Development of X-Ray Excitable Luminescent Probes for Scanning X-Ray Microscopy

Ultramicroscopy 77 (1999) 23}36

Development of X-ray excitable luminescent probes for scanning X-ray microscopy

Mario M. Moronne*

Innovative Microscopies Group, Life Science Division, 1 Cyclotron Rd., M/S 6-2100, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA Received 13 December 1997; received in revised form 7 December 1998

Abstract

Transmission soft X-ray microscopy is now capable of achieving resolutions that are typically 5 times better than the best-visible light microscopes. With expected improvements in zone plate optics, an additional factor of two may be realized within the next few years. Despite the high resolution now available with X-ray microscopes and the high X-ray contrast provided by biological molecules in the soft X-ray region (j"2}5 nm), molecular probes for localizing speci"c biological targets have been lacking. To circumvent this problem, X-ray excitable molecular probes are needed that can target unique biological features. In this paper we report our initial results on the development of lanthanide-based #uorescent probes for biological labeling. Using scanning luminescence X-ray microscopy (SLXM, Jacobsen et al., J. Microscopy 172 (1993) 121}129), we show that lanthanide organo-polychelate complexes are su$ciently bright and radiation resistant to be the basis of a new class of X-ray excitable molecular probes capable of providing at least a "vefold improvement in resolution over visible light microscopy. Lanthanide probes, able to bind 80}100 metal ions per molecule, were found to give strong luminescent signals with X-ray doses exceeding 10 Gy, and were used to label actin stress "bers and in vitro preparations of polymerized tubulin. ( 1999 Elsevier Science B.V. All rights reserved.

PACS: 07.85.Y

Keywords: X-ray microscopy; Fluorescent probes; Lanthanides

1. Introduction have been used to study systems as diverse as in- tracellular signaling, protein processing, nuclear In the past two decades, cell biology has been transcription sites, gene localization, and the struc- revolutionized by the development of highly speci- ture of the cytoskeleton. Fluorescent labels with "c #uorescent labels that can accurately target bio- high quantum e$ciencies combined with high- logical molecules of interest. Fluorescent probes quality confocal systems, are routinely used to vis- ualize biological structures with resolutions approaching the visible light limit of about 200 nm. * Tel.: #1-510-486-4236; fax: #1-510-486-5664. The most direct approach to increase resolution E-mail address: [email protected] (M.M. Moronne) beyond this limit is to use techniques that can

0304-3991/99/$ } see front matter ( 1999 Elsevier Science B.V. All rights reserved. PII: S 0 3 0 4 - 3 9 9 1 ( 9 9 ) 0 0 0 0 2 - 9 24 M.M. Moronne / Ultramicroscopy 77 (1999) 23}36 operate at shorter wavelengths. Electron micros- ventional light microscopy. Unfortunately, it was copy (EM) can achieve atomic resolution but also determined that organic #uors that have pro- requires specimen preparation and handling ved so useful in visible light microscopy are very methods that are potentially damaging to biolo- labile to soft X-rays and are not stable enough to be gical materials and becomes progressively less ef- used as X-ray probes. fective for samples more than a few tenths of a In this report, we describe the "rst detailed re- micron thick. In addition, molecular labels for EM sults of soft X-ray biological probes based on lan- are mostly limited to derivatized gold particles that thanide polychelate conjugates. We show that tend to su!er from problems of sparse labeling and lanthanide probes are of su$cient brightness and di$culties in penetrating cells. stability under X-ray excitation to be used for de- In recent years, soft X-ray microscopy signing high-resolution luminescent biological (j"2}5 nm) has achieved resolutions close to labels. Lanthanides were chosen for study since 30 nm [2]. Future improvements in zone plate their luminescence is based on their atomic spectral lenses are expected to give resolutions approaching properties and not the highly X-ray labile con- 10 nm. The ability of X-ray microscopes to image jugated ring systems common to conventional #u- hydrated biological specimens up to 10 lm thick orescent dyes. The fact that lanthanides such as Tb using native elemental X-ray absorption, makes it and Eu are common components of phosphors an attractive alternative to EM, eliminating the used in CRTs bears testimony to their ability to need for plastic embedding and sectioning. How- tolerate enormous doses of radiation. In addition, ever, despite the improved resolution of X-ray lanthanide compounds have essentially no overlap microscopes, and the high-quality morphological between their excitation and emission bands, i.e., detail they can provide, there has been a lack of they have very large Stokes shifts ('100 nm). As molecular probes capable of localizing speci"c cell a result, their per atom quantum yields are not proteins and other important subcellular targets. signi"cantly depressed when assembled into multi- To circumvent this problem, labeling agents are meric structures compared with single atoms [6]. needed that are designed speci"cally to operate This property makes it practical to increase lan- using X-ray excitation. In this regard, two ap- thanide probe brightness by building multi-residue proaches have been taken including dark-"eld chelates (polychelates) capable of binding a large X-ray microscopy using silver enhanced gold label- number of lanthanides in a single-probe molecule. ing as in EM [3], and scanning luminescence X-ray In contrast, conventional organic #uors such as microscopy (SLXM, [1,4,5]). Gold labeling may #uorescein show signi"cant self-quenching when prove very useful; however, it is subject to many of more than a few groups are attached to probe the same caveats as applies to its use in EM, and is molecules such as antibodies. further limited by the requirement for silver en- In this study we have built polychelates probes hancement that makes it impractical to study mul- that contain 80}100 lanthanides per molecules. tiple labels in the same specimen. Photostability tests of these compounds show Previously, Jacobsen et al. [1] adapted the that lanthanide complexes continue to produce scanning transmission X-ray microscope at the Na- useful luminescence with radiation doses well in tional Synchrotron Light Source (NSLS, Brook- excess of 10 Gy. Further, luminescent yields are haven National Laboratory) to measure X-ray su$cient to provide high-quality signals with spa- excited visible light luminescence. These workers tial resolution down to 50 nm or better. In addition made use of the fact that in a scanning con"gura- to these measurements, we present the results of tion, microscope resolution is determined by the biological labeling experiments of actin stress "bers spot size of the exciting radiation rather than the in "broblasts and in vitro preparations of micro- wavelength of the visible luminescence. With this tubule assemblies using polychelate conjugates of approach, Jacobsen and colleagues obtained a lat- secondary antibodies, and a biotinylated poly- eral resolution of about 75 nm on P31 phosphor chelate that is readily adapted to avidin}biotin beads, close to a three fold improvement over con- labeling protocols. M.M. Moronne / Ultramicroscopy 77 (1999) 23}36 25

