LETTERS TO THE EDITOR

1 ULTRAVIOLET FLUORESCENCE MICROSCOPY

1.1 TO THE EDITOR:

After having just read the article by John M. Messinger II, “Ultraviolet-Fluorescence Microscopy of Paint Cross Sections: Cyclopheptaamylose-Dansyl Chloride Complex as a Protein-Selective Stain,” JAIC 31(1992):267–74, I feel compelled to comment on two specific aspects of it, and, if space permits, to make a more general observation on the presentation and review of scientific papers in the JAIC.

One major point of the paper is a comparative study of the relative efficacy of cross sectional staining methods currently in use in conservation. In particular, the performance of two reactive fluorescent dyes (FITC, LISSA), and two lipid soluble fluorochromes (RB, DCF), are compared to a new “reactive” fluorescent dye introduced by the author (DC-C7A), along with two traditional visible light staining materials (AB2, SB). Apparently all of the fluorescent materials (RB, DCF, FITC, LISSA, and DC-C7A), applied by the author to “standard” test films were illuminated, and viewed under identical conditions. The excitation wavelength of light was 365nm (390 max), and the resultant fluorescent light from the stained samples was viewed with a filter to permit wavelengths beyond 420nm to be seen. While these might be the appropriate illumination conditions for viewing the DC-C7A material advanced by the author, they are not the appropriate wavelengths with which to excite, or view, the other fluorescent materials used in this study. For the record, RB and LISSA absorb light maximally at around 545nm and fluoresces at 565nm (or higher), DCF absorbs at 512nm and fluoresces at 526nm. FITC absorbs at 498nm and fluoresces at 518nm. By using a very low wavelength of excitation light (365nm), the author has insured that most of the fluorescent materials examined in this study would not be excited to fluoresce, and therefore be detectable by any direct means. It is a bit like trying to compare the performance of two flashlights, and leaving the batteries out of one of them; of course one won't work as well as the other. The excitation and emission wavelengths of most of the commonly used fluorescent materials that have been adapted for staining purposes are well established in the literature. Your readers might be interested in obtaining from Kodak the free pamphlet JJ-169 entitled Optical Products, a concise source for the spectral properties of a wide range of commercially available fluorescent materials.

The author has apparently used a 1987 conference preprint as the source for his “methodology” for the application of the fluorescent dyes used in this study (R. C. Wolbers and G. Landrey, “The use of direct fluorescent dyes for the characterization of binding media in cross sectional examinations,” AIC preprints 15th Annual Meeting, 1987, 168–202.) While this preprint was essentially only a general survey of methods for microchemical characterization of paint cross sections in use at that time, it was quite specific about the fluorescent dye structures we used, their concentrations (which the author seems to have arbitrarily changed for this study), and the individual dye reactions. It also stressed the use of appropriate excitation of wavelengths, delivery solvents, and the importance of rinsing away of any unreacted dye materials (with a particular coverslipping medium), to produce the most unequivocal readings possible.

The apparent “reaction” or “non reaction” of a stain is intimately tied to a user's ability to both understand and carry forward a specific chemical reaction on a target substrate. Taking Lissamine as an example, it is a fluorescent dye that carries a sulfonyl chloride (acid chloride) group on it. Acid chlorides react with any free amido, alcohol, or thiol groups that are available for binding. I have used LISSA to mark out proteinaceous materials in cross sections, but generally only after the samples have been pretreated (by selective reduction, or heating, or the like) to uncover as many reactive groups or sites as possible. As I indicated in the 1987 preprint, free amido groups react with acid chlorides, salts of amides do not. Without some kind of pretreatment step, most of the egg proteins, for instance, would give little or no reaction (they're heavily “salted”), and even some collagenous proteins (which are relatively poorly substituted, in free amido groups, but are hydroxyl rich), would react weakly, at best.

The notion of a single stain “identifying” a particular compound is a problematic one. Again, using LISSA as an example: it will react with any compound that contains an alcohol, thiol, or free amino group; this could certainly include proteins, but, as well, carbohydrates, or, for that matter, a myriad of other organic structures found in both natural and synthetic paints. The author seems surprised at one point at the retention LISSA on a “pure” oil standard material; but there is certainly enough glycerol (or sorbitol), and other ingredients such as driers, etc. in many drying oil preparations of modern origin, to react with LISSA substantially. The point is that LISSA (and FITC) are specific functional group reactive dyes. They can certainly be used to “mark out” proteins, if used carefully, and if enough of the appropriate reactive groups are present on a substrate to make them viable reagents. But no stain would ever be used singly, or in isolation, to unequivocally identify a given material. It's more likely that LISSA would be used as part of a carefully considered series or “panel” of stains (or reactions) to more fully, and reliably characterize a proteinaceous material, based on a number of functional groups or features intrinsic to proteins (they are autofluorescent, they can act as acids/bases; they contain thiol groups or not; they may bind certain metal ions such as Ca, etc.).

