Mixing, dipping, and fixing: the experimental drawing techniques of Thomas Gainsborough

The Morgan Library & Museum, New York, owns twenty-five works on paper by the 18th-century English artist Thomas Gainsborough. Scholarly publications over the past 20 years, as well as Gainsborough’s own writings, have highlighted his proclivity toward innovative methods and experimentation. In particular, a letter that the artist wrote in 1773 reveals details of his secret recipe for making oils on paper, such as his recommended use of lead white and the unorthodox practice of dipping his works in skim milk, possibly to prevent the pigments from discoloring. About a dozen of Gainsborough’s creations were included in a 2018 exhibition at The Morgan entitled Thomas Gainsborough: Experiments in Drawing. On this occasion, an in-depth scientific study aimed to explore the artist’s work as a draftsman, with a special focus on his mastery of materials, his technical innovations, and his development of an original approach to drawing. Initially, a selection of artworks was examined using magnification along with transmitted and raking light to improve surface visualization and to investigate the structure of each piece. Further photographic documentation with ultraviolet and infrared light was performed to gather preliminary information on the variety of white pigments employed, on the wet and dry chalk techniques used in certain works, as well as on the possible presence of coatings and underdrawings. Subsequently, scientific analysis by means of X-ray fluorescence (XRF) and Raman spectroscopies aimed at characterizing the white pigments present in The Morgan’s drawings, which mostly consisted of calcite and lead white. Moreover, a combination of advanced micro-sampling tools, i.e. polyvinyl chloride (PVC)-free erasers and fine polishing films, ad-hoc sample preparation methods, highly sensitive proteomics analysis via nano-liquid chromatography/mass spectrometry (nano-LC/MS), and sophisticated bioinformatics data processing was employed to assess Gainsborough’s use of skim milk as a “secret fixative” on some of his works. Results have revealed the presence of specifically bovine milk in all of the samples evaluated to date. Notably, only through the combined use of such advanced technical resources can the interrogation of all milk proteins retrieved from the samples provide evidence for the presence of a milk fixative and open the discussion about milk processing methods in the 18th century. In addition to granting conservators and art historians a deeper understanding of the complexity of Gainsborough’s drawing techniques, this study paves the way for further investigations to probe the use of casein-based fixatives by other artists working on paper such as Degas and Van Gogh.


Introduction
Celebrated in his lifetime as an outstanding artist and renowned painter of portrait and rural landscapes, Thomas Gainsborough (1727-1788) is a central figure in the history of British art and, more generally, of 18th-century art. Although most famous for his portraits and romantic landscape paintings, Gainsborough was also a prolific draftsman, with over 900 of his drawings presently known. As most artists of his generation, Gainsborough did not draw directly from nature; instead, he composed landscapes in his studio, relying on his sketches but also laying out stones, branches, leaves, and soil of various colors on his work table to recreate an ideal scene. This practice allowed him to experiment with materials and create dramatic scenes using multiple media on paper, such as graphite, chalk, charcoal, oil, watercolor, and pastel.
The Morgan Library & Museum, New York, owns twenty-five works on paper by Gainsborough, including highly completed landscape studies, an unfinished oil sketch, and a rare preparatory study for a painting (Fig. 1). About a dozen of these artworks were included in an exhibition entitled Thomas Gainsborough: Experiments in Drawing, which was held from May 11th to August 19th, 2018. A small catalogue that contains updated descriptions of The Morgan's works, as well as a technical study of Gainsborough's materials and techniques as observed in The Morgan's collection, accompanied the exhibition [1].
Conservation and art historical publications over the past 20 years have clearly demonstrated that Gainsborough was an innovative and experimental artist, having developed some unique drawing and painting methods during the decades of his career. By disrupting the traditional canons of drawing, he created real "experiments on paper" in which the landscape is key to creating original visual effects, each time different, to refine new techniques, and to investigate a new approach to art on paper. A crucial source of information about Gainsborough's singular working process is a letter, held at the Yale Center for British Art, which the artist addressed to his friend William Jackson in January 1773. In this letter, Gainsborough shared a brief account of his complex, extremely secret recipe for making oils on paper, using unorthodox technical solutions that subverted the academic principles of contemporary practice. The significance of this document for The Morgan drawings lies in the artist's documented use of certain white pigments and his process of intricate layering of various media and fixatives. According to Gainsborough's recipe, this type of drawing is created using lead white ("your lights of Bristol-made white lead which you buy in lumps at any house painters"). The paper sheet is then allegedly immersed multiple times in milk ("dip it all over in skim'd milk"), possibly to protect the pigments from discoloring, and adhered while wet to a temporary wooden frame or stretcher lined with paper. After the addition of supplementary light and dark layers as necessary, the drawing is dipped again in milk. Colors are then added, and gum Arabic glazing reported to be applied ("float it all over with Gum water, 3 ounces of Gum Arabic to a pint of water"). Finally, the piece is varnished on both sides to keep it flat ("let that dry & varnish it 3 times with Spirit Varnish"), after which the paper is removed from the strainer and mounted. Gainsborough's description of his secret recipe for making these works ends with a theatrical sign off: "Swear now never to impart my secret to anyone living".
