Imaging of BrdU-Labeled Human Metaphase ChromosomesWith a High Resolution Scanning Ion MicroprobeRICCARDO LEVI-SETTI,1 * JANM. CHABALA,1 KONSTANTIN GAVRILOV,1 RAFAEL ESPINOSA III, 2AND MICHELLE M. LE BEAU2

1The Enrico Fermi Institute and Department of Physics, The University of Chicago, Chicago, IL 606372Section of Hematology/Oncology, Department of Medicine, The University of Chicago, Chicago, IL 60637

KEY WORDS chromosome banding-nucleoside SIMS mapping-sister chromatid exchange-fragile site-Giemsa

ABSTRACT Detailed maps of the A-T distribution within human mitotic chromosomes labeledwith BrdU are obtained with a high resolution scanning ion microprobe through the detection ofbromine by imaging secondary ion mass spectrometry (SIMS). Corresponding maps of the emissionloci of the molecular ion CN2 describe the overall DNA, RNA and protein distribution in thechromosomes. Several chromosome preparations exhibit base-specific banding patterns (SIMS-bands) which mimic the well known G- or Q-bands resulting from conventional staining methods foroptical microscopy. SIMS-bands are more noticeable in mitotic cells at the first cell cycle and after insitu denaturation or Giemsa staining. Sister chromatid exchanges (SCE) at the second cell cycle andbeyond, occurring both spontaneously and promoted following cell culture exposure to the chemicalaphidicolin (an inhibitor of DNA replication), can be visualized readily from the relative label signalintensities between sister chromatids. The comparison of base-specific label maps with CN2 maps,in conjunction with the appearance of base-specific banding patterns, is informative about proteinsurvival and/or removal following different chromosome preparation protocols. In addition, theresulting condensation state of the chromosomes can be appraised during SIMS analysis from thesample topography (imaged via the collection of mass-unresolved secondary ions). We demonstratethat imaging SIMS is a powerful complement to existing methods for the study of bandingmechanisms and for the elucidation of chromosome structure. The advantages of this novelapproach to the systematic and quantitative study of cytogenetic phenomena and methodologies arestill largely untapped.Microsc. Res. Tech. 36:301–312, 1997. r 1997 Wiley-Liss, Inc.

INTRODUCTIONThe University of Chicago Scanning Ion Microprobe

(UC SIM) utilizes Secondary Ion Mass Spectrometry(SIMS) imaging tomap the distribution of elements andseveral chemical compounds in soft and mineralizedbiological material, at a typical image resolution of30–40 nm (Levi-Setti, 1988). The instrument uses a40–60 keV energy, focused heavy ion probe to scan thesurface of a material, leading to the emission of second-ary ions which are constituents of the material itself.These secondary ions are then collected and analyzedby a mass spectrometer, and recorded to generatemass-resolved images. Because elemental isotopes canbe detected by this approach, the method lends itself tothe use of isotope tracers that need not be radioactive.High-resolution SIMS imaging microanalysis with theUC SIM has already proven to be of unique value in anumber of biological applications (Chabala & Levi-Setti, 1987; Levi-Setti, 1988). More recently, we appliedthe SIMSmethodology to themapping of labeled nucleo-sides in human metaphase chromosomes (for a review,see Levi-Setti and Le Beau, 1992), and in Drosophilapolytene chromosomes (Levi-Setti et al., 1994a). Herewe discuss recent advances in the study of humanmetaphase chromosomes.

The technique was first tested using bromodeoxyuri-dine (BrdU, a thymidine analog) and 14C-thymidine aslabels, to explore the feasibility of studying sisterchromatid exchange (SCE) and fragile sites by thismethod (Hallegot et al., 1989). SCEs result from aspontaneous recombination, by which homologous DNAsequences cross over between sister chromatids, andare thought to be associated with various pathologies.Their rate of occurrence may be promoted by chemicalsand may serve as an index of the mutagenic propertiesof these chemicals. Fragile sites are loci where chromo-somes are prone to break and undergo rearrangements,including homologous and non-homologous recombina-tion. In a few cases, these fragile sites are known to berelated to genetic abnormalities. However, their patho-logical role, genetic and molecular basis, and possibleconnection with SCE, are generally not known. Thelabel signatures in these experiments were the second-ary ion 81Br (free of the PO3 interference that affects theother Br isotope at mass 79) and 28CN2 (14C14N), also

Received 5 January 1996; accepted in revised form 5March 1996.*Correspondence to: Prof. Riccardo Levi-Setti, The Enrico Fermi Institute, The

University of Chicago, 5640 S. Ellis Avenue, Chicago, IL 60637, U.S.A.Grant from: National Science Foundation, Grant number: BIR 9317959.