A second major point of the paper that concerns me is the “synthesis” of the DC-C7A stain, and its concomitant use as a protein tagging reagent. The author purports to have made a particular compound, a “complex of dansyl chloride and beta cyclodextrin” according to the method of Kinoshita et al. (1975), essentially by the additions of a solution of dansyl chloride in acetone, drop-wise to an aqueous solution of beta cyclodextrin. Presumably, the fate of the dansyl chloride (as a sulfonyl chloride) will be to either be hydrolyzed to its free sulfonic acid form upon addition to the water in the aqueous cylcodextrin solution, or to combine with the cyclodextrin (through one of the -OH groups present on it), to form a dansyl sulfonate ester of cyclodextrin. But the author seems to suggest that the synthesized complex still retains a sulfonyl chloride group. This leaves me, as a potential user of this material, to guess the exact composition of the “fine yellow precipitate.” It is a critical point; it makes a great deal of difference in terms of the actual mechanism of staining (or absorption) this new compound will exhibit on an intended proteinaceous substrate. So what then was the compound the author made? How pure was it? (Melting point? Infrared spectrum? Homogeneity by TLC?) As a potential user of this new “protein” stain, what are its spectral properties? Its absorption wavelength? Its fluorescence maxima? I don't have a Zeiss Universal microscope, nor a 487702 filter cube; can I view this new fluorochrome on my microscope, with a slightly different set of filter conditions?

The last point I wish to make is that certain fundamental principles of experimental design seem to have been opted out of in this study. Would phrases like, “Exact numbers of the number of staining attempts for each stain on each media were not recorded. In fact, a particular stain was used to stain a particular media from 1 to 10 times because most cross sections contained three or more known layers.” pass unchallenged in other scientific journals such as JACS or Anal. Biochem.? Precise record keeping, randomization techniques for sampling and testing, proper use of controls, blind or double-blind evaluations, appropriate statistical techniques, and the use of both replicate and repeat measures in experimental designs are necessary if our field wishes to advance its professionalism and credibility.

AUTHOR'S REPLY:

Before I comment on Richard Wolbers' letter to the editor, I would like to state that part of the thrust of my paper was to show that all of these stains (including DC-C7A) can give false indications for many reasons. I attempted to show this through many repeatable experiments, of which I merely intended to report results. I felt compelled to explain why, only when my results were at odds with those reported elsewhere. Now I would like to discuss what I perceive to be the salient points of Wolbers' letter.

Wolbers' most important point is that of the illumination of conditions used, and in particular the Zeiss catalog no. 487702 filter set. The use of this filter set did indeed put DC-C7A in a favorable light (DC-C7A has its maximum excitation wavelength at 335nm, labeled membranes showing maximum fluorescence at 490nm). Use of this filter set would increase the number of protein samples appearing to stain marginally with FITC. This filter set was used because I was under the impression that these were the conditions Wolbers used for viewing most samples. Indeed, Wolbers states in section A14 of his Theoretical and Practical Workshop on the Cleaning of Paintings (1989):

“By broadening the wavelengths of light allowed from the lamp to the sample and from the sample to the viewer … background features that lightly fluoresce … will also be seen and allow for a more contextual reading of Rhodamine stained materials, in the presence of other types of materials. Typically I've used a combination of 360–420nm light as a broad band of excitation wavelengths, and a 430–700nm emission filter in a cube type arrangement to examine cross sections of paint and finish materials. The broad excitation in the near ultraviolet excites and makes fluorescent a number of natural and synthetic materials used in art-making as well as most of the dyes used to stain for them.”

The preceding quote has been taken from a long paragraph and perhaps I am quoting him out of context, because upon rereading this and other materials he makes it evident that he also uses other filter arrangements. However, based on the above and the fact that I noticed early on that the fluorescence of LISSA in the presence of proteins was quite prominent with the Zeiss catalog no. 487702 filter set, I assumed that it would be acceptable for a wide range of conditions. In any event, his criticism is valid. It must now be emphasized that my results with lissamine rhodamine B sulfonyl chloride (LISSA), fluorescein isothiocyanate (FITC), rhodamine B (RB), and dichlorofluorescein (DCF) are only observed when “broad” excitation and “broad” emission wave-lengths are used. My conclusions on the ultimate efficacy of these materials is retracted. I sincerely regret any inconvenience or confusion readers may have experienced due to my failure to use optimal filter sets. Additionally, I am remiss in relying on sources of information that have not gone through a review or editing process.