The Morgan's 2018 exhibition provided an impetus to undertake an in-depth technical study of Gainsborough's works for inclusion in the associated catalogue. Initially, a selection of his drawings housed at The Morgan Library were inspected using magnification along with different types of illumination, including transmitted and raking light, in order to better visualize the artworks' surface and appreciate the structure of each piece. Further photographic documentation and examination with ultraviolet (UV) and infrared (IR) light were carried out in an attempt to gain insight into Gainsborough's working methods, with a special focus on the possible application of coatings and the presence of underdrawings. In a second phase of this project, scientific analysis with both non-invasive and micro-invasive instrumental techniques aimed at characterizing the different white pigments found on the artist's works, as well as his potential use of fixatives. Identification of the white pigments was performed non-invasively by means of X-ray fluorescence (XRF) spectroscopy and, upon careful removal of a few microscopic samples, with a benchtop Raman spectrometer. On the other hand, the detection and characterization of milk-based fixatives on Gainsborough's drawings was accomplished using minimally invasive sampling techniques, such as polyvinyl chloride (PVC)-free erasers and fine polishing films, combined with proteomics analysis via nano-liquid chromatography-electrospray ionization-mass spectrometry/mass spectrometry (nano-LC-ESI-MS/MS, henceforth referred to as nano-LC/MS).
Typically, the very thin layers of nearly invisible coatings or fixatives that might be present on a drawing cannot be identified using traditional analytical approaches such as Fourier-transform infrared (FTIR) spectroscopy, because a clean sample would be extremely difficult to isolate and the residues of paper substrate that are carried along would likely mask any signal from the target analytes. Similarly, protein analysis via gas chromatography/mass spectrometry (GC/MS) poses a great challenge, as the extensive sampling required in order to gather a sufficient amount of material would not be considered safe for the artwork. The present study, on the other hand, demonstrates that the combined use of minimally destructive micro-sampling methods [2], allowing safe sampling of the drawings, with trace-compatible sample preparation protocols, highly sensitive analytical techniques, and sophisticated bioinformatics data processing enables to achieve high levels of sensitivity and accuracy in the collection and interpretation of results [3]. Only thanks to the implementation of such multi-step methodology can the interrogation of all milk proteins retrieved from the samples provide evidence for the presence of a milk fixative and open the discussion about milk processing methods in the 18th century. In addition to granting conservators and art historians a deeper understanding of the complexity of Gainsborough's drawing techniques, this work paves the way for further investigations aiming to probe the use of casein-based fixatives by other artists working on paper such as Degas and Van Gogh.

Experimental
XRF Analysis was performed using a handheld Bruker Tracer III-V TM energy dispersive XRF analyzer, with Peltier-cooled advanced high-resolution silver-free Si-PIN detector with a 0.2-μm beryllium (Be) window and average resolution of approximately 142 eV for the full width at half maximum of the manganese (Mn) Kα line. The system is equipped with changeable filters, and a rhodium (Rh) transmission target with maximum voltage of 45 kV and tunable beam current of 2-30 µA. The size of the spot analyzed is approximately 3 × 4 mm. Analysis was performed using 40 kV, 30 µA, 180-s acquisition time, and a titanium (Ti)-aluminum (Al) filter, by positioning the instrument at a ≈ 1-mm distance from the artwork's surface.
Raman Analysis was conducted using a Bruker Senterra Raman spectrometer equipped with Olympus 20× and 50× long working distance microscope objectives and a charge-coupled device (CCD) detector. A continuous wave diode laser, emitting light at 785 nm, was used as the excitation source, and two holographic gratings (1800 and 1200 rulings/mm) provided a spectral resolution of 3-5 cm −1 . The output laser power was kept between 10 and 25 mW, while the number of scans and integration time were adjusted to prevent damage from overheating and according to the Raman response of the samples examined. Spectra were interpreted by comparison with published literature and library databases available at The Met.