MICROSCOPY RESEARCH AND TECHNIQUE 36:301–312 (1997)

r 1997 WILEY-LISS, INC.

devoid of significant interferences and copiously emit-ted from DNA. Additional information concerning thechemical structure of the chromosomes can be obtainedthrough mapping with the latter molecular ion in itsnormal 26CN2 form, which accurately describes thecumulative distribution of proteins, as well as of DNAand RNA, providing detailed histological SIMS maps.Subsequent studies using the above labels did succeedin detecting the differential label staining expected forsister chromatids, as well as possible examples of SCE(Levi-Setti and Le Beau, 1992). It also became apparentthat the nucleoside distribution along the chromatidswas not uniform, and that banding patterns could berecognized. In a few examples, these ‘‘SIMS bands’’appeared similar to the well known patterns resultingfrom the trypsin-Giemsa stain (G-bands) and the fluo-rescent quinacrine mustard stain (Q-bands). The SIMS

bands acquire particular significance, as will be dis-cussed further below, because by this approach we hopeto shed light on the true biochemical nature of thebanding patterns observed with the more conventionalstaining methods. This is an issue that has not yet beenresolved (Sumner, 1990) and that has been thought, inrather unspecified terms, to involve DNA-protein inter-actions. The ability to differentiate isotope-tagged DNAfrom proteins using the SIMS technique is particularlyrelevant in regard to this hypothesis.Although encouraging, our previous attempts to map

base-specific label distributions in human metaphasechromosomes were frustrated by the relatively poortransmission (,0.2%) of the RF quadrupole mass filterincorporated in the UC SIM for SIMS analysis at thattime. The resulting detected counts were often insuffi-cient to construct images with statistically useful de-

Fig. 1. Schematized representation of the components of theUniversity of Chicago Scanning Ion Microprobe (UC SIM). By meansof a ‘‘switchyard’’ quadrupole, the secondary ion beam can be directedeither into an Extranuclear 300 RF quadrupole mass filter, or into a

modified Finnigan MAT 90 magnetic sector mass spectrometer. (FromLevi-Setti et al., 1994, reproduced by permission of The Royal Micro-scopical Society).

302 R. LEVI-SETTI ET AL.

tail. In addition, difficulties were encountered in obtain-ing satisfactory spreading of mitotic cells on conductingsubstrates (e.g., Au foils), which exhibit hydrophobicbehavior. The limitations resulting from low countstatistics have been largely overcome in the presentinvestigation, which takes advantage of a radical up-grade of the UC SIM. The instrument is now coupled toa high performance magnetic sector mass spectrometerand exhibits a better than fifty-fold increase in SIMSsensitivity (Chabala et al., 1995; Levi-Setti et al.,1994b). At the same time, a satisfactory protocol hasbeen developed to obtain consistently well-distributedchromosome spreads on conducting substrates (Levi-Setti et al., 1995). It is now feasible to pursue several ofthe original goals of our program, concentrating here onthe SIMS analysis of alcohol-acid-fixed, BrdU-labelledchromosomes. Here we present details of the appear-ance and occurrence of base-specific SIMS-banding inrelation to cell division and to in situ procedures such asDNA denaturation and Giemsa staining. These proce-dures will be shown to deplete the protein content of thechromosomes to some extent. The differential BrdUlabelling of sister chromatids as a function of thenumber of cell divisions will be shown to enable theroutine detection of SCEs occurring in normal mitoticcell cultures, and at an increased rate following expo-sure of the cultures to the chemical aphidicolin. The UCSIM can also yield images of the surface topography of asample by collecting the overall ion-induced secondaryion (ISI) emission from the sample; this feature will beused to illustrate the morphological changes associated

with the different chromosome preparation proceduresmentioned above.

MATERIALS AND METHODSSample Preparation

Because of the irreversible structural incorporationof the BrdU label into DNA, the standard culture andspecimen preparation techniques employed for opticalmicroscope observation are adequate for SIMS analy-sis. However, the requirements that the substrates beconducting to prevent electrical charging, and at thesame time wettable by the mitotic cell fixative, necessi-tates the development of a special methodology. Thechromosome sampleswere obtained fromhuman periph-eral blood lymphocytes of healthy individuals, culturedin BrdU-containing medium following the protocol pre-viously described (Levi-Setti and Le Beau, 1992). Rel-evant to the present context is the incubation time (72hours), which allows, on the average, three cell divi-sions to occur. After arresting the cells in mitosis withColcemid (0.05 µg/ml for 45 minutes, leading to a widerange of chromosome condensation levels), the cultureswere treated with hypotonic KCl (0.075 M, 8 minutes)followed by 4–6 changes of fixative (absolute methanol:glacial acetic acid, 3:1). This method of fixation isknown to remove some of the histone (primarily H1)and non-histone proteins from the chromosomes (Sum-ner, 1990). However, estimates of this loss vary over awide range. As will be shown below, our observationssuggest that a substantial amount of these proteins is

Fig. 2. (a): 81Br2 SIMSmap of a BrdU-labelledmitotic cell, obtained by adding 7 single-passmaps, 35minutes overall scan time. Three chromo-somes overlap an unrelated cell nucleus. Theimage displays 9.1 3 105 counts. Sister chroma-tids are equilabeled, indicating that the cell isundergoing the first mitosis following exposure toBrdU. 45 µm full scale. (b): Same as the image in(a), shown in inverted contrast to facilitate com-parison with optical staining. (c): Magnified de-tail of (b), exhibiting banding patterns (SIMS-bands). 2.0 3 105 counts displayed. 23 µm fullscale. (d): 26CN2 map of area in (c), from asingle-pass scan of 2.2 minutes duration. 1.2 3106 counts displayed. The structure of the base-specific map in (c) is not visible in this map, whichincludes the overall contribution from DNA, RNAand proteins.