The chemistry of DC-C7A was questioned. This subject is discussed at length elsewhere in the literature, and I did not wish to burden the reader with it. A good review of cyclodextrin chemistry can be found in J. L. Atwood, J.E.D. Davies, and D.D. MacNicol. eds. Inclusion Compounds, Vol. III. Academic Press. 1984. Chapters 11–13. Dansyl chloride is a common reagent. Briefly, the cyclodextrin used in this study consists of seven glucose units arranged in a ring, such that the hydroxyl groups present on the repeat units are arranged on the outside of the ring. The inside of the ring, a lumen 7.8A in diameter, is hydrophobic. No bond-forming reactions occur between dansyl chloride and the cyclodextrin. Rather the dansyl chloride takes up position in the lumen of the cyclodextrin forming a “clathrate.” The molecule of dansyl chloride is encapsulated. The driving force for complex formation appears to be precipitate formation. The hydrolysis reaction between water and sulfonyl chlorides is thermodynamically favorable, but, kinetically it is very slow (particularly with dansyl chloride). In fact, dansyl chloride can be recrystallized from water; the major impurity, dansyl sulfonic acid, is highly water soluble. Thus, when adding dansyl chloride in acetone to the cyclodextrin in water the sulfonyl chloride is left intact. Being encapsulated, the sulfonyl chloride requires a significantly longer reaction time in order for this fluorochrome to mark the intended materials. In general, molecular encapsulation with a cyclodextrin lowers the volatility and slows the rate of oxidation, hydrolysis, dissolution, etc. of the encapsulated compound. The IR spectrum of DC-C7A is consistent with that reported by Kinoshita et. al. but does not reveal the sulfonyl chloride due to C–H, C–O, O–H bending modes of the oligosaccharide falling directly on top of the S–O stretching mode. If Wolbers is doubtful of the nature of the fine yellow precipitate, I believe he should take it up with Kinoshita et. al. In any event, the fact remains that DC-C7A does work as reported on known compounds and in a reproducible manner.

The statement: “… a particular stain was used to stain a particular media from 1 to 10 times because most cross sections contain three or more known layers”, can be explained by way of example. One particular area of a painting contained a bottom layer of acrylic gesso, over this was zinc white in linseed oil and a top layer of egg white. Another area was also on acrylic gesso, but over this was titanium white in linseed oil and a top layer of egg yolk. If cross-sections of both of these areas are prepared and both are stained with LISSA, then acrylic gesso will have been stained twice (once in each sample) with LISSA while the other media will have been stained only once. In this study I had the luxury of having unlimited numbers of known samples to prepare and to examine. (Typically, five to ten cross sections were made of any area examined.) Thus, there was little need to restain a particular specimen (i.e., cross section) with more than one particular stain. The exception was in the case of the included photographs. Two cross sections were restained in order to make it easier for the viewer to interpret the photographs. In particular, fig. 9 and fig. 12 are the same cross section and rhodamine B was applied before the Sudan black. Fig. 10 and fig. 13 are the same cross section and FITC was applied before the LISSA.

Wolbers devotes two paragraphs to a discussion of the nature of the staining process at the molecular level, making the pedestrian observation that, to use a stain most effectively requires an understanding of the molecular basis of the staining process. He goes on to demonstrate this using as an example, the staining of proteins via the sulfonyl chloride in LISSA. Wolbers assures us that “acid chlorides will react with any free amido, alcohol, or thiol groups that are available for binding [sic]” noting, however, that “free amido groups react with acid chlorides, salts of amides do not.” Regrettably, Wolbers apparently does not understand that amides are essentially neutral, not forming salts except at very low pH, and that free amides are poorer nucleophiles than alcohols, thiols, and amines. Consequently, while he uses “pretreated” samples (has he published details of these pre-treatments?) in LISSA staining to “uncover as many reactive sites as possible,” if those sites were only the amides of a polyamide chain, there would be little reaction between protein and stain. However, if the protein contains amino acids, such as lysine, with amino groups on the side chain, these, if not protonated, can react with the sulfonyl chloride group of the LISSA and bond it to the protein structure.

Although at no point in my paper did I assert that any one single stain could be used for identifying a particular compound or class of compounds, Wolbers seems to conclude that I have done so. In fact, my whole paper is full of qualifying terms such as “may have,” “some,” and “appears to be” along with some caveats about DC-C7A. I feel the field of conservation can benefit from reports on the problems or inconsistencies of DC-C7A.

The last point Wolbers wished to make in his letter appears to me to be critical not only of my scientific and statistical standards, but of those of the reviewers and editors of the Journal. In making his point, he compares these standards unfavorably to those of two journals in which he has never been published. Many members of this field have been attending seminars and reading manuscripts written by Wolbers and yet not one bit of this work has appeared in any reviewed journal. While I believe that most, if not all, of his research is likely to be good and useful, is it right for him to accuse me and the Journal of not working to a set of standards that he has not applied to his own work in more than a decade in this field?

FROM THE EDITOR:

In the review and editorial process John Messinger's paper was treated not as a scientific paper per se but as a paper concerned with the application of an existing methodology of particular interest to paintings conservators. JAIC is an omnibus journal with a broad readership and many papers submitted to us for publication encompass more than one discipline. Because conservation and science are often combined in one paper, it is often difficult to ensure appropriate review of all facets of a paper. Starting in fall 1992 JAIC initiated a policy of assigning papers, when appropriate, to more than one associate editor, representing different disciplines. The implementation of this policy works in two ways. It means that a scientist editor can comment on the scientific component of a paper primarily concerned with conservation. And, equally important, a conservator editor can comment on a scientific paper so that it is put in context for the conservator reader. As this policy makes itself felt, I hope that JAIC will be able to provide the best information possible to its varied and talented readership.

LETTERS WELCOMED

The editor welcomes letters concerning the specific contents of papers published in JAIC. Letters accepted for publication will be printed in whole or in part at the editor's discretion.