Visual examination Visual examination of the artworks was performed using a Nikon SMZ-1270 stereo zoom microscope with 6.3-80× zoom range, equipped with stops and 0.5× lens. Examination under UV light was carried out by means of light model Q-22B, with peak intensity at 6"-1200 μW/cm 2 . A Fuji IS Pro Digital camera with Forensic Zeiss lens kit, Peca #916 and Wrattan 2E filters, was used to collect UVA-induced visible fluorescence photographs. Illumination was provided by CFL BLB bulbs working at 13 W and 120 V. IR photography was performed using a Pixelteq Spectrocam SWIR camera system including RAPTOR 320 × 256, a filter wheel, and a Coastal Optic 60-mm lens; all IR images were captured using a 1332-nm filter.
Sampling and sample preparation for nano-LC/ MS Sampling of the Gainsborough drawings was performed following a protocol described by Kirby and coauthors, in which the use of both PVC-free erasers and fine polishing films was first demonstrated on photographs [2]. In detail, PVC-free erasers (Staedtler, 525 B) were cut into small blocks of approximately 3-5 mm and held with tweezers for rubbing under the microscope. Small discs of 5-6 mm in diameter were cut from fiber optic polishing film discs with aluminum oxide (Precision Fiber Products) of 6-and 15-μm grits. These discs were adhered to polystyrene rods of varying diameter (Walther) and held in a nickel Starette Pin Vise (Grainger) for rubbing under the microscope (Fig. 2). Rubbing duration and pressure varied for each sample until the deposition of material could be observed on the eraser or fine polishing film (appearing as surface grime or as a grayish/whitish residue, respectively). Microscopic examination of the works' surface after sampling did not show any evidence of surface disruption or media loss. Based on a treatment method previously developed by some of the authors of this article [3], including the adoption of filter-aided sample preparation [4], erasers and fine polishing films were handled as follows prior to nano-LC/MS analysis: upon sampling, these proteintrapping supports were placed in a lysis buffer containing 4% sodium dodecyl sulfate, 8 M urea, 0.2% deoxycholic acid, and 50 mM dithiothreitol. After washing, alkylation was induced by adding 55 mM iodoacetamide. Buffer was then exchanged to 50 mM ammonium bicarbonate and 0.2% deoxycholic acid (pH 8.8) for digestion with trypsin and lysC enzymes. Liquid-liquid extraction was performed using ethyl acetate, while trifluoroacetic acid was employed for acidification. Several washing steps were carried out before and after phase transfer. The aqueous phase was then entirely evaporated by means of a speed vacuum system, after which a minimal volume of 0.1% formic acid was added. A list of the artworks' locations from which samples were retrieved is provided in Table 1.
Nano-LC/MS The peptide mixture was analyzed by means of an UltiMate 3000 nano-LC system coupled with an electrospray orbitrap Fusion TM Lumos TM Tribrid TM mass spectrometer (Thermo Fisher Scientific, San Jose, California), using (A) 0.1% formic acid and (B) 0.1% formic acid in 80% acetonitrile. 1 μL of peptide digest was initially loaded onto a 5-mm C18 PepMap TM trap column with 300-µm inner diameter (Thermo Fisher Scientific) at a flow rate of 10 µL/min with 0.1% formic acid. Peptides were then eluted from the trap column onto an analytical 50-cm C18 PepMap TM column with 75-mm internal diameter (Thermo Fisher Scientific) with a 4-40% linear gradient of solvent B in 235 min. The elution flow rate for the latter step was set at 300 nL/min. The mass spectrometer operated in positive ion mode at a 1.9-kV needle voltage. Data were acquired using Xcalibur 4.1 software in a data-dependent mode. MS scans were recorded in the m/z 375-1500 range at a resolution of R = 120,000 (at m/z 200), while MS/MS scans were collected in the orbitrap at a resolution of R = 30,000 (at m/z 200). Ions between +2 and +7 charge were selected for high-energy collisional dissociation (HCD) fragmentation. Reference samples, blanks, as well as positive and negative controls were injected and tested alongside the historic samples to prevent contamination and carryover.