303ION MICROPROBE IMAGING OF HUMAN CHROMOSOMES

still present after the above exposure to acid fixative.Instead of using glass slides as substrate, as is routinefor optical microscope observation, the cell suspensionswere dropped onto glass coverslips that have beenlightly Au-coated and treated with a non-ionic surfac-tant (Triton X-100, 0.1%), and air dried. This procedureensured uniform wetting of the substrate by the cellsuspension until dry, a feature that is essential toobtaining well distributed chromosomes which are freeof deformity. For SIMS analysis, the samples werefurther coated with a thin, sputter-deposited layer ofgold.Several variants to the above standard preparation

protocol were introduced, both at the culture level and

in situ, to explore the chemical basis of a number ofcytogenetic phenomena and procedures: 1) To promotethe occurrence of SCE and the expression of fragilesites, some of the BrdU-containing cultures were ex-posed to the chemical aphidicolin (0.4 µM) for 24 hoursat 37°C (Rassool et al., 1992). 2) Toward elucidatingaspects of the staining process, some of the sampleswere Giemsa and trypsin-Giemsa stained, followingstandard procedures. 3) To explore the chemical changesassociated with DNAdenaturation, some of the sampleswere subjected to the procedures for fluorescent in situhybridization (FISH), except exposure to the DNAprobe, according to the formamide/SSC protocol ofRowley et al. (1990).

Fig. 3. (a): Sum of 8 81Br2 SIMS scans totaling 36 minutes, 9.8 3106 counts, 38 µm vertical full scale. The label intensity ratio betweensister chromatids of this BrdU-labeled cell averages close to theexpected 1:1/2 ratio for second division. (b): Magnified detail of (a),showing structure in the packing of chromatin, but no recognizeable

banding. 3.6 3 105 counts, 19 µm full scale. (c): Same as in (a), shownin inverted contrast. Three SCEs are marked by pointers. (d): Anotherdetail of the same chromosome spread, exhibiting 2 additional SCEs.3.5 3 105 counts, 19 µm full scale.

304 R. LEVI-SETTI ET AL.

Scanning Ion MicroprobeAn exploded schematic of the recently upgraded UC

SIM, in the configuration employed in this investiga-tion (magSIMS), is shown in Figure 1. It differs fromthat originally described (Levi-Setti et al., 1986) by itscoupling with a high performance magnetic sector massspectrometer (modified Finnigan MAT 90) for SIMSanalysis. The primary column is unchanged, relative toprevious descriptions, and for the present work was setto focus a 6 pA Ga1 ion beam, extracted from a liquidmetal ion source, to a spot 35 nm wide. The ejectedsecondary ions are collected and energy-analyzed by aspherical 90° electrostatic sector. The secondary iontransport system has been entirely redesigned, and theoverall collection and transmission efficiency in themagSIMS mode is now estimated at ,20%, at a massresolution m/Dm , 550, independent of mass. Com-pared with the former RF quadrupole mass filterperformance, where the transmission efficiency droppedby a factor of 5 with increasing mass, at a massresolution m/Dm , 200, this represents an improve-ment in overall detection efficiency by factors of 20–100over the periodic table of the elements (Chabala et al.,1995). To ensure even transmission of the SIMS trans-port system while the point of primary ion impact isscanned over the field of view, the secondary ion beam issynchronously descanned (dynamic emittance match-ing) so as to be constantly redirected toward the centerof the entrance slit of the magnetic spectrometer.

ImageAcquisition and ResolutionThe SIMS signals are detected by an active film

electron multiplier (ETP AF820, pulse counting) ca-pable of accepting count rates up to 50 MHz. Images,containing 512 3 512 picture elements in a squareraster from single scans, are stored in digital memoryfor subsequent analysis with a KONTRON IMCO im-age processing system. Mitotic cells are located byscanning the samples using the 81Br2 SIMS signaldisplayed on a CRT, at fields of view 160 µm wide.Analytical maps are then recorded from single passscans at fields of view of width in the range of 20–50 µm,to encompass the entire mitotic cell. The 6 pA Ga1

probe erodes the chromosomes very slowly: We esti-mate that about 300 scans, 40 3 40 µm2 in size, at adwell time of 1 µs/pixel, are needed to completely erodea set of BrdU-labelled chromosomes prepared by thestandard preparation procedure above. Counting ratescannot exceed 256 counts/pixel, the limit imposed bythe 8 bit image processor memory, to avoid imagesaturation. This limit is generally reached for theabundantly emitted CN2 in a single pass scan at adwell time of 1 µs/pixel. However, for the 81Br2 label,the sum of 5–10 such single pass maps for a particularmitotic cell is needed (added together after registrationof the images to compensate for occasional minorspecimen drift) before the computer saturation limit isreached. Such stacks of sequential images effectivelyprovide an in-depth view of elemental distributions(SIMS tomography) and may be used for 3D image

Fig. 4. (a): 81Br2 SIMS map of BrdU-labelledmitotic cell cultured in the presence of aphidicolin,shown in inverted contrast. The average numberof SCEs/cell in this sample is 2.5 greater than for anormal culture. Sum of 9 single-pass maps, total-ing 39.3 minutes. 3.7 3 105 counts, 47 µm fullscale. (b): The same cell as in (a), imaged with26CN2. Some structure, due to the condensationpattern of the chromatin, is detectable here. 3.0 3106 counts from a single scan in 2.2 minutes. (c):Detail of another aphidicolin-treated cell, exhibit-ing 5 SCEs. 1.8 3 105 81Br2 counts acquired in 4single-pass scans totaling 17 minutes. 17 µm fullscale. (d): 26CN2 map corresponding to the area in(c), revealing a break (possibly due to a fragilesite) in the q arm of the left chromatid of theuppermost left chromosome. This occurs at thelocation of an SCE, as shown in (c). 5.1 3 106counts from a single scan in 2.2 minutes.

