49 carbon numbered long-chain esters,bserved in sample 73291, are reported in Fig. 4. The fragmentslearly indicate that several esters co-elute, in fact, long-chainsters are known to provide one main fragment characteristicf the acid moiety of the ester, namely [CnH2nCO2H + H]+

15]. On the contrary, the mass spectra recorded for theontents from unguentaria, present several peaks and seem tondicate that every chromatographic peak could be associatedith several (from 2 to 6) aliphatic long-chain monoestersith the same molecular weight. For example, in the mass

pectrum of ester with 50 carbon atoms (Fig. 4a), the peaks at/z 257, 285, 313, 341, 369 and 397 indicate the co-elutionf triacontanyl-hexadecanoate, octacosanyl-octadecanoate,exacosanyl-eicosanoate, tetracosanyl-docosanoate, docosanyl-etracosanoate and eicosanyl-hexacosanoate, respectively. The

ass spectrum of each chromatographic peak provides theualitative composition of the isomeric mixture and allows us

o quantitatively evaluate the various isomers present. Fig. 5resents the distribution (relative %) of the ions at m/z 257, 285,13, 341, 269, 297, 425 and 453 in each group of isomeric estersor each unguentaria residues. Interestingly, the composition of

Page 6

E. Ribechini et al. / J. Chromatogr. A 1183 (2008) 158–169 163

F rbon7

tqc

nbblhpmbcpa

ottu

apieeafsoth

ahaa

ig. 4. Mass spectra of long-chain monoesters containing 50 (a) and 49 (b) ca3291 in Fig. 3.

he various isomeric mixtures of each chromatographic peak isuite constant in the seven samples investigated, suggesting aommon origin.

The co-occurrence of a series of long-chain n-alkanols,-alkanes, long-chain n-alkanoic acids and even carbon num-er long-chain palmitate wax esters confirms the presence ofeeswax [15]. However, the presence of odd carbon numberedong-chain wax esters together with even carbon numberedomologues in which the fatty acyl moiety is not exclusivelyalmitic acid strongly suggests the presence of another waxyaterial, seemingly confirming the initial observations made

y DE-MS. Significantly, aliphatic long-chain wax esters withhain lengths ranging between 30 and 56 carbon atoms and com-rising C26 and C28 n-alkanols esterified with C18–C22 fattycids are common constituents of plant epicuticular waxes [22].

However, there may be another explanation for the origin

f the unusual mixtures of the long-chain esters present inhe samples; they are possibly secondary products (transes-erification products) formed during the preparation of thenguents, i.e. during the heating of beeswax with pine resin and

dP

i

atoms observed during the HTGC–MS of the contents of glass unguentarium

second fatty acyl lipid containing material. There are somerecedents for this in archaeological materials, for example,n ceramic vessels from Roman Britain, unusual triterpenoidsters, comprising betulin, lup-2,20(29)-dien-28-ol and lupeolsterified to fatty acids (palmitic and stearic acids), werettributed to the intentional mixing of birch bark tar and animalat [23]. More recently [24] several longiborneyl palmitates andtearates, unknown as natural products, observed as componentsf the organic contents of Egyptian canopic jars were attributedo secondary transesterification products formed during theeating of animal fat with conifer oil.

Moreover, in all the unguentaria residues several diterpenoidcids with abietane skeleton were identified, including: 7-ydroxy-15-oxo-dehydroabietic acid as the major component,ccompanied by lower abundances of 7-oxo-dehydroabieticcid, 15-hydroxy-dehydroabietic acid, dehydroabietic acid, and

idehydroabietic acid, thereby confirming the presence of ainaceae resin [13].