MS data processing Raw data files generated from nano-LC/MS spectral acquisition were searched using PEAKS 8.5 software against UniProtKB/Swiss-Prot 2020_02 database and, then, against a restricted Uni-Prot Bos taurus database (Proteome ID UP000009136). Parameters for data analysis were set as follows: 3 missed cleavages maximum; precursor ion error tolerance 10.0 ppm; fragment ion error tolerance 0.02 Da. Carbamidomethylation was set as a fixed modification, while oxidation (methionine, proline), deamidation (asparagine, glutamine), phosphorylation (serine, threonine), and lactosylation (lysine, arginine) were set as variable modifications. Protein false discovery rate (FDR) was set to 0.1% with -10lgP ≥ 18 score threshold for peptides. A manual filter of at least two different non-overlapping peptides was set for protein identification. Data resulting from the injection of positive and negative controls, as well as blanks, were carefully examined prior to injecting the historic samples to verify the possible presence of contaminants or residues from carryover. Reference samples were also used to verify retention times, peak shapes, peptide signal to noise ratio, the occurrence of contaminants resistant to washing, as well as the overall validity of the procedure adopted. For the historic samples, MS/ MS spectra of each peptide of interest were inspected manually, and the presence of various peptide species was investigated by searching unknown spectra against the whole nrNCBI database via Basic Local Alignment Search Tool (BLAST).

Results
A summary of the results obtained from visual examination, as well as XRF and Raman analysis of a selection of Gainsborough drawings, is provided in Table 1, along with an indication of the sampling locations for caseinbased fixative identification with nano-LC/MS.

Results of XRF and Raman spectroscopy
Identification of the white pigments in a selection of Gainsborough drawings from The Morgan Library holdings was carried out using a combination of in situ, non-invasive XRF analysis and micro-invasive Raman measurements on microscopic samples. Results have shown that, in addition to a prominent use of lead white as described in Gainsborough's 1773 letter, calcite is also present in white areas and highlights of several works among those examined. The data collected strongly suggest that, in all drawings selected for analysis, these two pigments were used individually. In addition to lead and calcium, indicative of the use of lead white or calcite pigments, XRF spectra display a series of trace elements that are attributable to the paper substrate. In detail, as also summarized in Table 1, lead white was detected in five

Results of visual examination
All the whites, both lead-and calcium-based, appear bright upon visual inspection with normal light. Close examination with transmitted and raking light, as well as UV photography, indicated that overall and selectively applied coatings or fixatives might be present on many of the drawings under study. , may suggest the presence of a protective coating. These observations led to formulate a theory according to which Gainsborough may have used his "secret ingredient", skim milk, on many of his drawings, as a fixative or working medium to manipulate chalk, graphite, and opaque white pigments more easily [5].

Results of proteomics analysis
Verification of reported fixatives, i.e. very thin coatings applied to protect or "fix" friable or sensitive artists' media, is relevant to Gainsborough as well as to many other artists working on paper. While Gainsborough's innovative methods for making drawings have been of great interest to scholars, obtaining a sufficient amount of sample from the many thin and delicate layers of such fixatives found on his artworks has posed a significant hurdle when using traditional analytical approaches. Indeed, in most cases, well-established instrumental techniques such as FTIR, GC/MS, and pyrolysis-GC/MS are not adequately sensitive for this type of investigations unless a sample of inappropriately large size for cultural heritage artifacts is removed. In the present study, high levels of sensitivity and accuracy in the collection and interpretation of results could only be achieved thanks to the combined use of minimally destructive micro-sampling tools (PVC-free erasers and fine polishing films), advanced sample preparation methods, and highly sensitive analytical techniques (nano-LC/MS), in association with sophisticated bioinformatics data processing. In the literature, the use of erasers for mass spectrometry analysis was first successfully demonstrated in the removal of specimens from parchment [6]. Kirby and coworkers further expanded the use of erasers for sampling of cultural heritage and introduced polishing films to sample coatings on photographs and works of art on paper [2]. Proteomic analysis by nano-LC/MS is the only way to manage extremely low sample amounts and, at the same time, address the likelihood of complex sample mixtures. As stated previously, Gainsborough reportedly used a wide range of media, including oil paint, chalks, watercolor, and possible gum Arabic isolating layers, as well as various fixatives and varnishes. As a result, every sample removed from any of his drawings may potentially contain more than one medium, and the amount of material relevant to protein analysis is only a fraction of the total sample collected. Therefore, the use of special sample preparation methodologies tailored to the treatment of the minute quantities of material gathered upon sampling by PVC-free erasers or fine polishing films represents a first, necessary step to isolate the protein fraction contained in the sample and ensure successful analysis. The subsequent application of a proteomic approach, i.e. the digestion of such proteins into peptides, enables all of the proteins to be detected and identified as each peptide sequence can be assigned to a protein by means of bioinformatics softwares.