305ION MICROPROBE IMAGING OF HUMAN CHROMOSOMES

reconstruction. To emphasize relevant features and tosmooth the graininess of low signal regions of theimages, gray levels may be minimally averaged usinglowpass routines as needed, without affecting the im-age resolution. The relationship between probe size andimage resolution as a function of signal count statisticsand pixel spacing (related to the magnification andraster size) has been extensively studied (Levi-Setti etal., 1987). The image resolution is defined in terms of anedge spread function, measuring the sharpness of theboundary separating two regions of different signaldensity. Experimentally, this quantity is obtained fromline scans across images of sharp boundaries. Only forsignal density differences across an edge in excess of,10 counts/pixel does the edge spread function ap-proach the probe size. Here, this condition is usuallysatisfied for the CN2 images and for the sum of several81Br2 maps, provided that the probe spacing in theraster be less than the probe diameter (for a raster of512 3 512 and a probe diameter of 35 nm, the field ofview must be 18 µm wide or less). At lower magnifica-tions, the edge spread function approaches, instead, thespacing between pixels, which can be greater than theprobe size. The images presented here were taken fromareas 20 to 50 µm wide, leading to an estimate of theedge spread function, for high counts/pixel regions, inthe range of ,40–100 nm.

BrdU Labeling and Its DetectionBrdU, incorporated in the chromosomes as a thymi-

dine analog, is the carrier of the Br label that isdetected by SIMS. Br is present in its two isotopic

components 79Br and 81Br in their natural abundances,50.54% and 49.46% respectively, and is predominantlyemitted as a negative ion. As previously shown (Halle-got et al., 1989), PO3

2 represents a major interferingmolecule at mass 79 (,39%), that could not be sepa-rated from 79Br2 at the limited mass-resolving power(Dm/m , 200) of the RF quadrupole mass filter used inprevious studies. We have shown (Levi-Setti et al.,1994b) that the magSIMS upgraded version of the UCSIM can indeed separate the two species at mass 79 (atDm/m , 3,400), at the expense however of a reducedSIMS system transmission. In view of the much im-proved detection sensitivity of magSIMS for 81Br2

(,1%) compared with that reported previously (0.01%from Levi-Setti and Le Beau, 1992), we opted as anunwarranted effort the use of the potential additionalcontribution at mass 79.As is well known, and reviewed in detail by Levi-Setti

and Le Beau (1992), the relative amounts of label thatare incorporated within sister chromatids from theBrdU-containing culture medium varies with cell cycle,because the label is only introduced in newly-synthe-sized DNA during chromosome replication, startingfrom unlabeled and one parental DNA. Thus at the firstcell cycle, each chromatid contains one labeled and oneparental (unlabeled DNA) strand: the sister chromatidscarry the same quantity of label, so that the relativelabel amounts stand in the ratio 1/2:1/2 (taking as 1 thecontent of fully labelled DNA). It follows from combina-torial arithmetics that after two S phases, one chroma-tid will carry the label in both DNA strands, the otheronly in one, and the label ratio between sister chroma-

Fig. 5. (a) 81Br2 map of a BrdU-labelled mi-totic cell stained with Giemsa. Here the Br con-tent of eosin overshadows the label, and origi-nates recognizeable Giemsa banding patterns ofequal intensity for both sister chromatids. 9.0 3106 counts from a single scan in 2.2 minutes. 40µm full scale. (b) 26CN2 map of the area in (a). Thebright halo surrounding the chromosomes is at-tributed to the thiazines residue from Giemsa.The apparent lack of emission from the chromo-some interior is a photographic reproduction arti-fact; the signal intensity within chromatids infact equals that of unstained chromosomes. 2 3107 counts from a single scan in 2.2 minutes. (c)Another example of a BrdU-labelled mitotic cellstained with Giemsa. 7.7 3 106 81Br2 counts from5 single-pass scans in 44 minutes. 47 µm fullscale. (d) Magnified detail of (c) showing welldefined SIMS bands. 2.1 3 106 counts, 25 µm fullscale.

