Analysis of inorganic fiber concentrations in biological samples by scanning electron microscopy.

in by scanning electron microscopy. j work 7 (1981) 101-108. Analyzing fibers with electron microscopic techniques involves several preparation steps, especially during the analysis of fibers in human tissue. The influ ence of these steps on the analytical result was studied in detail. Fiber number was unaffected by the mild sonication of fiber suspensions analyzed with scanning electron microscopy. Significant fiber losses appeared during the filtration of fiber suspensions through 0.8-flm pore-size Nuclepore membranes. Shrinkage of the tissue during dry ashing broke long fibers and consequently increased the fiber concentration and affected the fiber length distribution. Dry ashing, however, removed more of the organic debris than wet ashing; thus specimens prepared with dry ashing were more suitable for scanning electron microscopic analysis. A fairly good correlation was demonstrated for the analysis of fibers with scanning and transmission electron micro scopy.

The assessment of airborne inorganic fiber concentration has tradi1ti.onally been performed by optical phase-contrast microscopy. For a Ibetter identification and quantification Of inorganic fibers, electron microscopy (EM), inclUlding scanning (SEM), transmission (TEM), and se,anningi transmission (STEM) electron microscopes fi'libed with X-ray and electron microanalytical equipment, has recently been introduced.
These techniques are also used for the analysis of inorganic fibers in human autopsy matedal, especially !lung tissue. Different microscopk methods often reqmre different preparation of the tissue samples. In a recent paper by Morgan & Holmes (16) different aspects of the preparation and analysis of lung tissue are discussed. They emphasize tha't drying, 1  low-temperature ashing of the tissue and ultrasonic dispersion of the ash are contramdicated. Such teclmiques have, on the o1fuer hand, previlously -been used by other authors (3,4,7). Different preparation techniques have, however, not been thoroughly compared.
In the present investigation we have studied 'the influeIliCe of different preparation steps on inorganic fiber concentration as measured by SEM.

Material and methods
The different prepamtion steps which may lead to serious ftber losses have been tested on Union Internationale Contre Le Oancer (UICC) standard crocidoli'te as-beStos and on fibers from autopsy material of subjects who died of asbestos-related diseases.
All fibers with an aspect ratio equal to or greater than 3:1 were counted. Fiber numbers were evaluated at a magnifioation of 4,500 X iJn a Jeol J'SM-35 SEM, and in each specimen either 100 fibers or the 0355-3140/81/020101-8 fibers in 200 view fields were counted. The calculation of :relative standard deviations (RSD) were /based on four samples.
All liquids ·used in 1Jhe analyses were filtered, and the SEM samples were goldcoarted before analysis. The resolution of the SEM was -tested on standard UICC ohrysotile on a Nuelepore fnter with a pore size of 0.8 pm. The TEM analyses were peflformed in a Philips 301 TEM at a magnification of 5,000 and 20,000 X.

Sample preparation
Stock solutions of the standard UICC samples were prepared by dispersing 1 mg of asbestos in a 100-ml water/ethanol solution :in ,an ultrasonic bath. Ethanol was added Ito avoid hydrophobic areas on the Nudepore membranes during the fHtration process.
The stock solution was diluted 1:100 with distiHed water. Different volumes of this solution were filtered through the membranes iby means of a water suction pump. Particle ag.glomeration and adhesion to ithe glass walls occurred during storage, and ,a new stock solution was prepared for each series of :analyses.
Tissue samples were cut from lungs preserved in formaLdehyde and either dried to 'Constant weight at 80°C or weighed ilIl the wet 'Conditbion. For .the determination of the dry-to-wet weight ratio, adjacent tissu'e samples were dried as described pre-Vli'ously.

Digestion of tissue
The following three procedures were used to remove oI1ganic material from the tissue samples: 1. Wet digestion (WD) in 1 N sodium hypochlorite in 0.1 N sodium hydroxide. Sodium 'hypochlorite (5-10 ml) was added to -the tissue in cel1ltrifuge glasses thart were kept at 60-70°C untilla complete reaction was obtained. A mixture of the sodium hypochlor1te solution and di~yl ether (5 ml) was thoroughly shaken and thereafter centrifuged at 4,000 rlmin for 20 min. Each time the procedure was repeated 'twice, the €'ther layer being removed with a pipeMe. The ether fraction was filtered through Celas silver membranes and analyzed for fibers wiJtJh SEM. The un-reacted sodium hypochlorite was neutralized by 1 N hydrochloric acid and sonioated. In one series of analyses the ether extraotion was omitted, and the WD samples were filltered directly through the Nudepore membranes.
3. Low-temperature ashing (LTA) at temperatures below 200°C in a Tracedab model 505 LTA. Generally 50 W of forward power wi'1ih an oxygen flow of 225 ml/min was USled. Some samples were also ashed at high power LTA, ie, at an energy input greater than 150 W of forward power at an oxygen flow of 150 ml/min. A oomplete ashing was obtained for 10-40 mg of dry tissue wiJthin 2-5 h, depending on .the size of 1fue tissue pieces. 'I1he ash remaining a:liter LTA was dispersed iJn 0.5 N hydrochloric acid, a few milli!lilters of etbhanol being added, and kept in the ultrasonic bath for 5 min. The tissue salts were then rapidly dissolved by the hydrochloric acid.

