© Biodiversity Heritage Library, http://www.biodiversitylibrary.org/; www.zoologicalbulletin.de; www.biologiezentrum.at Bonn. zool. Beitr. Bd. 46 H. 1-4 S. 203-231 Bonn, Juni 1996 The subspecific status of European populations of the striped field mouse Apodemus agrarius (Pallas, 1771) based on morphological and biochemical characters Axel Hille & Holger Meinig Abstract. Patterns of geographic variation in 13 populations of Apodemus agrarius from Kaliningrad (GUS) to Macedonia were investigated by means of skull morphology (14variables)andinasubset of4populationsbyelectrophoresis (44enzymes encodedby 57 gene loci). Genetic distanceanalysis ofbiochemical data failed to indicate clusters of populations differentiatedatthe subspecificlevel. Morphological differences weremainly size-dependent. Linearskulldimensionscouldbeattributedtonon-genetic,environmental adaptationswiththeexclusionofmolarswhichseemtoberelativelyinvariableagainsten- vironmental conditions. Selective constraints to modify parts of the dentition seem to require stronger changes in the genetic program that mayvarybetween different popula- tionstoalowdegree.Lookingatallresults,A. a. kahmannishowsconvergentsizerelation- ships toA. a. istrianus. A. a. kahmanniis ingeographic contact with populations ofthe nominalrace,anditslargercranialproportionsarepossiblyaresultofclinalsizevariation. By contrast, A. a. istrianus is geographically isolated and appears to establish specific genetical characteristics as expressed by ahighly significantlyreduced heterozygosityand morphological features similar to those ofA. a. kahmanni. Key words. Mammalia, Rodentia, Apodemusagrarius, subspecies, geographic variation, Europe, craniometry, Multiple Group Principal Component Analysis, electrophoresis, genetic distances. Introduction The striped field mouse {Apodemus agrarius) inhabits a wide geographical range between central Europe in the west and China and Korea in the east (Musser & Carleton 1993). In Middle Europe three subspecies ofApodemus agrarius have been discussed: A. a. henrici von Lehmann, 1970 from Germany, regarded by some authors (e.g. Böhme 1978) as a synonym ofA. a. agrarius, A. a. istrianus Krystufek, 1985 from Slovenia, and A. a. kahmanni Malee & Storch, 1963 from Macedonia. WhileA. a. kahmanniis regarded as valid by most authors (Böhme 1978, Kahmann &Einlechner 1992), the status ofA. a. istrianuswas recentlyquestioned byKahmann & Einlechner (1992). A. a. henrici was described from Germany (v. Lehmann 1970). Although we had no material from the type locality ofA. a. agrarius in Russia, we follow Böhme (1978) insynonymizing henriciwith agrarius. A. a. istrianusoccurs in Slovenia and NE Italy (Krystufek 1985, 1991, for Italy see Sala 1974 and Zulian 1987). According to Krystufek (1985, 1991, pers. comm. 1995) its populations are geographically separated by a gap from east Slovenian populations which represent A. a. agrarius. Kahmann (1961) reported on findings from Ribnica, a place right between the two current areas, but he left no voucher specimens and Krystufek (1985) could not con- firm this locality after intense collecting. Other authors, however, suggested that all © Biodiversity Heritage Library, http://www.biodiversitylibrary.org/; www.zoologicalbulletin.de; www.biologiezentrum.at 204 A. Hille & H. Meinig A. agrarius from the area of former Yugoslavia and NE Italy should be referred to subspecies kahmanni (Djulic & Vidinic 1964, Ondrias 1966, Soldatovic et al. 1971, Kahmann & Einlechner 1992). Descriptions of subspecific divergence among populations of the striped field mouse in Europe were to a great extent based on external morphological traits, mainly differences in size. In this paper, we compare patterns of morphological differentiation among populations assignable to the 3 subspecies currently recognized to their patterns of biochemical differentiation, in order to account for genetic relationships that define evolutionary units such as subspecies. Inasmuch, we follow the concept of Smith & Patton (1988) to consider those entities to have both character (morphological and genetical) and geographic continuity as appropriate infraspecific units to be recognized in a formal taxonomy. While from the Oriental range ofthe species onlylittle karyotypic (Bulatovaet al. 1991) and biochemical data are available (Wang 1985, Zhao & Lu 1986, Liu et al. 1991), the scarce data on Euro- pean populations are widely scattered in the literature (Britton-Davidian et al. 1991; Filipucci 1992; Gemmeke 1980; Gill et al. 1987; Hartl et al. 1992; Niethammer unpubl.). But, dealing with small sample sizes, they seem not to be sufficient to fully characterizeinfraspecificgeneticvariabilityofA. agrarius. The purpose ofthis study was to assess the taxonomic status of European populations of A. agrarius at the border ofits range in western