2. Materials and methods was substituted with thiols by reacting with N- succinimidyl-3[2-pyridyldithio] propionate (SPDP, 2.1. Synthesis of polychelate luminescent probes Molecular Probes) then reduced with dithiothreitol (DTT) to generate two-to-three free sulfhydryls per Two approaches to polychelate probe design chain. The maleimide-antibody and polylysine-SH were implemented: (a) similar to the method of derivatives were then mixed and reacted overnight Can" et al. [6], a goat anti-rabbit secondary anti- yielding a goat anti-rabbit polylysine conjugate. body was "rst linked with polylysine then acylated Precipitated material was removed by centrifu- with the conjugate of diethylenetriamine pen- gation. To the supernatant was added the DMSO taacetic acid (DTPA) and p-aminosalicyclic acid solution of DTPA-pAS monoanhydride prepared (pAS), and (b) a biotinylated polychelate was made as described above. The "nal DMSO concentration by "rst conjugating polylysine with biotin and rho- was kept below 20%. Aggregates were removed by damine succinimidyl esters, followed by extensive centrifugation and the conjugate dialyzed against substitution with DTPA-pAS. The inclusion of pAS 10 mM Tris bu!er pH 8.0. From estimates of the in the chelating subunit serves two purposes. When "nal protein content using BCA protein reagent conjugated to DTPA, pAS acts as an e$cient (BioRad, CA), and the Tb binding capacity mea- sensitizer of Tb luminescence [7], and may sured by spectro#uorometry (SPEX Fluorolog, contribute signi"cantly to Tb X-ray excitation. In SPEX Industries, NJ), each IgG bound 80 Tb per addition, it is a useful marker during probe syn- antibody. thesis and conjugation, since it facilitates measurement of Tb loading, and can be measured with a sensitivity in the picomolar range by spectro- 2.1.3. Biotinylated rhodamine polychelate (BiotinRP) #uorometry. General preparation methods are To prepare the biotinylated rhodamine poly- & e given below. chelate, 20 mg of a 20 kD polylysine ( 100 -NH per chain, Sigma) was dissolved in 1 ml of anhyd- e 2.1.1. DTPA-pAS monoanhydride rous DMSO (94 mM -NH). The solution was The monoanhydride reagent of DTPA-pAS was then made 5 mM in rhodamine succinimidyl ester prepared by adding 0.714 g (2 mmol) of DTPA di- (Molecular Probes) and biotin-LC-succinimidyl anhydride to 6 ml of anhydrous DMSO and 0.7 ml ester (Pierce Chemical, IL.) and incubated for 3 h of triethylamine. To this solution, 0.356 g (2 mmol) at 373C. The reaction mixture was then added of pAS dissolved in 3.3 ml of DMSO was added dropwise with constant stirring to DTPA-pAS dropwise with constant stirring to give a "nal monoanhydride leaving a "nal "vefold molar ex- e 200 mM concentration of chelator. The mixture cess of the anhydride to -NH groups. The was allowed to react for 3 h at room temperature mixture was then incubated overnight at room tem- and used without further puri"cation. Each reagent perature; any precipitate formed was removed by was obtained from Aldrich Chemicals (Milwaukee, centrifugation. The supernatant was then dialyzed WI). repeatedly in the dark against HBS (10 mM HEPES, 150 mM NaCl at pH 7.3) using dialysis 2.1.2. Goat a-rabbit polychelate conjugate cassettes with a 10 kD membrane cuto! (Pierce The antibody conjugate was prepared in co- Chemicals). operation with Molecular Probe, Inc. (Eugene, Rhodamine was used in the polychelate to Oregon). Goat a-rabbit IgG secondary anti- provide a useful #uorescent tracer to test labeling body was "rst reacted with succinimidyl 4-(N- protocols using conventional #uorescence light maleimidolmethyl)-cyclohexane-1-carboxylate microscopy. In addition, the chromophore pro- (SMCC, Molecular Probes). This introduces two to vides a convenient visual marker for following the three maleimide groups into the antibody for later polychelate through subsequent puri"cations and crosslinking to polylysine. Polylysine (20}25 kD, conjugations. Fig. 1 shows the basic polychelate e 100}120 -NH per chain, Sigma Chemical Co.) probe design. 26 M.M. Moronne / Ultramicroscopy 77 (1999) 23}36