Thus, qualitatively the chemical compositions were very sim-lar for the all residues analysed, although there were some

Page 7

164 E. Ribechini et al. / J. Chromatogr. A 1183 (2008) 158–169

F 269, 2t

qdle[(ocdTatmo7fdhm(

iasa

apCct

3

lfi(wd2hh9i(

ig. 5. The relative abundance distributions of ions of m/z 257, 285, 313, 341,he contents of the glass unguentaria.

uantitative differences. For example, as shown in Fig. 6, theistribution of the various components present in the totalipid extracts determined by HTGC–MS, is somewhat differ-nt for each sample and differs from that of fresh beeswax12,19]. This difference could be for the following reasonsnone of which are mutually exclusive): (i) the occurrencef a second waxy-lipid material, as discussed above and (ii)hemical changes occurring in the composition of beeswaxue to anthropogenic and/or natural factors [19,21,25,26].he presence of free long-chain n-alkanols (Fig. 6a), whichre absent in the fresh beeswax, may be related to the par-ial hydrolysis of characteristic palmitate esters providing a

ixture of free palmitic acid and long-chain alcohols. More-ver, in the samples from unguentaria 74676, 74677, 74678,4680, 74682, the n-alkane profile (maximising at C31) dif-ers from that of fresh beeswax (maximising at C27) due to aisappearance or a depletion of the lowest molecular weightydrocarbons (C23, C25 and C27), and to the consecutive enrich-ent in the highest molecular weight hydrocarbons (C31, C33)

Fig. 6b).In the case of the Oplontis residues, the changes in chem-

cal composition of beeswax could be due to: (a) the heatpplied to mix the constituents of the balms and/or render themufficiently fluid for application to the skin, (b) the high temper-tures reached by the ash that completely covered the Oplontis

sa11

97, 425 and 453 in each group of isomeric long-chain monoesters observed in

rea after the volcanic eruption, and/or (c) natural degradationrocesses undergone by beeswax during 2000 years of burial.learly, all these factors may have worked together to lead to theompositions observed in the unguentaria residues from Oplon-is.

.3. GC–MS

The analyses carried out by means of GC–MS gave simi-ar results for all the seven vessel contents. The TIC obtainedor the acidic fraction of sample 73291 is shown in Fig. 7 ands characterised by the presence of the following compounds:i) linear saturated fatty acids containing 12–34 carbon atoms,ith palmitic, stearic and lignoceric acids as the most abun-ant fatty acyl moieties; (ii) long-chain n-alcohols containing4–34 carbon atoms; (iii) long-chain hydroxyacids, namely 14-ydroxyhexadecanoic acid, 15-hydroxyhexadecanoic acid, 21-ydroxydocosanoic acid and 23-hydroxytetracosanoic acid; (iv),10-dihydroxyoctadecanoic acid (present as threo and erythrosomers); (v) oleic acid, the only unsaturated fatty acid present;vi) �,�-dicarboxylic acids, sebacic acid, azelaic acid and

uberic acid at low abundances; (vii) several tricyclic diterpenoidcids with abietane skeletons, encountered eluting in the range8–28 min, namely 7-hydroxy-dehydroabietic and 7-hydroxy-5-oxo-dehydroabietic acids, followed by 7-oxo-dehydroabietic

Page 8

E. Ribechini et al. / J. Chromatogr. A 1183 (2008) 158–169 165

Fig. 6. The relative abundance distributions of n-alkanols (a), n-alkanes (b) and fatty acids (c) from the TLEs of the contents of the glass unguentaria from Oplontis.

Page 9

166 E. Ribechini et al. / J. Chromatogr. A 1183 (2008) 158–169

Fig. 7. TIC of the trimethylsilylated acid fraction of sample 73291 obtained by GC–MS. Ax:y are fatty acids of chain length x and degree of unsaturation y; diAx

a lengta hydroa ydroa

ae

pbtpn

wohs

Fl

re �,�-dicarboxylic fatty acids of chain length x; OHx are n-alkanols of chaint position X; 9,10OHA18:0 are 9,10-dihydroxyoctadecanoic acids; DDA is didecid; 7ODA is 7-oxo-dehydroabietic acid; 7OH15ODA is 7-hydroxy-15-oxo-dehnd I.S.2 are the n-tridecanoic acid and n-hexadecane internal standards.

cid, 7-hydroxy-15-hydroxy-dehydroabietic acid, dehydroabi-tic acid and didehydroabietic acid.