306 R. LEVI-SETTI ET AL.

tids will be 1:1/2. At the third cell division, ratios 1:1and 1:1/2 should occur with equal probability. Becausethe SIMS signal is proportional to the concentration ofthe element being detected, it is possible to identify thenumber of divisions of a particular mitotic cell in thepresence of BrdU from the measured ratio of Br labelintensities between sister chromatids. Importantly, atthe second division and beyond, where one chromatidmay carry twice the amount of label as the other, it ispossible to detect SCEs by simple inspection of theSIMS images. In the present experiment the cells werecultured for the average time (72 hours) necessary for 23-cell cycles to occur; therefore, examples of all 3 aboveoccurrences are represented in the results to follow.It should be emphasized that the above predictions

are valid, provided the entire label content of thechromosomes is recorded. This would require the totalsputter erosion of the sample, a goal that is seldomattainable. Because one of our 40 3 40 µm2 SIMS scansprobes only the outermost layers of the samples (,6monolayers), the condensation form of the chromo-somes and the spatial distribution of the proteins mayaffect the actual surface density of DNAdetected by theprobe. As a consequence, the ratios of label intensitiesbetween chromatids that we measure may exhibitsubstantial variation in individual scans, approachingthe predicted values only after averaging over a largenumber of chromosome maps.

RESULTSSeveral mitotic cells were located and analyzed for

each of the sample categories. Representative examples

of 81Br2 and 26CN2 SIMS maps are presented here forseveral of these categories.A few color-coded correlativemaps are shown in the attached color plate.

BrdU-Labeled ChromosomesA partial mitotic cell resulting from the first division

is shown in Figure 2. Forty-three chromosomes are wellseparated. Figure 2a is a 81Br2 map where each de-tected count is displayed to form a bright image,reminiscent of fluorescent staining microscopy. Figure2b is an inverted contrast version of the same, tofacilitate comparisonwithGiemsa-stained samples seenby transmission in the optical microscope. We note firstthat the well-separated chromatids of each chromo-some have similar average image density, as expectedfrom equal incorporation of BrdU in the first celldivision. This is quantitatively confirmed by measuringthe average signal intensity in each chromatid, relativeto that within the cell nucleus, taken as a calibrationstandard. This ratio is close to 1/2 of a similar ratiomeasured for the brighter chromatid of chromosomesfrom the second cell division (where both DNA strandsare labeled), indicating that we observe here only onelabeled strand. We also note that, although the labeldensity is essentially continuous for most chromatids,SIMS bands are discernable, suggestive of Q-bands inFigure 2a, or G-bands in 2b, although perhaps not assharply defined. These banding patterns are moreapparent in the magnified detail of Figure 2c. Also to benoted is the homologous appearance of the SIMS bandsfor sister chromatids, as expected for homologous labeldistributions (identified here as A–T sequences). The

Fig. 6. (a) 81Br2 map of a BrdU-labelledmitoticcell after DNA denaturation with FISH protocol.6.7 3 105 counts from 7 single-pass scans, 31minutes, 36 µm full scale. (b) 26CN2 map of thearea in (a), showing fine condensation structurenot visible in undenatured samples. 2.4 3 106counts from single scan in 2.2 minutes. (c) Detailof another denatured mitotic cell, with markerspointing at 3 SCEs. 1.9 3 105 counts, 7 single-passscans in 31 minutes, 24 µm full scale. (d) 26CN2

map of the area in (c), showing coiling of thechromatid fibers. 4.0 3 106 counts from singlescan, 4.4 minutes.

307ION MICROPROBE IMAGING OF HUMAN CHROMOSOMES

overall chromosomemorphology revealed by thesemaps,together with the SIMS banding patterns, identifiesmost of the chromosomes at first sight. Finally, Figure2d shows a 26CN2 map, corresponding to the detail ofFigure 2(c). Here the structure present in the 81Br2

maps is generally absent, the images of sister chroma-tids often merge together. This masking must be attrib-uted to the remainder of base pairs and proteinsconvoluted with the A–T pairs. The brighter border ofthe chromosome images is attributable to edge effects(artifacts of the SIMS technique), not present in thelabel images, also suggesting a protein coating of theDNA.Figure 3 shows 81Br2 SIMS maps of a partial meta-

phase cell from the second cell division, in both normaland inverted contrast. Here the differential label con-tent of sister chromatids is clearly visible, and mea-sured, on the average, to be in the expected ratio 1:1/2.This leads to the detectability, by inspection, of severalexamples of SCEs, pointed out by markers. The imageintensity for the chromatids with only one labeled DNAstrand, comparable to that of the chromatids in Figure2, appears smaller than the latter due to photographicreproduction constraints. The SIMS banding patternsevident in the singly-labeled chromatids of Figure 2 arereplaced in the doubly labeled chromatids of Figure 3by an essentially continuous knotted structure, sugges-tive in some details of the coiled structure of thechromatid fiber of metaphase chromosomes discussedby, e.g., Gasser et al. (1986). The average rate ofoccurrence of SCEs, measured from several mitoticcells belonging to the same cell culture from an indi-

vidual healthy donor (control), is approximately 4.3/cell, or about 9.4% of the chromosomes. This figure doesnot include apparent SCEs occurring at the centromere(often included in the count), which cannot be distin-guished from a simple twist in the chromosome arms.