2.3. Microtubule assemblies

Monomeric tubulin was dissolved to a concentration of 1 mg/ml in bu!er containing 10 lm taxol, 2 mM MgCl, 150 mM NaCl, 1 mM GTP, and 10 mM phosphate bu!er, pH 7.4. Polymerization was allowed to proceed at 373C for 30 min. Ten ll of the polymerized suspension was then allowed to settle for 10 min on polylysine treated formvar coated grids, wicked o! with "lter paper, and air dried. The dried tubules were then "xed with 4% paraformaldehyde in PBS pH 7.4 for half an hour. The "xed tubules were gently rinsed 3 times for 10 min with HBS and labeled immediately (see Labeling Protocols, below)

2.4. Labeling protocols Fig. 1. Biotinylated DTPA polychelate probe (Tb-BiotinRP) with p-aminosalicylic acid sensitizer and rhodamine tracer. Total cell actin (F- and G-actin) in 3T3 cells was labeled by "rst incubating with a rabbit anti-actin primary (Sigma) suspended in 1% bovine serum albumin in TBS (10 mM Tris, 150 mM NaCl, pH 2.2. Cells 7.4) for 1 hr. This was followed by three 10 min rinses in 1% BSA-TBS, then treatment for 30 min NIH 3T3 mouse "broblasts were cultured in with the goat anti-rabbit-polychelate-Tb antibody Dulbelco modi"ed Eagles medium containing 10% (50 lg/ml). After three washes in BSA-TBS, a "nal calf serum (Gibco) at 373C with 5% CO. Wi38 rinse was done with 15 mM NaCl and the speci- embryonic lung "broblasts were similarly grown mens allowed to air dry. except 10% fetal calf serum was used. Specimen To label cell F-actin with biotinRP, "xed per- substrates consisted of either formvar coated 200 meabilized 3T3 specimens were incubated for mesh gold "nder grids, or 1000 As thick silicon ni- 20 min with a 20 : 1 dilution of phalloidin}biotin tride windows (3.5;3.5 mm) etched in 100 lm (Molecular Probes) in HBS. Specimens were rinsed thick 12;12 mm silicon squares. Fibronectin coat- twice in HBS, then once with Tris bu!ered Super- ing was found to greatly improve cell adhesion to block (TSB, Pierce Chemicals). This was followed both grids and silicon nitride windows and was by incubation with streptavidin (10 lg/ml in TSB) used routinely. In either case, substrates were for 30 min. Subsequently, specimens were rinsed placed in petri dishes, covered with media, then once with TSB, three times with HBS, then treated layered with cells and cultured for 1}3 d until with Tb loaded biotinRP (25 lg/ml) for 30 min. a convenient but subcon#uent density was ob- Finally, samples were rinsed 3 times in HBS and tained. Specimens for actin labeling were then twice in distilled water before air drying. washed with phosphate bu!ered saline (PBS) pH Microtubule assemblies deposited on formvar 7.4, permeabilized for 30 s with 0.5% Triton X-100 coated EM grids were "rst labeled for 30 min with in PBS, then "xed with a solution containing 4% an anti-b tubulin mouse monoclonal (CalBiochem, paraformaldehyde, 0.2% glutaraldehyde, 0.5% CA) diluted 1 : 50 in TSB. After three 10 min rinses Triton X-100, 5 mM MgCl, 150 mM NaCl, 1 mM in TSB, the tubules were incubated for 30 min in EGTA, and 10 mM phosphate at pH 7.4 for 1 h. biotinylated goat anti-mouse (CalBiochem, 1 : 50 in Specimens were subsequently stored in PBS at 43C TSB) then rinsed three times in TSB for 10 min. until ready for labeling. This was followed by incubation with streptavidin M.M. Moronne / Ultramicroscopy 77 (1999) 23}36 27