Europe. The present multivariate examination ofskull proportions in combination with a rigorous analysis ofprotein variationshould give answers whether certainpopulationgroups warrant recognition as subspecies or not. Materials and methods Morphometry Measurements: Inthecraniometricpartofthestudyweexaminedatotalof 158skulls stemm- ing from 13 populations between Kaliningrad (GUS) in the north and Lake Dojran (Macedonia)inthesouth(Fig. 1). Onlyyoungadultandadult specimens ofbothsexes (tooth- wear classes 3—5 according to Adamczewska-Andrzejewska 1973) were measured in order to reduce variance bias in size and shape introduced into the samples by ontogenetically caused variation. Thesexeswerenotseparated(populationssampledandabbreviationcodesaregiven in the legend to Fig. 1). Skullsarestoredinthe followingcollections: Zoologisches ForschungsinstitutundMuseum Alexander Koenig, Bonn (ZFMK); Senckenberg Museum, Frankfurt/M. (SMF); Slovene Museum ofNatural History, Ljubljana (PMS); Staatliches Museum für Naturkunde, Görlitz (MNG); private collection H.-J. Pelz, Münster (CP); private collection H. Meinig, Wer- ther/Westf. (CHM). 14 measurements were taken, measurements 1 to 9 (Fig. 2) with a digital calliper (Mitutoyo digimatic) to the nearest 0.01 mm, measurements 10 to 14 with a binocular (Zeiss GSZ) with an enlar—gement of50. Allmeasurements—were taken by one ofus (H.M—.). Abbreviations used are: Cbl cond—ylobasallength(1), zBr —zygomaticbreadth(2), IoC— interorbitalconstric- tion (3—), RoM rostral breadth (4), NL nasalia len—gth (5), MBr mastoid breadth (6), A—PF length of ant—erior palatine foramen (7), MxT maxi—llary tooth-row length (8), D diastema—(9), MIL lengthoffirstuppermolar(10), M—lBr breadthoffirstuppermolar (11),—M2Br breadthofseconduppermolar(12), M3Br breadthofthirduppermolar(13), ID incisive diameter (14). Statistical analyses Population genetic measures: Allelic frequencies were computed for each population derived fromindividualelectrophoretic genotypes bygene-counting as implemented inthe BIOSYS-1 © Biodiversity Heritage Library, http://www.biodiversitylibrary.org/; www.zoologicalbulletin.de; www.biologiezentrum.at — Fig. 1: Geographical—origin of the populations examined: 1 Zehlau, K—aliningrad area (KAL), GUS (7);—2 Prenzlau, Brandenburg (PRE), Germany—(12); 3 Berlin (ber), Germany (13); 4— Harz, Lower Saxony (har), Ger—many (15); 5 Görlitz, Saxony (goer), Germa—ny(18), 6 Osthessen(ohe), Germany(16); 7 —Tiszacsege, Hortobagy(tis), Hungar—y (7); 8 Radenci, Mura rijeka (rad)—, Slovenia (21); 9 Brezice (BR—Z), Slovenia (6); 10 Ajdovscina (AJD), S—lovenia (11); 11 Rovinj (rov), Croatia (15); 12 Banja Bansko (bba), Macedonia (3); 13 Lake Dojran (doj), Macedonia (13). Codes for populations studied morphologically and biochemically are given in capitals, codes for populations studied only morphologically are given in lower letters; the numbers of skulls measured are given in parentheses. © Biodiversity Heritage Library, http://www.biodiversitylibrary.org/; www.zoologicalbulletin.de; www.biologiezentrum.at 206 A. Hille & H. Meinig program of Swofford & Selander (1981); allele frequency estimates for an isofemale Fl line sample from Kaliningrad area (KAL) were corrected for introduced bias not exclusively screening for polymorphism in samples from the wild (Long 1993). The amount of genetic divergence between populations was computed by Nei's unbiased standard genetic distance D (Nei 1978). A phenogram of the genetic relationships among populations was obtained performing the unweighted pair group arithmetic average cluster analysis (UPGMA, Sneath & Sokal 1973). Standard errors on each bifurcating node were calculated as the standard deviation ofall pairwisedistances between all OTUsjoining the nodes withinthe cluster con- secutively (Nei et al. 1985). Cranial morphometric analyses: Morphological relationships among geographic samples wereassessedbyfoursubstantialtechniquesutilizingseveralstatisticalroutinesoftheSYSTAT version 5.03 for DOS (Wilkinson 1990), the BMDP-PC90 package (Dixon 1990) and the NTSYS-pc ver. 1.60 (Rohlf 1990) for IBM-compatible computers. Techniques for verification of natural groupings (in this case subspecies) should have the propertynottobebiasedbyinformationofgroupmembership, thatis anaprioriassignment ofspecimens to these groups (Humphries 1984). As an exploratorytechnique for discovering structure in data the Principal Component Analysis (PCA) is widely used in systematic studies. Here we employ Multiple Group Principal Component Analysis (MGPCA; Thorpe 1983, 1988). It provides a multivariate means to assess the within-group components of character variation when using intercorrelated linear measurements. By pooling the within- group variance-covariance matrices derived from log-transformed cranial variables it contri- butes better to among-group