The neutral fraction of sample 73291 (Fig. 8a) shows theresence of long-chain n-alkanols, with an even carbon num-

ers in the range 24–34. These n-alcohols were also present inhe acidic fraction, as in the adopted condition of our extractionrocedure they were separated in the two fractions (acidic andeutral) [26]. A series of n-alkanes (C25–C33 maximising at C27)

3

wt

ig. 8. TIC (a), and summed mass chromatogram (b) of m/z 396, 394, 382 and 255 of nength x; OHx are long-chain alcohols of chain length x; 1,x-1OHx are �,�-1diols of c

h x; XOHAx are hydroxy fatty acids of chain length x and with hydroxy groupabietic acid; DA is dehydroabietic acid; 7OHDA is 7-hydroxy-dehydroabietic

abietic acid; 7OH15OHDA is 7-hydroxy-15-hydroxy-dehydroabietic acid; I.S.1

ith odd-carbon number was also present, together with a seriesf long-chain alkandiols, namely, 1,23-tetracosandiol, 1,25-exacosandiol and 1,27-octacosandiol. Finally, campesterol,tigmasterol and �-sitosterol were observed eluting between

1 min and 33 min, as better shown in Fig. 8b.

GC–MS analyses revealed in all residues the presence of aide range of classes of organic compounds largely attributable

o the presence of Pinaceae resin and beeswax in all the residues.

eutral fraction of sample 73291 obtained by GC–MS. Cx are n-alkanes of chainhain length x; camp is campesterol; stigm is stigmasterol; �-sito is �-sitosterol.

Page 10

E. Ribechini et al. / J. Chromatogr. A 1183 (2008) 158–169 167

F ntenta

MpaodbTa

atpi

ig. 9. The relative abundance distributions of fatty acids derived from the conalysis.

oreover, in all samples, with the exception of sample 74677,hytosterols were found, namely, campesterol, stigmasterolnd �-sitosterol. Their presence along with the simultaneousccurrence of oleic acid, of �,�-dicarboxylic acids and dihy-

roxyoctadecanoic acids seems to indicate that together witheeswax and coniferous resin, the presence of a plant lipid.his interpretation concurs with the findings of the DE-MSnd HTGC–MS analyses, discussed above. While phytosterols

t(tr

s of glass unguentaria and from beeswax by alkaline hydrolysis and GC–MS

re very common compounds in the plant world, their tendencyo oxidise readily and the fact that they are usually present inlant lipids in low concentrations, means they rarely surviven archaeological organic residues. Interestingly, in the Oplon-

is samples phytosterol contents constitute on average 5–10%w/w) of the lipid present, while in the vegetable oils it theyypically constitute only ca. 1%. The latter evidence seems toeinforce the hypothesis regarding the presence of an addi-

Page 11

1 omat

to

biaanftb

ManebbAactacfahc

4

acttjuracnotatohostoHaoar

bifacpwpo

iigaTiwalfiel

A

SiD

R

[[

68 E. Ribechini et al. / J. Chr

ional plant lipid, most likely a wax, rather than a vegetableil.

The latter hypothesis appears to be confirmed by the distri-utions of fatty acids present in the various samples, as shownn Fig. 9. The histograms clearly show that the patterns of fattycids are inconsistent with the sole presence of beeswax. Oleicnd stearic acids, rather than palmitic acid, are the main compo-ents, and the long-chain saturated fatty acids are too abundantor beeswax. Significantly, of the long-chain fatty acids, n-etracosanoic acid is not the dominant homologue, as it woulde in beeswax [12,15,19,26].