(10 lg/ml in TSB) for 30 min, three rinses in HBS, a formvar coated gold EM grid and dried. The then a 30 min incubation with Tb loaded biotinRP. X-ray beam was then focused on the antibody- Samples were then rinsed three times with HBS, polychelate "lm and irradiated with 30.9 As X-rays twice with water, and "nally air dried. while the luminescence signal was simultaneously measured. This wavelength was chosen because it 2.5. Scanning luminescence X-ray provides the highest available #ux from the X1-A microscopy (SLXM) beamline. To further boost the incident X-ray #ux for luminescence imaging, helium gas was used to Fig. 2 shows the basic X-ray microscope con"g- purge the air space between the sample and zone uration used in the present studies. It is essentially plate-exit window assembly. X-ray transmission identical to the original design of Jacobsen et al. [1] calculations were done using X-CAL 6 software for with the exception that a higher numerical aperture Macintosh (Oxford Research Group, Richmond, objective is used for light collection and the photo- CA), which incorporates the X-ray data tables of multiplier has been replaced with a more quantum Henke et al. [8]. e$cient avalanche photodiode (APD, EG&G, Ca- nada). The APD design was described previously [5]. Taken together at least a "ve fold improvement 2.6. Fluorescence light microscopy in detection e$ciency is expected over the earlier PMT design with signi"cantly reduced dark noise Visible light #uorescence images were taken us- (5}20 cps dark count rate for the APD). A zone ing an Olympus OM-2 microscope equipped with plate lens with a 45 nm outer zone width was used a Xillix Model 1400 cooled CCD camera. Camera throughout these experiments with a nominal operation was controlled via a Power Computing Rayleigh resolution of 56 nm [1]. Power Center 150 equipped with a Snapper image For X-ray photobleaching experiments, Tb acquisition board (Technical Instruments, CA) and loaded antibody-polychelate was deposited on IPLab Spectrum software.

Fig. 2. Scanning X-ray microscope con"gured for luminescence detection (SLXM) using an avalanche photodiode. 28 M.M. Moronne / Ultramicroscopy 77 (1999) 23}36

3. Results luminescence was followed over time from the cen- tral region of the antibody "lm. 3.1. Luminescence vs. dose To estimate the absorbed dose, it is necessary to determine the thickness of the antibody probe An X-ray excitable probe must be able to tolerate layer. This is done by (a) measuring the #ux trans- exposure to 2}5 nm X-rays most commonly used mitted through a region of formvar without probe, for X-ray microscopy. X-rays of these wavelengths I&, and (b) through probe and formvar region, I&>. are in the range of the K-shell ionization energies of Eqs. (1)}(3) are then used to estimate the thickness carbon (284 eV) and nitrogen (410 eV) so that ab- of the probe "lm: sorption of a single photon by a chemical dye such I "I e\IDVD (through formvar), (1) as #uorescein is likely to destroy the conjugated & aromatic rings essential to its #uorescence. This is " \I&V& \IV I&> Ie e (through probe and in addition to radical damage resulting from hy- droxyl and other highly reactive species generated formvar), (2) by X-rays in the media. I&> To determine the radiation stability and lumines- "e\IV, (3) I cent yield of a model lanthanide probe, a comparat- & a \ \ ively thick "lm of a terbium loaded goat -rabbit where I is the incident #ux (photons cm s ), k k polychelate conjugate was deposited onto a for- & and are the formvar and probe X-ray absorp- \ mvar coated EM grid. Fig. 3 shows the X-ray tion coe$cients (cm ), and x& and x are their transmission image (left) and the corresponding respective thicknesses. luminescence image (right) for this compound. From measurement of the Tb binding capacity Thicker regions of the "lm show up as dark areas in of the antibody (80 Tb per molecule), the chem- the transmission image and bright areas by ical formula for DTPA-pAS substituted polylysine, luminescence. Also, various cracks can be seen and a formula for protein, an elemental throughout the dried "lm that are useful for calib- composition for the antibody conjugate of ration purposes. To evaluate the quantum yield of CONHS Tb is obtained. This gives an k the probe (visible light emitted/X-ray absorbed) estimated mass absorption coe$cient ( )of and its radiation stability, the X-ray excited 1.3;10 cm/g at a wavelength of 30.9 As [8]. As-

Fig. 3. X-ray transmission (left) and luminescence (right) images of terbium loaded goat a-rabbit antibody. Antibody terbium capacity equaled 80 Tb per IgG or 6% metal by weight. (Cross hairs indicate center of X-ray beam used to collect data in Fig. 4.) M.M. Moronne / Ultramicroscopy 77 (1999) 23}36 29