Filtration
Due to 1fu.eir high filtration efficiency, porous membranes are normally used for the ,air filtra'1Jion of mkrofibers. Compared to tlle structure of plain Nudepore membranes, the porous structure of the membranes is a poor ba~kground for SEM work, hut penetra1ion of the Nuclepore may occur depending on the pore sire (4,8).
In this study we have compared 0.2-and 0.8-pm Nuclepore membranes versus 0.8-Fm Gelman Metricel polyvinyl chloride (PVC) filters using standard UICC croci-doUte prepared from water suspensions. The PVC filters were prepared for SEM according to the method published by LeGuenet al (5). Samples for the TEM analy..ses were prepared with a modified Jaffe Wick washing teohnique (2).

Ultrasonic treatment
The samples weI'e 'treated for 5 min in an U1trasonic bath. VigoTO<uS sonication may spM fibers and should be avoided. This possibility was investigated by comparing the direct :£iltration of standard solutions whi~h had been mildly sonicated in a bath (less than 0.2 Wlml) to solutions v~gorously sonicated with a probe with a power owtput greater 1Jhan 3 WlmI.

Results
Larg,e initer-and :intra:laboratory variations may occur in nber counbing ras H is based on subjective observations of the partrcles with the use of different equipment. In most papers an estimation of the coeffident of variation of the method is not ind:iJcated. For the standard UICC samples analyzed in SEM at 4,500 X a relative standard deviation (RSD) of 10 0 10 was found for repeated counts on the same samp1e. CounHng different samples prepared from >the same stock solution gave an RSD for high nber loadi'ngs (ie, grearter than 10,000 fiJbers/mm 2 ) of 16010 fOrI' tlhe Gelman fi1teI1sand 10010 for the Nuclepore membranes. For lower fiber 10ading r s the precision was even bet/trer fo,rtlhe Nuclepor,e membranes. The RSD for the tissue samples was better than 20 010, includimg intra-and interlobar fi/ber density variation (4). A rrelationship between the magnificaltion ,employed and fiber numbers has previously been demonstrated by Gylseth et a:l (4).   In table 1 the results from the comparison of the different filter types are given, and a significant higher filteration efficiency is indicated for Nuclepore 0.2-pm and PVC 0.8-pm than for Nuclepore 0.8-jlm membranes.
Tarb'les 2 and 3 show the fiber counts in two dirHererut series with different agitation meihods. There was no significant in-     crease in fiber concentratton roter the samples had been treated in an ultrasonic ba!th. Table 3   partly be avoided either by ultrasonication or by the use of a "rubber policeman" before filtration (table 3). A comparison of the different digestion procedures against direct fHtra!mon of standard UICC cI1oddolite suspensions gave a mean recovery of 85 and 90 % for LTAand WD/e1Jher extraction, respectively. HTA at 600°C ror 1 h gave a re- cO\l1ery of less than 20 0/0. When the temperature was reduced to 500°C, 'the recovery increased to 55 0/0.  Samplesashed at high-power LTA and filtered on 0. 8-,um Nu'Clepore membr.anes were compared to adjacent samples ashed at 'low-power LTA and filtered on 0.2-,um NU'clepore membranes {table 4). On the average the latter technique gave six times higher filber concentrations <than the former. WDand LTA were compared from adja'cent tissue samples (table 5). After Table 6. Fiber concentrations in adjacent tissue samples after low-temperature ashing (samples 1, 2, 3) and wet digestion (1', 2', 3') followed by two ether extractions. As pointed out by Ascroft & HeppleS'ton (1) and Morg,an & Holmes (6), fiber losses or even an iJncrease in fiber nurrib'€r may oocur during the digestion {WD) of tissue sample for quanti,ta'tive fiber estimation. Errors may also ,be iJntroduced by LTA (6).
We have employed LTA for some yea,rs, and ourexp'€rienoe is til,at, under carefully Discussion complete WD, the solutions were directly filtered onto the membranes. WD gave, on the average, 35 % lower fiber concentrations than LTA. LTA treatment of the wet-digested and SEM-mounted Nudepore membranes did not improve the resu]ts. In fact, the LTA treatment of the Nuclepore membranes gav'€ a filter residue which could be misinterpreted as fibers.
In 'a $'€Cond series of analyses of adjaC'€TIt tLssue s'amples from another lung with the ethere~traction procedure, the data given 1n table 6 were obmined; fuey gave concentrations that averag'ed 37 Ofo low'€r than those for LTA. For two corresponding samples of the same lung, the diameter and lengJth of 100 !tbers were ev,aIluated in each sample. No difference appeared between 'the Itwo 'as r,eg,ards the d1ameter diStribution, Whereas WD, in comparison to LTA gave a significantly larger proportion of Ilonger f1bers (fig 7).