In summary, on the basis of the results obtained by DE-S, HTGC–MS and GC–MS analyses, the occurrence of long-

nd mid-chain fatty acids, long-chain n-alkanols, long-chain-alkanes and, odd- and even-carbon number long-chain waxsters can be explained based on the presence of a mixture ofeeswax and plant waxy-lipid. This interpretation is supportedy the occurrence of unusually high abundances of phytosterols.n alternative explanation is that the observed distribution of

lkyl lipids is in fact altered beeswax, however, for this to be thease we would have to invoke a degradation pathway involvinghe hydrolysis of long-chain monoesters, depletion of n-alkanesnd palmitic acid and the formation of secondary transesterifi-ation products. The resulting chemical profile would be uniqueor archaeological beeswax. The most likely explanation is that

combination of anothropogenic and diagenetic phenomenaas produced the characteristic compositions observed in theontents of all seven unguentaria from Oplontis.

. Conclusions

Presented above are the results of chemical investigationsimed at characterizing the organic components of the originalontents of seven Roman glass unguentaria recovered duringhe excavation of Villa B in Oplontis (Naples, Italy; 1st cen-ury BC to 1st century AD). The contents of such ointmentars have been suggested to be emollients for skin, scentednguents, make-up, religious rituals balms or medicinal prepa-ations. Historical documentary sources point to mixtures of

lipid-base with terpenic resins being used to manufactureosmetics, unguents and ointments in Roman times, but untilow this has never been confirmed by chemical analysis ofrganic materials surviving at archaeological sites. The unguen-aria from Oplontis provided a unique opportunity to undertake

systematic investigation of their enigmatic contents. Due tohe likely complex nature of such materials a combinationf analytical approaches was employed to ensure compre-ensive analysis: DE-MS, HTGC–MS of TLEs, and GC–MSf neutral and acid fractions obtained by alkaline hydroly-is. DE-MS provided chemical ‘fingerprint’ information forhe major components, requiring only low microgram amountsf sample, no sample pre-treatment, and the analysis is fast.TGC–MS and GC–MS afforded compositional information

t the molecular level, indicating the presence of a wide rangef compound classes characteristic of the presence of beeswaxnd Pinaceae resin, together with another waxy-lipid mate-ial.

[[

[

ogr. A 1183 (2008) 158–169

The mixtures of components point to the vessel contentseing residues of the preparation of cosmetics, i.e. moisturis-ng creams or skin balms. Beeswax was most likely the baseor the cosmetics, with pine resin being included for its pleas-nt aroma and possibly antioxidant properties. The other lipidomponents, namely the high abundance of phytosterols and theresence of long-chain monoesters in which the acyl moietiesere not exclusively palmitic acid, appears to be the residue oflant lipid extracts derived from epicuticular waxes of flowersr leaves.

The results are consistent with historical sources document-ng that scented unguents were prepared using a base substance,.e. vegetables oils, animal fats or waxes, with aromatic and fra-rant components being extracted from resin, flowers, spicesnd/or aromatic wood by hot and/or cold maceration [1,17].he cold maceration technique, known today as “enfleurage”,

nvolves smearing a layer of a fatty or waxy substance on aooden board on to which aromatic plant materials were placed

nd the scent components allowed to diffuse into the fatty/waxayer over a period of 1–3 days. The process was repeated withresh plant materials until that the desired degree of fragrances reached. In hot maceration, the aromatic components werextracted by steeping and stirring plant materials in a heatedipid-based extractant.

cknowledgments

The authors would like to thank Prof. P.G. Guzzo of theoprintendenza Archeologica di Pompeii (Naples, Italy) for hav-

ng provided archaeological samples. E.R. would like to thankr. A. Beconcini for his help in preparing Fig. 1.

eferences

[1] A. Lucas, J.R. Harris, Ancient Egyptian Materials and Industries, Dover,Mineola, NY, 1962.