k suming a density of 1.3 g/cm this gives of through the sample (92%). This means that the 1.7;10 cm\. From Eq. (3) and the measured luminescent signal is coming predominantly from values for I&> and I&, an estimated thickness (x)of the front face of the probe "lm. To obtain an 1.44 lm is obtained for the irradiated region accurate estimate of the photobleaching para- marked in Fig. 3. meters, the e!ects of attenuation are taken into To get the dose delivered to the sample, the account as follows. actual number of photons absorbed per unit mass At any depth x, a slab with an absorption coe$c- l in the "lm must be calculated. Several corrections ient V receives X-ray photons with an intensity are needed for this estimate including: " \IVV (a) the e$ciency of the gas proportional counter I(x) Ie . (5) (GPC) used for X-ray detection (66%; Kirz, per- The X-ray intensity absorbed in some thickness dx sonal communication); is (b)X-ray transmission loses in the air gap (2 mm) " k \IVV between the sample and GPC (&63% transmis- I!" I Ve dx. (6) j" s sion, 30.9 A); The dose rate in photons per unit mass per second (c) the use of He in the air gap between the zone at position x is then plate and the sample to increase the incident #ux for luminescence; I k e\IVV dx DoseRate" V "I k e\IVV, (7) (d)the use of a larger monochromator exit slit width o dx (400 lm versus 200 lm) for luminescence com- o k pared to the transmitted X-ray measurement. where is the density and is the mass absorption k o"k coe$cient ( / ). Eq. (7) may be converted to The last two factors contribute an increase of Gy/s by multiplying by 1.9864;10\/j. about fourfold to the incident intensity. Correcting Bleaching can be represented as a dose-depen- for all these e!ects gives the X-rays absorbed dur- dent change in the initial quantum yield, Q (visible ing luminescence imaging versus X-ray transmis- light photons/X-ray photon). A single exponential sion as decay process would then be given by + " " \I "%" \IR"%0!% I!" 4I!"/0.66/0.63 I!"9.6. (4) Q Qe Qe , (8) In theory, a 45 nm zone plate will produce a spot where k is the bleach constant. Thus, for any time size of 56 nm diameter. However, since a large slit t and any depth x, the luminescent rate will be width was used (400 lm), the spot size is increased " ! k \IVV k \IVV because the beam is (a) less monochromatic and (b) Lum(x, t) Q exp[ ktI e ]I Ve there is a loss of spatial coherence. The latter factor (Q for time t at position x) predominates. The demagni"cation of the exit slit width is proportional to the ratio of the distance (absorbed photon rate at x) (9) between the source and ZP (3.1 m) and ZP to image l or for multiple decay processes with a non-bleach- plane (0.0013 m). Thus, a 400 m exit slit gives able component, Q , a spot size about 0.168 lm. Spectral blurring and ! " k \IVV position scanning noise probably increase the spot Lum(x, t) I Ve size further. Therefore, for purposes of calculation a value of 0.2 lm diameter is used for the area of ; + ! k \IVV # A QG exp[ kGtI Ke ] Q!B. illumination. Combined with the 1.44 lm thickness G estimate of the probe, the dose for a 100 msec. dwell (10) time is calculated to be 2.53;10 Gy averaged throughout the thickness of the antibody. However, Finally, to calculate the luminescent rate as a func- because of the thickness of the "lm, there is a very tion of time for a thick sample, we must integrate large degree of attenuation of the beam as it passes (10) with respect to the "lm thickness and the "nite 30 M.M. Moronne / Ultramicroscopy 77 (1999) 23}36 dwell time (dwt) used for sampling so that a Q! of 0.004. This gives a total initial Q of 0.01. Correcting for the numerical aperture of the de- VI k e\IVV (H>)$UR " V tector objective lens (N.A."0.85) and the e$ciency Lum Rate (tH) dwt (H)$UR of the avalanche photodiode (&50%), the estimated initial Q is 0.2 visible light photons/X-ray. + ! k \IVV # A QG exp[ kG tI Ke ] Q?B dt dx, (11) Since the terbium antibody conjugate contains 6% G Tb by weight, about 95% of the X-rays are where j"0, 1, 2,2C of samples ("200). absorbed by carbon, nitrogen, and oxygen, i.e., Tb Fig. 4 shows the bleaching of the terbium-anti- captures only one X-ray out of twenty absorbed. If body conjugate as a function of the exposure time the terbium emission were due solely to X-ray "t to two exponential decays and a non-zero asymp- absorption by the lanthanide, then for every X-ray tote using least squares and numerical integration photon absorbed, four visible light photons would of (11). In terms of dose, the estimated 1/e compo- have to be generated. It is di$cult to see how nents for bleaching are 3.3;10 and 2.1;10 Gy single X-ray absorption events in Tb could lead to with apparent initial QGs of 0.0031, 0.0027, and multiple visible photon emissions, suggesting that secondary electrons account for a signi"cant frac- tion of the total visible light produced.

3.2. Biological samples

Fig. 5 shows some of the "rst scanning luminescence X-ray microscope (SLXM) biological images obtained indicating the feasibility of using the lanthanides for luminescent labeling. Fig. 5A and B shows 3T3 mouse "broblasts labeled "rst with a rabbit anti-actin primary antibody followed by a goat a-rabbit terbium conjugate. Using the anti- actin primary has the e!ect of labeling both monomeric and "lamentous actin and generally produces a disperse labeling of the cytoplasm as well as micro"laments. Fig. 5C and Fig. 5D shows examples of actin labeled using a biotinylated polychelate in Wi38 "broblasts (see Methods). In these two examples, the primary label is phalloidin}biotin, which binds very speci"cally to "lamentous actin. After treating with the primary, the cells were incubated with streptavidin then treated with the biotinylated polychelate that binds to unoccupied valencies on the streptavidin. Again, "laments are clearly distinguishable. Fig. 6. shows a series of images taken with increasing magni"cation and resolution of 3T3 cell actin "laments labeled with Tb loaded biotinRP. In these images, "laments are clearly evident. The im- Fig. 4. X-ray dependent photobleaching of Tb loaded goat age dwell time in each case was a 100 ms/pixel. It is a-rabbit probe. Top graph shows the luminescence signal as a function of dose "t to two exponentials. Bottom curve shows also clear that with decreasing step size more and the accumulated photons counts with exposure. Data points more image detail is visible. Signal intensities along were accumulated every 100 ms. labeled "laments range from 100 to 150 counts M.M. Moronne / Ultramicroscopy 77 (1999) 23}36 31