Methodologically, proteomics, and in particular the socalled "bottom-up" strategy used in this study, is based on four main steps, as follows: (i) Protein extraction from the sample matrix containing organic and/or inorganic components; here, this step was conducted using an adapted filter-aided procedure, as described in the Experimental section. (ii) Controlled hydrolysis of proteins into peptides using one or more enzymes; in the present work, the digestion was performed using trypsin and lysC, i.e. endoproteases that cleave proteins at the carboxyl side of lysine amino acid for both enzymes and of arginine amino acid for trypsin. (iii) Analysis of the resulting peptide mixture by means of an analytical workflow adapted to very low sample amounts; here, through the coupling of nano-LC and high-resolution tandem MS, peptides are purified, separated, amino acid sequences are obtained by the elucidation of fragment ions resulting from the peptide fragmentation spectra and, then, coded to represent their origin (N-terminal charged fragment ions are classed as a, b or c, while the C-terminal charged ones as x, y or z) and length (for example, a 10 amino acid peptide could create fragments y 1 + through y + 10 with the corresponding b ions) [7]. (iv) MS data processing by means of bioinformatics tools that integrate genomic or protein databases, allowing for the identification of proteins and their modifications.
In the present case, despite the extremely low sample amounts investigated, remarkably high protein sequence coverages were achieved, leading to the unequivocal identification of different milk proteins in all of the samples analyzed, as shown in Table 2. As a representative example, the results obtained for sample S2C2 from Coastal Scene with Shipping, Figures, and Cows (2017.89) are presented in the following. For this specimen, proteomics analysis revealed high sequence coverage (over 50%) for alpha S1 casein. Spectra display good signal to noise ratios, as well as a well-defined fragmentation pattern characterized by an almost complete y and b ions series from the protein peptides, ensuring unambiguous sequence assignments. The data shown in Fig. 4, for instance, enables identification of the HQGLPqEVL-NENLLR sequence corresponding to the 23-37 peptide of alpha S1 casein. In this spectrum, y 1 + to y + 10 and b 2 + to b + 11 fragment ions are detected; Q28, i.e. glutamine at position 28 of the sequence, referenced in lower case, is deamidated; and variants of deamidation were also identified on positions Q24, N32, and N34, consistently with the fact that this is a common modification in old proteins.
Along with alpha S1 casein, ten milk proteins belonging to three main types have been successfully characterized in sample S2C2 [8,9]: (i) caseins, (ii) whey proteins, and (iii) milk fat globule membrane proteins (MFGM), representing about 78%, 17%, and 1-2% of the total bovine milk proteins present, respectively. Belonging to the first group, the alpha S1, alpha S2, beta, and kappa caseins are low molecular weight proteins of 20-25 kDa and are found in cow's milk in 38%, 10%, 35%, and 12% respective proportions [10]. Several of these proteins exhibit variants and they are heavily post-translationally modified in protein biosynthesis with varying level of phosphorylation and glycosylation [11]. As displayed in Table 2, these four caseins were successfully identified in sample S2C2 from drawing 2017.89 with high sequence coverages. Several of their post-translational modifications such as phosphorylation were also detected and localized, while no lactosylation, a non-enzymatic modification indicating the potential heating treatment of milk, was identified with the current experimental design [12][13][14]. As an example of these results, Fig. 5 illustrates the 118-134 sequence of alpha S1 casein, exhibiting a phosphorylation on the serine at position S130 in both reference and historic samples, whereas only a peptide from a reference heated milk sample shows lactosylated K128 and K120 lysine residues; among these graphs, spectrum C corresponds to the 121-134 sequence, indicating a trypsin cleavage of the K120 lysine from sequence 118-134.
The second type of milk proteins found in sample S2C2 from drawing 2017.89 is whey proteins. Among them, several of the most abundant ones, such as betalactoglobulin (LGB), bovine serum albumin (BSA), and immunoglobulins (IgG), representing approximately 60%, 10%, and 10% of total whey proteins, respectively [10], were identified with sequence coverages over 50% for LGB and between 20 and 50% for BSA and IgG. In addition to those, other whey proteins such as serotransferrin and lactoferrin/lactotransferrin were also found in the matching data. Remarkably, one of the typically most abundant whey proteins, namely alpha-lactalbumin, accounting for 20% of total whey proteins, was not detected in the historic samples analyzed. This protein, however, was identified in all the liquid and dry reference samples of modern milk examined with high sequence coverage (over 80%); analysis of reference milk-dipped slides and papers upon micro-sampling with PVC-free erasers and polishing films also yielded positive identification, albeit with lower sequence coverages (between 30 and 40%) compared to what was expected based on the scores obtained for other milk proteins identified. Alphalactalbumin is a metal-binding protein and its ability to interact with other organic compounds such as lipids has been the topic of extensive accounts in the literature [15]. At this stage, the reason for the failed or poor detection of this protein in the Gainsborough samples, compared to other proteins that appear to have similar physicochemical properties or action mechanisms, remains unclear [16,17]. Ongoing research by the authors of this article is currently investigating drawing replicas in an attempt to shed light on the possible mechanisms involved.