[2] M. Serpico, R. White, in: P. Nicholson, I. Shaw (Eds.), Ancient EgyptianMaterials and Technology, Cambridge University Press, Cambridge, UK,2000, p. 390.

[3] T. Bardinet, Le Papyrus Medicaux de l’Egypte Pharaonique, Fayard, Paris,1995.

[4] H. Rackham, Pliny the Elder: Natural History, Loeb Classical Library,Heinemann, London, 1968.

[5] P. Walter, P. Martinetto, G. Tsoucaris, R. Breniaux, M.A. Lefebvre, G.Richard, J. Talabot, E. Dooryhee, Nature 397 (1999) 483.

[6] M. Cotte, P. Dumas, G. Richard, R. Breniaux, P. Walter, Anal. Chim. Acta553 (2005) 105.

[7] R.P. Evershed, R. Berstan, F. Grew, M.S. Copley, A.J.H. Charmant, E.Barham, H.R. Mottram, G. Brown, Nature 432 (2004) 35.

[8] J. Perez-Arantegui, J.A. Paz-Peralta, E. Ortiz-Palomar, J. Archaeol. Sci. 23(1996) 649.

[9] A.C. Chapman, H.J. Plenderleith, J. Chem. Soc. CXXIX (1926) 2614.10] G. Zolotovich, A. Popov, Soap Perfumery Cosmet. 42 (1969) 425.11] P. Walter, E. Welcomme, P. Hallegot, N.J. Zaluzec, C. Deeb, J. Castaing,

P. Veyssiere, R. Breniaux, J.L. Leveque, G. Tsoucaris, Nano Lett. 6 (2006)2215.

12] M. Regert, C. Rolando, Anal. Chem. 74 (2002) 965.13] M.P. Colombini, F. Modugno, E. Ribechini, J. Mass Spectrom. 40 (2005)

675.14] F. Modugno, E. Ribechini, M.P. Colombini, Rapid Commun. Mass Spec-

trom. 20 (2006) 787.

Page 12

atog

[

[

[

[[

[

[

[[[

957.

E. Ribechini et al. / J. Chrom

15] R.P. Evershed, in: E. Ciliberto, G. Spoto (Eds.), Modern Analytical Meth-ods in Art and Archaeology, Wiley-Interscience, New York, 2000, p. 177.

16] A.M. Pollard, C. Heron, Archaeological Chemistry (RSC Monographs),Royal Society of Chemistry, Cambridge, UK, 1996, p. 239.

17] M. Serpico, R. White, in: P. Nicholson, I. Shaw (Eds.), Ancient EgyptianMaterials and Technology, Cambridge University Press, Cambridge, UK,2000, p. 430.

18] R.P. Evershed, C. Heron, L.J. Goad, Analyst 115 (1990) 1339.19] R.P. Evershed, S.N. Dudd, V.A. Anderson-Stojanovic, E.R. Gebhard, J.

Archaeol. Sci. 30 (2003) 1.20] F. Modugno, E. Ribechini, M.P. Colombini, J. Chromatogr. A 1134 (2006)

298.

[

[

r. A 1183 (2008) 158–169 169

21] C. Heron, N. Nemeck, K.M. Bonfield, D. Dixon, B.S. Ottaway, Naturwis-senschaften 81 (1994) 266.

22] T.J. Walton, Methods Plant Biochem. 4 (1990) 105.23] S.N. Dudd, R.P. Evershed, Tetrahedron Lett. 40 (1999) 359.24] A. Charrie-Duhaut, J. Connan, N. Rouquette, P. Adam, C. Barbotin,

M.F. de Rozieres, A. Tchapla, P. Albrecht, J. Archaeol. Sci. 34 (2007)

25] M. Regert, S. Colinart, L. Degrand, O. Decavallas, Archaeometry 43 (2001)549.

26] M.P. Colombini, G. Giachi, F. Modugno, P. Pallecchi, E. Ribechini,Archaeometry 45 (2003) 649.