Fig. 5. Actin labeled samples. A and B are 3T3 mouse "broblasts labeled with a goat a-rabbit antibody Tb-polychelate conjugate; C and D are Wi38 fetal lung "broblasts labeled with Tb loaded BiotinRP. above background levels. This compares quite fa- enced by the luminescent particulates that appear vorably with typical confocal measurements, in throughout the images. Using the probes which the actual number of photons used for image rhodamine #uorescence, these particulates do not construction can be much less but are displayed at appear by light microscopy until the samples are 256 levels. Fig. 6 also illustrates a technical artifact air dried. At this juncture, we are not certain as to that has not been completely overcome with air the cause, but it may be due in part to surface dried polychelate labeled specimens. This is evid- tension related damage to the labeled cytoskeletal 32 M.M. Moronne / Ultramicroscopy 77 (1999) 23}36

Fig. 6. 3T3 cell on silicon nitride window labeled with Tb-BiotinRP. B expanded area just above center of frame A. Boxed area in B expanded in frame C. Images are 128;128;100 ms. per pixel with the indicated step sizes.

Fig. 7. 3T3 cell actin and tubulin labeled using Tb-BiotinRP and visualized by its rhodamine tracer using epi#uorescence light microscopy. structures during drying. For comparison Fig. 7 ing, which require serial ethanol}water steps with shows wet actin and tubulin specimens labeled with increasing alcohol, appear to extract the poly- the biotinRP by light microscopy giving the classic chelate and, so far, have proved unusable. labeling morphology seen with conventional or- The necessity to use dried samples, however, can ganic #uors. Unfortunately, initial attempts to use ultimately be avoided using a cryogenic stage. Al- dehydration techniques such as critical point dry- though X-ray imaging of labeled wet specimens can M.M. Moronne / Ultramicroscopy 77 (1999) 23}36 33