The last type of protein identified in sample S2C2 from drawing 2017.89 is represented by MFGM. The structure of MFGM consists of lipid (40%), protein (60%), and cholesterol [18,19]. More accurately, a phospholipid monolayer surrounds the triacylglycerol core, followed by a proteinaceous coat connecting the monolayer to the outer phospholipid bilayer. Glycoproteins such as butyrophilin and lactadherin (milk fat globule-EGF factor 8) are distributed over the external membrane surface. Lactadherin, a 47-kDa protein, was identified in the sample discussed with sequence coverage below 20%. Despite such low score value, in this case, a particularly good spectral quality enabled the unambiguous attribution of five characteristic peptides. Xanthine dehydrogenase/oxidase, a 146-kDa high molecular weight protein located between the outer glycerolphospholipd bilayer and the inner glycerolphospholipid monolayer that surrounds the triacylgrycerol core, was also identified with very low sequence coverage (below 5%) in another sample, namely S2B1, from the same drawing.
The presence of MFGM proteins in all of the historic samples analyzed is difficult to interpret in relation to Gainsborough's 1773 letter, in which the artist describes his own practice of dipping his drawings specifically in skim milk. In the mid-17th into the 18th century, English dairy and cheese production was distinctly artisanal in character, with very low levels of mechanization. Farmers in the most favored dairying regions, which included the Plain, Somerset, and North Wiltshire, began to commercialize and produce a milk surplus, which, in combination with the advances in the English transportation system, caused farmers to seek non-local markets. London was one of the largest beneficiaries of these developments. Farmers also took advantage of the combination of milk and cheese production. For example, the county of Suffolk had a reputation for high-yielding dairy herds and, since the 16th century, much of the capital's cheese and butter was supplied via a relatively short coastal passage from ports such as Yarmouth, Woodbridge, and Ipswich. However, since cream from the milk was required for making cheese, and farm cheese dairies were researching ways to utilize the remaining heavily skimmed milk, they employed it to produce an inferior "flet" (i.e. skim) milk cheese [20].
Due to industrialization and process improvement, modern skim milk is much more efficiently defatted than milk in 18th-century England and, consequently, what was considered to be skim milk three centuries ago may still have contained a certain amount of MFGM [21]. On the other hand, milk treatment such as cooling, heating, and centrifugation can damage or alter MFGM, causing the release of some membrane-associated proteins from MFGM into the aqueous phase (i.e. skim milk) [22]. In addition to the factors described above, the size of MFGM also induces differences in protein composition [23][24][25][26]. Hence, the designation of "skim" milk in the context of 18th-century dairy production is very complex. Further studies are currently underway, including replicating 18th-century milk defatting, to gain knowledge of the milk protein composition that may most closely correspond to 18th-century skim milk.
Most of the peptide sequences detected in the present study are common to several animal species; however, some are indicative of the biological origin of the proteinaceous material under investigation and, therefore, may be used to draw conclusions in that regard. For example, in the case of sample S2C2 from drawing 2017.89, both the 23-37 peptide of the alpha S1 casein displayed in Fig. 4 and the 19-37 peptide with a missed cleavage listed in Table 3 belong to the Bovinae subfamily (Bos taurus, Bos mutus, Bos indicus, Bos grunniens, Bison bison bison). Other discriminant peptides from the most abundant proteins identified in this sample, namely alpha S1, alpha S2, and kappa caseins, as well as beta-lactoglobulin, are also shown in Table 3. Based on the documentary information available on Gainsborough's working Table 2 List of the main milk proteins identified in the Gainsborough drawings for selected locations. Dark gray, medium gray, and light gray colors represent protein sequence coverages respectively over 50%, between 20 and 50%, and below 20%; asterisks indicate sequence coverages below 5%; blank rectangles mean lack of identification  III, 61  III, 62  III, 63a  III, 63  Samples B S1C2 S2B1 S2C2 S1B2 S1C2 S2B2 S2C2 S1C2 S3B2 S1C S4B S3B2 S2C2 S1A

Discussion
The various coating materials described by Gainsborough in his 1773 letter to Jackson include skim milk, gum Arabic, and varnish, all of which, however, are mentioned only in relation to his multi-layered oils on paper. The presence of yellowed coatings, both overall and selectively applied, on some of The Morgan's drawings clearly suggests that the artist likely used different procedures and materials to fix or protect his various creations.