BrdU Labeling PlusAphidicolin TreatmentExposure to aphidicolin (Rassool et al., 1992) of

cultures from the same peripheral blood sample as thecontrol results in a marked increase in the average rateof occurrence of SCEs: about 10.3/cell, or 23% of thechromosomes, ,2.5 times that for the control. Ex-amples of these mitotic cells are shown in Figure 4. The81Br2 map of the metaphase cell of Figure 4a exhibits 5SCEs and 7 centromeric twists. The identification ofthese features is facilitated by comparison with theCN2 map of Figure 4b, which characterizes the chromo-some appearances and allows a precise localization ofthe centromeres. A higher magnification detail of an-other mitotic cell is shown in Figure 4c and d. Here 5SCEs and 1 twist can be identified, the uppermost leftSCE occurring at the site of a break in the left chroma-tid (a possible fragile site induced by aphidicolin)visible in the CN2 map.

BrdU Labeling Plus GiemsaThe Br content of eosin (C20H6Br4Na2OS) in Giemsa

(a mixture of eosin with the thiazines methylene blueand azure B) was found to overshadow (by at least afactor of 10) the BrdU label, masking the differentiationin label content of sister chromatids. Nonetheless it is

Fig. 7. Topographic images of mitotic cells,obtained withmass-unresolved secondary ions. (a)and (b) show 2 different condensation states ofBrdU-labelled chromosomes, 37 and 39 µm fullscale respectively. (c) Collapsed appearance ofchromosomes after denaturation in 70% for-mamide 4 3 SSC, 70°C, following FISH protocol,36 µm full scale. (d) Highly condensed and re-solved chromatid appearance after Giemsa stain-ing, 47 µm full scale.

308 R. LEVI-SETTI ET AL.

still useful to investigate the chemical changes associ-ated with Giemsa staining using SIMS. The 81Br2 mapsof 2 Giemsa-stained chromosome spreads are shown inFigure 5a and c. Figure 5b is a CN2 map correspondingto (a), and Figure 5d a magnified detail of 5c. Severalfeatures stand out in these SIMS maps. First, theapparent absence of CN2 emission from the chromo-somes in 5b is a visual artefact of the photoreproduc-tion. In fact, the measured CN2 signal intensity withinthe chromosomes corresponds closely to that measuredfor e.g., Figures 2d or 4d in the absence of Giemsastaining, although the chromatids appear better sepa-rated and closely mimic the 81Br2 images in the presentcase. This comparison to non-stained samples suggeststhat some of the proteins not closely packed with DNAmay have been removed by staining, and that theoriginal histone proteins-DNAcomplexmay either havebeen left untouched, or partially replaced by the nitro-gen-carrying components of Giemsa (thiazines). Sec-ond, CN2 originates from a bright halo surrounding thechromosomes, with signal intensity about double thatof the chromosome interior. This halo can be attributedto the residues of thiazines (eosin contains no nitrogen)left by capillarity after in situ Giemsa staining. Moreimportantly, the Br2 distribution, in this case represent-ing the eosin distribution, shows transverse bands, notsurprisingly similar to the Giemsa patterns. We shallreturn to these observations, related to the bandingmechanism of optical stains, in the discussion section.

BrdU Labeling Plus DNADenaturationTwo examples of mitotic cells subjected to the FISH

procedure (without DNA probe) are shown in Figure 6.

Both originate from the third cell division, because ofthe presence of chromosomes exhibiting 1:1 and 1:1/2label ratios between sister chromatids in the 81Br2

SIMS maps (Figs. 6a and c). Two SCEs are marked bypointers in Figure 6c. The label intensities within thechromatids are comparable to those measured for non-denaturated samples. Of particular interest are thecorresponding CN2 maps of Figure 6b, and especiallythat of Figure 6d. In contrast to the rather uniformmaps for CN2 previously shown for non-denaturatedsamples (see Fig. 2d and 4d), these maps exhibitremarkable structure and evidence of a coiling arrange-ment of the chromatid fiber. Perhaps unexpectedly, theCN2 signal within the chromatids is about twice asintense as that for non-denaturated chromosomes, seem-ingly in contradiction with the evidence that much ofthe protein coating, normally widening the slenderstructure of the coiled chromatin, has been removed bythe denaturation process. We propose an explanationfor this apparent discrepancy by referring to the conden-sation state of the chromosomes, revealed by theirtopography. This is shown in Figure 7, where Figures7a and b show the appearance of two different presenta-tions of mitotic cells among our samples, prior to anyfurther in situ procedure; Figure 7c shows the result ofdenaturation, and Figure 7d that of Giemsa staining.The denatured chromosomes in Figure 7c appear flat-tened with a thin raised rim, suggestive of the collapsedchromosome state reported (Gormley and Ross, 1972;Sumner, 1990) following treatment with hot 23 SSC,part of the ASG (acetic/saline/Giemsa) staining methodof Sumner et al. (1971). We can surmise that theintensity of the SIMS signal is a function of the aspect

Fig. 8. Comparison of optical microscope im-ages of trypsin-Giemsa stained mitotic cells, withbase-specific SIMSmaps fromBrdU-labelled cells.(a) and (b) optical microscope, 50 and 33 µm fullscale respectively. (c) 81Br2 SIMS map, seconddivision, sum of 16 single pass maps recorded forSIMS tomography, 5.6 3 105 counts, 70 minutes,39.1 µm full scale. (d): 81Br2 SIMS map, thirddivision, sum of 9 single-pass scans, 6.7 3 105counts, 39.3 minutes, 32 µm full scale.