Fig. 8. X-ray absorption and luminescence images of Tb-BiotinRP labeled in vitro polymerized microtubules. Luminescence image is 150;150;0.1 lm per pixel with 100 ms. dwell time. be done in a silicon nitride wet cell, the doses required for luminescence imaging lead to severe sample damage at room temperature [9]. By comparison, early tests with cryogenic stages indicate that biological samples can tolerate doses of soft X-rays of 10 Gy or better when frozen at liquid nitrogen temperatures [10]. A further example of lanthanide probe labeling is given in Fig. 8 that shows the transmission (A) and luminescence images (B) of labeled in vitro polymerized microtubules. Using the data obtained from bleaching of the antibody polychelate "lm, we can estimate the Tb content per pixel in the luminescent image. For example, from the data in Figs. 4 and 2.4;10\ g of Tb will yield approxim- ately 300 ($6%) detected visible light photons for a 4.7;10 Gy exposure. From this we can generate the lanthanide density map shown in Fig. 9. In this Fig. 9. Terbium density map of labeled microtubules from Fig. 8. "gure, the peak signal corresponds to 9.9 ($0.5); 10\ g of Tb per 0.1 lm pixel. we can calculate the expected signal from an ideally labeled 24 nm diameter microtubule 50 nm in 4. Discussion length covered by a 6 nm single layer of the goat antibody probe containing 6% terbium. A 2.0;10 From the X-ray photobleaching data and the Gy exposure should yield about 1030 visible light estimated X-ray quantum e$ciency (see Results), counts. This could be improved several fold using 34 M.M. Moronne / Ultramicroscopy 77 (1999) 23}36 probes such as biotinRP that can contain &25% the resolution limit is determined by the excitation Tb by weight. Additional improvements would re- volume generated by the radiation. In the case of sult from better light collecting optics such as an oil soft X-rays, secondary electrons generated by the objective (N.A."1.4, &two fold signal increase) beam can travel some distance from where they are or by parabolic mirror designs such as that de- produced and potentially give rise to probe excita- veloped by Oxford Instruments (UK) for use in tion outside the radiation cone of the focused beam. cathodoluminescence that have almost a full 2% At present, there is little information about the collection angle. Taken together, these improve- mean free path of soft X-ray generated secondary ments could provide nearly a 10 fold safety margin electrons in biological materials. However, it has in cases where labeling is not close to ideal. been estimated that absorbed X-rays of less than The photobleaching behavior in Fig. 4 suggests 500 eV deposit their energy within a range of that multiple factors contribute to lanthanide exci- 5}20 nm in organic solids such as polystyrene tation. There are three likely relevant mechanisms [11,12]. These values are su$ciently small that the including: (a) direct excitation of the lanthanide by apparent beam spread owing to secondary elec- an N level photoelectric event, (b) direct secondary trons is not likely to compromise resolution at the electron excitation of the lanthanide, and (c) sec- 20}50 nm level. ondary electron excitation of a nearby organic #uor A potentially useful property of lanthanide labels (pAS) that can excite the lanthanide by intra- such as Tb and Eu, is the comparatively long life- molecular energy transfer. As indicated in Results, times of their excited states (100}1000 ls, [13]). the latter appears to be a signi"cant contributor to This makes it possible to eliminate essentially all probe luminescence at low-to-moderate doses of unwanted sources of luminescence other than the X-rays. P-aminosalicyclic acid when conjugated to lanthanide. Objective lenses, support "lms such as DTPA-Tb is an e$cient sensitizer of Tb lumines- formvar, and bio-organic material in general, gen- cence [7]. Secondary electrons generated in the erate some background luminescence when irra- probe by the incident X-ray beam can give rise to diated by soft X-rays. (Background count rates in pAS excitation that is rapidly transferred to the the experiments reported here were typically lanthanide. A direct hit on pAS by an X-ray, how- 400}500 cps.) However, the #uorescence lifetime of ever, is likely lead to photolysis and a gradual such materials will usually be under a microsecond. reduction in the Q of the probe. The fact that at By suitable X-ray pulsing and acquisition delays, least two exponentials are required to "t the all but the long-lived lanthanide luminescence can bleaching data suggests that even after pAS photo- be e!ectively blocked. This approach has been suc- lysis, the remaining organic material can provide cessfully used to image lanthanide labeled biolo- secondary electrons that excite the lanthanide, but gical specimens with signal-to-noise ratios that are are degraded by continued X-ray exposure. The about 400 times better than comparably labeled importance of pAS to probe brightness is further #uorescein specimens [14]. The long lifetimes of the supported by the recent preliminary "nding that lanthanides, however, can be a disadvantage for inclusion of two pAS ligands per DTPA in the rapid sample imaging. In order to preserve resolu- polychelate increases the luminescence yield several tion, mixing of the afterglow from a previously fold with X-rays (data not shown). This "nding illuminated spot with a newly irradiated pixel must could prove very useful in the development of be prevented. This can be done by imposing ac- future polychelate probes with higher quantum quisition delays between the sampling of each pixel e$ciencies. at the cost of longer image acquisition times. The fact that secondary electron excitation ap- Using di!erent lanthanides such as Tb and Eu, pears signi"cant raises an issue regarding the limit- which have very distinct and easily resolved emis- ing resolution achievable using luminescent probes. sion lines, also makes it possible to do simultaneous It was stated earlier that for a scanning system the multiple label imaging. This will be a great advant- resolution is determined by the spot size of the age since the high resolution provided by SLXM illumination. However, a more precise view is that will be best appreciated in studies of the functional M.M. Moronne / Ultramicroscopy 77 (1999) 23}36 35 geometry of subcellular organelles that are at pres- probe design in a few minutes. The ideal micro- ent too small to resolve without resorting to thin scope con"guration would also include a cryogenic sectioning and other preparative demands of elec- tilt stage for full 3-D reconstruction of specimens tron microscopy. In this context, McEwen et al. labeled with multiple probes. [15] showed that X-ray-based 3-D tomographic reconstructions are feasible at 50 nm resolution us- 5. Conclusion ing dose fractionation so that sample irradiation is kept within tolerable limits ()10 Gy with cryo). In this paper we have demonstrated that lan- The same dose fractionation principle applies for thanide polychelates have the requisite brightness the case of luminescent labeled samples. Therefore, and radiation stability to form the basis of a new it should be possible to perform 3-D reconstruc- class of X-ray excitable molecular probes that can tions on multiple probe labeled specimens without take advantage of the high-resolution capability of sectioning at 50 nm resolution. scanning X-ray microscopes. In contrast to conven- Although the potential advantages of SLXM are tional organic dyes, lanthanide polychelate clear, for biological application certain practical complexes were shown to yield strong luminescent considerations must be taken into account. At pres- signals with doses of soft X-rays exceeding 10 Gy. ent using the NSLS STXM to collect even the On the basis of X-ray photostability tests, calcu- modest signal intensities shown in Figs. 5 and 6, lations indicate that doses in the range of 10 Gy and 8, very long dwell times must be used should be capable of producing speci"cally labeled (*100 ms). For a 256;256 pixel image, acquisi- images with signal-to-noise ratios comparable to tion times close to 2 h are required. Lengthy imag- good visible light CCD microscopes, but at 5 times ing times could in part or in whole be overcome by better resolution. For practical application, how- amplifying the labeling using multiple sandwiching ever, brighter X-ray sources or probes with im- passes of the probes or enzyme based ampli"cation proved luminescent yields will be required for methods such as that used by de Haas, et al. [16]. productive biological applications because of the However, this increases the risk of non-speci"c long acquisition times necessary with currently background staining and a potential loss of resolu- available scanning X-ray microscopes. tion. A better alternative is make a brighter probe with a higher quantum e$ciency. As indicated earlier, initial photon yield measurements for Acknowledgements a probe containing two pAS residues per chelate site, increase probe brightness by several times. An The author gratefully acknowledges the invalu- additional pAS residue could be substituted giving able discussions held with Prof. Robert Glaeser three donor pAS rings per chelate and very likely (UC Berkeley), and Dr. Jon Nagy (LBNL) through- improve the brightness of the probe even further. out the course of this work, and the expert technical Labeling with lanthanide loaded latex spheres is help of Ms. Donna Hamamoto and Mr. Peter Gin also a possibility [1]. In principle, a 20}30 nm lan- (LBNL). The author further acknowledges the sup- thanide loaded latex sphere could bind thousands port by the US. Department of Energy, O$ce of of lanthanides to a target site; however, as with gold Health and Environmental Research, and the Of- particles of this size, good labeling can be di$cult "ce of Basic Energy Sciences under contract num- to achieve because of cell penetration and steric ber DE-AC03-76SF00098. In addition, this work hindrance problems. would not have been possible without the help of The alternative approach to obtaining faster Profs. Janos Kirz and Chris Jacobsen of Stony acquisition times is to operate a SLXM using Brook and other the members of the X1A beamline a brighter third generation X-ray source such as the team including Sue Wirick, Xioadong Zhang, Advanced Light Source (ALS) in Berkeley. A dedi- Shawn Williams, and Steve Wang. Data was taken cated biological instrument at the ALS could using the X-1A STXM developed by the group of provide high-resolution images using the current Kirz and Jacobsen [17,18], with support from the 36 M.M. Moronne / Ultramicroscopy 77 (1999) 23}36