Fixatives are materials used to enhance the attachment of friable pigments onto a surface, and have been in use for centuries. They may be applied by brush, by spraying with an atomizer, or by dipping. In the 18th century, discussions about fixing pastels, the most The data shows a triply charged ion at m/z 910.0919 (error Δm 0.9 ppm), enabling identification of the 118-134 peptide of alpha S1 casein, which includes two lactosylations (lower case K118 and K120) and one phosphorylation (S130). b, c MS/MS spectra of peptides identified in sample S2C2 from Coastal Scene with Shipping, Figures, and Cows (ca. 1780;2017.89). The data shows (B) a triply charged ion at m/z 694.0209 (error Δm 0.5 ppm), enabling identification of the 118-134 peptide of alpha S1 casein, which includes one phosphorylation (S130); and (C) a doubly charged ion at m/z 831.3934 (error Δm 0.8 ppm), enabling identification of the 121-134 peptide of alpha S1 casein, which includes one phosphorylation (S130) and one deamidation (Q123) friable of media, were taking place. Research in this area conducted by Shelley and other authors indicates that most fixatives employed consisted of natural resins or some form of animal glue [27][28][29]. These materials, however, would significantly alter the color, saturation, and texture of the pastels, and, because of this reason, they were not well liked by professional artists. Gainsborough's desire for an invisible fixative is understandable in light of the fact that he worked in various wet and dry media and was reportedly fearless in his experimentation.
Since the Renaissance, artists have known that lead white would darken under certain conditions and, by the 18th century, using this pigment in watercolors was considered a risky choice, as sulfur in the air would prompt the conversion of the white basic lead carbonate into gray lead sulfide. Writing on the need for careful preparation of lead white, Edward Norgate, English miniature painter and court musician to James I, states: "You cannot be too carefull of this, the neglect whereof hath bene the spoile of many in the Vatican Library at Rome, as alsoe in the curious Villa of Cardinall Burghese and elsewhere. For these being heightened with white, are in the heigthninge become soe black, rusty and dis-coloured as I told you of before" [30]. An unpublished technical study by Sophie Crombie, an undergraduate student at the Courtauld Institute of Art, in the 1980s attempted to re-create Gainsborough's oils on paper, as discussed in his 1773 letter. Interestingly, the author noted that varnishing lead white chalk without a fixative does not work, because the pigment is simply brushed away, indicating that it was poorly bound to the paper support and required being worked with a substance that would improve adhesion. According to Crombie, dipping in milk does appear to act as a fixative, adhering the chalk securely to the paper and allowing for a varnish (or other coatings) to be subsequently applied [5].
The general protein hydrophilicity promotes a good compatibility to polar surfaces, such as paper [31], and an effective barrier to apolar gases, such as oxygen and carbon dioxide. In light of these observations, coating industries are manipulating proteins' physicochemical properties in order to modulate selectively their barrier, mechanical, and surface properties for use in coatings and packaging [32]. A parallel is possible with proteinbased coatings on artworks, as reported in a work exploring the protective properties of egg white as a coating for paintings [33], and, as in the present case of study, with milk proteins used as a protective coating for works of art on paper.
Observations of discolored coatings and subtle surface phenomena on many of The Morgan's drawings, as well as the artist's detailed letter of 1773, strongly suggested that a milk fixative might be present on Gainsborough's works on paper containing lead white. As explained above, results of non-invasive analysis and examination of microscopic samples revealed that both calcite and lead white were used in many of his artworks, which led to further questions about the possible presence of a fixative or coating that might protect other pigments, in addition to lead white, from darkening. All of the lead white-containing areas analyzed appeared to be in excellent condition, with no apparent cracking, flaking or darkening, although these phenomena have been observed in Gainsborough drawings housed in other collections [5,34].
Micro-sampling in association with qualitative proteomics provided important new information. As seen in Tables 1, 2, and 3, all of the drawings tested by this method, regardless of the sampling location (front or back, background or painted area), the date of the artworks, or the white pigment used in them, were found to contain casein and other milk proteins specific to Bos taurus. This result was unanticipated, since it was expected that only the works with lead white might show milk proteins. These data would strongly suggest the following: (i) Gainsborough applied a milk fixative on his drawings whether the white pigment employed was calcite or lead white, or the medium was graphite, chalk, watercolor, or oil paint; (ii) in all cases examined, the milk fixative was applied to both sides of the paper support; (iii) Gainsborough could have been using milk fixatives as early as 1750. The discovery that milk was present on chalk, graphite, and wash drawings prompts further indepth research to investigate all drawing types and styles in the artist's oeuvre not only to understand the extent of his use of milk fixatives, but also to evaluate the possible presence of other materials such as gums and natural resins. The present study indicates that dipping in milk was part of Gainsborough's process of making drawings over a long period of time. Furthermore, based on the data collected thus far, this practice does not appear to relate only to drawings with multiple layers (i.e. III, 61; III, 62; III, 63a), but even to purely graphite works (2005.82) and simple watercolors made with lead white wash (2014.32).