309ION MICROPROBE IMAGING OF HUMAN CHROMOSOMES

Figure 9.

ratio of the chromatin exposed to the primary probe. Acollapsed state would then favor a better secondary ionyield than a more compacted normal state, or a con-tracted state such as that indicated as a result ofGiemsa staining in Figure 7d.

DISCUSSIONThe images presented in the previous section are

illustrative of the direct information that high resolu-tion SIMS imaging can provide concerning chromo-somal structure and composition. In conjunction withBrdU labeling, the method enables the detection ofSCEs at the base-specific level in images of comparablecontrast and resolution to those obtained with fluores-cent dye techniques. Depth compositional informationfrom sequential SIMS mapping (although not exploitedin this presentation) is also available to explore thebase-specific 3D structure of the chromosomes. In con-junction with the label maps, the corresponding imagestakenwith the CN2 secondary ions provide complemen-tary information about the overall chromosomemorphol-ogy that may occasionally be valuable. Thus, in oneexample shown (Figures 4c and d), this comparison hasmade it possible to detect a break in one chromatid(possibly due to a fragile site), coincident with the occur-rence of an SCE. Furthermore, the difference betweenthese two sets of images, representing respectively theDNA core and the overall DNA-protein complex, revealsthe level of protein digestion due to the fixation process andsubsequent in situ procedures.At the same time, images ofthe surface topography indicate the condensation stateof the chromosomes under investigation.Before discussing our observations that may be rel-

evant to the mechanism of chromosome banding, it is

appropriate to compare in Figure 8 the optical micro-graphs of trypsin-Giemsa stained chromosomes (Figure8a and b), with base-specific SIMS maps of BrdU-labeled chromosomes (Figure 8c and d). As previouslynoted in regard to Figure 3, fully-labeled chromatidsshow an essentially continuous label (or A-T pair)distribution, with only occasional suggestion of band-ing, never differentiated at a level comparable to theGiemsa bands. Even after denaturation, the correlationbetween SIMS and Giemsa banding (see Figure 8) forfully-labeled chromatids is at best rather weak. Only inmonofilarly labeled chromosomes, such as those inFigure 2, banding suggestive of the Giemsa patterns,that can be associated with possible regions of A-Tenrichment are discernable. These observations mili-tate against attributing Giemsa banding exclusively topatterns of base-specific condensation, at least in com-pacted metaphase chromosomes such as those exam-ined in the present survey. The pattern of overallchromosomal condensation becomes visible instead inthe CN2 maps of denatured chromosomes. Again how-ever, this pattern, although structured, appears notselectively differentiated. In view of the Br content ofeosin, the detection of Giemsa-like bands in the Br2

maps of Giemsa-stained chromosomes, at an intensitylevel that is higher than that provided by BrdU label-ing, simply reinforces the notion that eosin is associ-atedwithG-positive bands. By and large, these observa-tions are consistent with the discussion of G-bandingsummarized by Sumner (1990). Hints of the possiblerole of divalent cations are found in the literature andare also referred to by this author. We are presentlyinvestigating the distribution of Ca and Mg in oursamples.It should be noted that our observations on denatur-

ated chromosomes represent a successful preliminaryto the localization of specific labelled probes, to develop,in analogy with FISH, a methodology that should becalled SISH (for SIMS in situ hybridization). This goalseems attainable using, e.g., BrdU-labelled probes (withchromosomes not carrying the same label), or probeslabelled with other thymidine analogs such as iodode-oxyuridine or fluorodeoxyuridine. By a similar ap-proach, using stable isotope labels such as, e.g., deute-rium it seems feasible to aim at the localization ofregulatory proteins, in conjunction with SISH, forspecific gene sequences. Further investigations alongthese lines, with the UC SIM, are being formulated atthis time.

ACKNOWLEDGMENTSThis report is based on work supported by the

National Science Foundation under grant BIR 9317959.We wish to thank Dr. Pamela L. Strissel for informativediscussions.

REFERENCESChabala, J.M., and Levi-Setti, R. (1987) High-resolution chemicalmapping of biological material by use of a heavy ion probe. In:Analytical Electron Microscopy—1987. D.C. Joy, ed. San FranciscoPress, San Francisco, CA, USA, pp. 203–207.

Chabala, J.M., Soni, K.K., Li, J., Gavrilov, K.L., and Levi-Setti, R.(1995) High-resolution chemical imaging with scanning ion probeSIMS. Int. J. Mass Spectr. Ion Processes, 143:191–212.

Gasser, S.M., Laroche, T., Falquet, J., Boy de la Tour, E., and Laemmli,U.K. (1986) Metaphase chromosome structure. Involvement oftopoisomerase II. J. Mol. Biol., 188:613–629.