O$ce of Biological and Environmental Research, [6] A. Can", M.P. Bailey, B.F. Rocks, Analyst 114 (1989) 1409. U.S. Dept. of Energy under contract DE-FG02- [7] M.P. Bailey, B.F. Rocks, C. Riley, Analyst 109 (1984) 1449. 89ER60858, and the National Science Foundation [8] B.L. Henke, E.M. Gullikson, J.C. Davis, At. Data Nucl. Data Tables 54 (2) (1993) 181. under grant DBI-9605045. The zone plates were [9] S. Williams, X. Zhang, C. Jacobsen, J. Kirz, S. Lindaas, developed by Steve Spector and Chris Jacobsen of J. Van't Hof, S.S. Lamm, J. Microscopy 170 (1993) 155. Stony Brook and Don Tennant of Lucent Tech- [10] G. Schneider, B. Nieman, P. Guttmann, D. Rudolph, nologies Bell Labs [19], with support from the G. Schmahl, Synchrotron Rad. News 8 (3) (1995) 19. NSF under grant ECS-9510499. [11] J.C. Ashley, C.J. Tung, R.H. Ritchie, IEEE Trans. Nucl. Sci. NS-25 (1978) 1566. [12] J.C. Ashley, IEEE Trans. Nucl. Sci. NS-27 (1980) 1454. [13] E. Soini, T. LoK vgren, CRC Crit. Rev. Anal. Chem. 18 (2) References (1987) 105. [14] L. Seveus, M. Vaisala, I. Hemmila, H. Kojola, G.M. Roo- [1] C. Jacobsen, S. Lindaas, S. Williams, X. Zhang, J. Micros- mans, E. Soini, Microscopy Res. Tech. 28 (2) (1994) 149. copy 172 (1993) 121. [15] B.F. McEwen, K.H. Downing, R.M. Glaeser, Ultramicros- [2] J. Kirz, C. Jacobsen, M. Howells, Quart. Rev. Biophys. 28 copy 60 (3) (1995) 357. (1) (1995) 33. [16] R.R. de Haas, N.P. Verwoerd, M.P. van der Corput, R.P. [3] H.N. Chapman, J. Fu, C. Jacobsen, S. Williams, J. Micros- van Gijlswijk, H. Siitari, H.J. Tanke, J. Histochem. copy Soc. Am. 2 (2) (1996) 53. Cytochem. 44 (10) (1996) 1091. [4] M.M. Moronne, C. Larabell, P.R. Selvin, A. Irtel von [17] C. Jacobsen, S. Williams, E. Anderson, M.T. Browne, C.J. Brenndor!, Microscopy Soc. Am. 52 (1994) 48. Buckley, D. Kern, J. Kirz, M. Rivers, X. Zhang, Opt. [5] A. Irtel von Brenndor!, M.M. Moronne, C. Larabell, Commun. 86 (1991) 351. P. Selvin, W. Meyer-Ilse, Soft X-ray stimulated high res- [18] X. Zhang, H. Ade, C. Jacobsen, J. Kirz, S. Lindaas, olution luminescence microscopy, in: V.V. Aristov, A.I. S. Williams, S. Wirick, Nucl. Instr. Meth. A 347 (1994) Erko (Eds.), X-Ray Microscopy IV: Proceedings of the 431. IVth International Conference on X-ray Microscopy, [19] S. Spector, C. Jacobsen, D. Tennant, Process optimization Bogorodski Peschatnik Publishing Co. Chernogolovka, for production of sub-20 nm soft X-ray zone plates. Russia, 1993. J. Vacuum Sci. Tech, 1997, in press.