A comparison of the LTA samples with SEM and TEM analysis has been per-fOl'IIled employing a magnification of 4,500 X for the SEM and 5,000 X and 20,000 X !for the TEM (table 7). The results for SEM/4,500 X versus TEMI 20,000 X ar'€ plotted on a log-log scale in  Sample oontrolled conditions, it is an effective and gentle method for the removal of organic materi,al. It entaHs low fiber losses and reprodUcible results. High plasma energies (ie, power input greater than 150 W) produce local overheating in the tissue pieces and ;introduce filber losses of significant degree. As pointed out, these effects may he avotded with low plasma energies (ie, power inpwt of less than 60 W) at an increased oxygen How (ie, approx 225 ml/min). T'he data show that LTA gives a slighitly increased fiber count when compared to WD, a finding indicating that fibers Ibreak during tissue shrinkage. The difference may, however, partly be caused by the short and ,thin fibers being covered by the organic debris left after WD. As expected, the diameter distrilbution is not affected, whereas the WD method, when compared to LTA, gives a length distribution sl,ightly biased towards longer fibers.
The whole ashed sample can be filtered with LTA, and thus any extractions introdudng fiber losses can be avoided. For SEM fiber counting i't~lso gives a better background. Furthermore, the most precise tissue mass indicator is dry tissue weight. W1henthe wet tissue is cut, weighing errors are introduced due 'to 'liquid loss. Fo,rmaldehyde also evaporates fairly rapidly, and therefore contributes to the error. Therefore the samples should be weighed in airtight beakers. Lungs which are intratracheally :£ixed contain a significant amount of liquid and have a low dry-towet weight ratio. Other lungs are infiltrated with tumors or consist of pneumonic areas of highly different density and liquid content, and therefore their weigJhtlweight ratios vary. The dry-to-wet bssue weight ratios variled from 6 to 15 0/0 in our <Study. If dry weight is used, paraffin-embedded material from retrospective studies may be analyzed; thus a comparison of data from retrospective and prospective Sltudies becomes possible.
With direct filtmtion of the suspension, fi:ber losses are Teduced when compared to those of 'the other pl'etreating s'teps. It is 'important to use membranes with a sufficiently smaU pore size to prevent fiber penetration. Our data show 1:hat 0.2-,urrn Nuclepore membI"anes give higher fiher counts than O.8-pm porous PVC membranes. The difference demonstrated may be due to -the etching of the PVC membranes necessary for the release of the fibers. Nuclepore membranes also give a better image for counting in SEM than the etched PVC membranes. A comparison of frber size distributions for 0.2-and 0.8-,um Nuclepore membranes showed that the 0.2-,um membranes contained both shorter and thinner fi'bers than the 8-,um ones. Therefore, the filtration of samples on Nuiclepore membranes should only be done with filters wtth a pore size of 0.2 ,urn or less; a compromise between filtration speed and efficiency must be made.
The use of sonication for the homogenization of fi;ber suspensions has also been discussed 'by Morgan & Holmes (6). There are two mai'n features, a homogeneous dispersion ,and a rapid dissolution of the tissue salits. Our data, along with those (jf Ohatfield et al (2) and Spumy et al (9), show that homogenization of fiber suspensions may well be performed !by ultrasonics if a sufficiently low power input is used (ie, 0.2 W/rdl or less). High-power sonication may significantly increase fiber number.
Before TEM analysis the Nudepore memlbranes ·a,re cut and transferred to TEM grids; thereafter the membrane is removed by dissolution in chlorororm. fiber losses may occur during these steps. These losses are iIndicated by the lower counts obtained faa: the TEM analysfus at a magnJificatton of 5,000 X when it is com-paTed to ,the SEM analysis a,t 4,500 X. If the TEM magnification is increased to 20,000 X, the loss is near:ly compensated for. At the same ,time the effec,ts of iner,eased magnification on the fiber number is demonstrated.
The dHferentpreparatilon methods may affect asbestos bodies differently than naked fibers. A study dealing with ,these problems is at present in progress in our laboratory.