Fig. 9. (a) 81Br2 SIMS map of portion of a BrdU-labelled mitoticcell. Sister chromatids are equilabeled, indicating first cell divisionwhere only one DNA strand of each chromatid is labeled. The image isshown on a pseudocolor scale which emphasizes a banded structuredue to base-specific (A-T) condensation (blue:zero signal; yellow–orange:maximal signal). A coiled structure of the chromatin fibers is alsoapparent in several chromosomes. (b) 81Br2 SIMS map of a BrdU-labelled mitotic cell, where the label intensity ratio between sisterchromatids averages close to the expected 1:1/2 ratio for second celldivision. Several occurrences of sister chromatid exchange (SCE) canbe detected in this micrograph. (c) False color superposition of a 81Br2

SIMS map (A-T pair distribution) of a BrdU-labelled mitotic cell(green), with the corresponding 26CN2 map (blue). The latter includesthe overall contribution fromDNA, RNAand proteins. This correlativepresentation emphasizes the occurrence of SCEs. (d) Portion ofBrdU-labelled mitotic cell cultured in the presence of the chemicalaphidicolin. The average rate of occurrence of SCEs in this sample is2.5 times that of a normal culture. The image is a color-codedsuperposition of 81Br2 (green) and 26CN2 (blue) SIMS maps. FourSCEs are pointed out by arrows. In this example, the 26CN2 distribu-tion extends beyond the chromatid outline, possibly due to residues ofnuclear material. (e) Portion of another BrdU-labelled mitotic cellcultured in the presence of the chemical aphidicolin. 81Br2 is colorcoded in white, 26CN2 in blue. Here the latter map helps identify abreak in the p-arm of the left chromatid in the chromosome located inthe NW corner. This break occurs in conjunction with an SCE and isprobably due to a fragile site. Several additional examples of SCEs areclearly identifiable in this correlative image. (f ) Correlative superposi-tion of 81Br2 (white) and 26CN2 (blue) SIMS maps for a BrdU-labeledmitotic cell stained with Giemsa. Here the Br content of eosinovershadows the label, and originates recognizeable Giemsa bandingpatterns of equal intensity for both sister chromatids. The bright halosurrounding the chromosomes is attributed to the thiazine residuefrom Giemsa.

311ION MICROPROBE IMAGING OF HUMAN CHROMOSOMES

Gormley, I.P., and Ross, A. (1972) Surface topography of humanchromosomes examined at each stage during ASG banding proce-dure. Exp. Cell Res., 74:585–587.

Hallegot, P., Girod, C., Le Beau, M.M., and Levi-Setti, R. (1989) Highspatial resolution SIMS imaging of labelled human chromosomes.In: Secondary Ion Mass Spectrometry-SIMS VII. A. Benninghoven,C.A. Evans, K.D. McKeegan, and H.A. Storms, eds. John Wiley &Sons, NewYork, NY, USA, pp. 327–330.

Levi-Setti, R., Chabala, J., and Wang, Y.L. (1987) Micro-secondary ionmass spectrometry: Physical and instrumental factors affectingimage resolution. Scanning Microsc. Suppl., 1:13–22.

Levi-Setti, R. (1988) Structural andmicroanalytical imaging of biologi-cal materials with heavy-ion probes. Ann. Rev. Biophys. Biophys.Chem., 17:325–347.

Levi-Setti, R., and Le Beau, M.M. (1992) Cytogenetic applications ofhigh resolution secondary ion imaging microanalysis: detection andmapping of tracer isotopes in human chromosomes. Biol. Cell,74:51–58.

Levi-Setti, R., Chabala, J.M., and Smolik, S. (1994a) Nucleotide andprotein distribution in BrdU-labelled polytene chromosomes re-vealed by ion probe mass spectrometry. J. Microsc., 175:44–53.

Levi-Setti, R., Chabala, J.M., Gavrilov, K., Li, J., Soni, K.K., andMogilevsky, R. (1994b) Imaging with a high-resolution SIM. In:Secondary Ion Mass Spectrometry-SIMS IX. A. Benninghoven, Y.

Nihei, R. Shimizu, and H.W. Werner, eds. John Wiley & Sons,Chichester, England, pp. 233–237.

Levi-Setti, R., Chabala, J.M., Gavrilov, K., Espinosa III, R., and LeBeau, M.M. (1995) Advances in imaging SIMS studies of BrdU-labelled human metaphase chromosomes. In: Secondary Ion MassSpectrometry-SIMS X. A. Benninghoven, ed. John Wiley & Sons,Chichester, England, in press.

Rassool, F.V., Le Beau, M.M., Neilly, M.E., van Melle, E., Espinosa III,R., and McKeithan, T.W. (1992) Increased genetic instability of thecommon fragile site at 3p14 after integration of exogenous DNA.Am. J. Hum. Genet., 50:1243–1251.

Rowley, J.D., Diaz, M.O., Espinosa III, R., Patel, Y.D., van Melle, E.,Ziemin, S., Taillon-Miller, P., Lichter, P. Evans, G.A., Kersey, J.H.,Ward, D.C., Domer, P.H., and Le Beau, M.M. (1990) Mappingchromosome band 11q23 in human acute leukemia with biotinylatedprobes: Identification of 11q23 translocation breakpoints with ayeast artificial chromosome. Proc. Nat. Acad. Sci. USA, 87:9358–9362.

Sumner, A.T. (1990) Chromosome banding. Unwin Hyman, London,England, p. 92.

Sumner, A.T., Evans, H.J., and Buckland, R.A. (1971) New techniquefor distinguishing between human chromosomes. Nature New Biol.,232:31–32.

312 R. LEVI-SETTI ET AL.