The question of what is yellowing in these works remains unclear and must be further studied. Because to date only the analysis of proteins was undertaken, future research must also consider whether polysaccharides or natural resins may be present, and if unexpected interactions with casein-based coatings leading to yellowing may occur. While Gainsborough noted that he used specifically skim milk, it is likely that at least some milk fat remained in his 18th-century processed milk, and that the amounts might vary in each fixative bath used. The interaction of the fat with the milk proteins and with Table 3 List of peptides that are characteristic of the Bovinae subfamily, with indication of their position in the corresponding proteins, mass to charge ratio (m/z), charge (z), measurement error (Δm), identification score (reported as -10lgP, i.e. converted P value; higher numbers indicate higher-confidence matches), retention time (RT), and animal species that contain that specific peptide. Deamidation variants of these peptides are not referenced in the table. Lower case letters in peptide sequences indicate the following modifications: n asparagine deamidation, m methionine oxidation, s serine phosphorylation, c cysteine carbamidomethylation (induced by sample treatment), w tryptophan oxidation, q glutamine deamidation or N-terminal glutamate conversion to pyroglutamate other possible coatings also prompts for further studies. Scientific investigation, beyond The Morgan's collection, of potential milk fixatives from a larger group of paper artworks by both Gainsborough and his contemporaries would solidify the observations gathered in the present research and enable a deeper understanding of the working methods of the time. In addition to this, it would be crucial to investigate whether milk actually does protect lead white from pollution and conversion to lead sulfide, or if another coating was present that preserved the lead white films in The Morgan's drawings.

Conclusions
The present article describes the first, comprehensive scientific study of a selection of drawings by 18th-century English artist Thomas Gainsborough in the collection of The Morgan Library & Museum, New York. Visual examination of the artworks combined with in-depth analysis by means of a variety of instrumental techniques have been carried out in order to explore the artist's mastery of materials and innovative practices for making drawings, as described in a letter that he wrote to his friend William Jackson in January 1773. Among the main topics investigated are Gainsborough's choice of white pigments and, most interestingly, his reported use of milk-based fixatives to protect the media and prevent pigments from discoloring. Results of this study have shown that, in addition to lead white, calcite was also employed in several works created during the last decade of Gainsborough's artistic production, which may attest to an intrinsic evolution in the artist's choice of materials over the course of his career. Moreover, the artist appears to have applied milk from domestic cows as casein-based fixative on his drawings from 1750-1785, regardless of the media and white pigments used. The data collected indicate that visual examination does not correlate to the actual presence of fixative, which can only be unambiguously confirmed through scientific analysis. Gainsborough's documented practice of dipping his artworks in a fixative bath was initially supported by normal light and UV photography of one of the drawings examined, namely Landscape with Horse and Cart Descending a Hill (ca. 1780; III, 63). In this work, a clean corner at the upper left, very distinct from the overall deteriorated coating, possibly corresponds to the location where the work might have been held by the artist when dipping it in milk. The consistent qualitative identification of the milk proteins sampled on all the drawing surfaces would appear to indicate a similar deposition across both the front and reverse of the drawings. This likely supports an overall dipping of the paper rather than a localized application of a fixative by brush or random non-specific transfer from storage or handling.
In conclusion, the present study clearly demonstrates the necessity of a multifaceted approach combining art historical research, examination of the available documentary sources, careful inspection of the artworks, as well as advanced micro-sampling tools, ad-hoc sample preparation methods, highly sensitive analytical techniques, and sophisticated bioinformatics data processing to shed light on artists and their practices. In addition to granting conservators and art historians a deeper understanding of the complexity of Gainsborough's materials and innovative drawing techniques, this work paves the way for further investigations aiming to probe the use of milk-based fixatives by other artists working on paper such as Degas and Van Gogh. Additional work will be also required to assess more accurately the milk fixative identified on Gainsborough's drawings as representative of 18th-century skim milk. In the near future, the cutting-edge methodologies and techniques combined here may similarly provide useful insight into different classes of materials, such as paper sizing as well as gum Arabic coatings, among others.