discrimination than ordinary PCA. The logarithmic trans- formation makes the covariance matrixindependent ofscaling ofmeasurements but standar- dizesvariancesandpreservesallometries(Jolicoeur 1963). Extractedprincipalcomponentsare interpretedaspatternsofcovariationinsizeandshape, butactuallydonotconfusethewithin- andbetweengroupdifferenceswhenseveralgroupsareused(Thorpe 1976). ThefirstMGPCA axis derived from the pooled within-group variance-covariance matrix can be interpreted as a general within-group allometric "size" vector if most of the original variables contribute withpositivesigns andequalmagnitudeto its eigenvector coefficients (Patton&Smith 1990). The first step of the procedure was the computation of character residuals from the log- transformed variables for each population sample derived from an analysis ofvariance using theMGLH routine ofSYSTAT. An ordinary PCAonthe covariance matrixoftheseresiduals produced eigenvectors to be cross-validated bymultiplyingthe score coefficientswiththe log- transformedvariables(usingSYSTAT'sweightingvariableoption). Alternatively,computation could be done using BMDP-PC90 tools. First the variance-covariance matrix was computed for each ofthe 13 groups (= populations), and thesewere pooledto produce asingle within- groupvariance-covariancematrixusingBMDPAM-module. Then fromthis matrixtheprinci- pal components were extracted by means of the BMDP4M-routine. The resulting component scores were used in bivariate plots in an attempt to separate the groups (= populations or subspecies) either "size" included or excluded (omitting MGPC-1 = "size-out" analysis). Following these latter consideration ofa "size-out" analysis (Thorpe et al. 1982), the "size- dependent" principal components (MGPC-1 and also MGPC-2) were excluded from sub- sequent analyses andthe component scores ofthe MGPCA2-14res. MGPCA3-14variates are regarded as size-independent 'characters', whichweresubjected as new variablesto adiscrimi- nant analysis to assess grossly size-free variation between populations. Individual scores on the first two canonical axes plotted against each other show size-independent shape variation among the populations. In a slightly different approach used as an independent means to subsume for effects of overall size on variation found among populations, cranial variables were first size-adjusted, using Burnaby's (1966) canonical variate analysis framework. Data were projected onto the hyperplane orthogonal spaced to the "size"loaded vector of the first principal component employing the ORTH option of the PROJ module of NTSYS. Individual scores on the adjusted principal components plotted against each other show size-independent discrimina- tion of the populations. © Biodiversity Heritage Library, http://www.biodiversitylibrary.org/; www.zoologicalbulletin.de; www.biologiezentrum.at European populations of striped field mouse 207 Fig. 2: Skull ofApodemusagrariuswiththe cranial measurements 1 to 9 indicated (measure- ments 10 to 14 not shown). For abbreviations see text. Linear Discriminant Function Analysis using the pooled variance-covariance matrix was performedtocomputethedistancesbetweendifferentsamplesmaximizingthebetween-group versus the within-group variance. It requires a beforehand allocation ofindividual specimens to one of the a priori determined groups (Neff & Smith 1979). We graphically demonstrate thedifferencesbetweenthegroups (= populations)byaNeighbour-Joiningtree(cf. Nei 1987) clustering theMahalanobis distances ofindividual canonical variable scores from group cen- troids. Findingclassificationfunctionswascomputationallyrealizedwiththe 'StepwiseDiscri- mination Analysis BMDP-subroutine 7M'. Clustering was done with NTSYS. Size and shape covary, and unless isometry pertains, such covariation implies a changing relationship between size and shape (Gould 1966). To study this finally, multivariate static allometric coefficients forthe 14 cranial variables were calculated to look at the influence of covariation of shape and form dimensions related to size differentiation (Leamy & Bradley 1982). In a first step we performed principal component analyses separately for each population sample (Smith&Patton 1988). Becausethe firstprincipalcomponent (PCI) ofour datasatis- fies interpretation as a general size factor, the position (= score) of an individual on PCI is a measure ofits overall body size, while the "raw" loadings (= elements ofthe eigenvector) © Biodiversity Heritage Library, http://www.biodiversitylibrary.org/; www.zoologicalbulletin.de; www.biologiezentrum.at 208 A. Hille & H. Meinig 0QcnQ< MM íMcqww<UmQ(JQQQ Q<W UQ MUQuOD <Q(J<i<<QQQUQQUwQ aCCX^^XxxX^^nX)X0^0~o.0_X00^0^00^-ara^"X^ar^^~^a^xa^s^—Xa-«^<Xa^"^X ©SC\aox©•0a0„DC'0a0aCE*o£a^DC PHhhhhhhhh«hHha<F'hHhhhhhhhhh<hhhhHi- hft<hhhhHhhft <Hhhh H 7D3j^^"3DoTD35^^'3o'3u'3u'3o'3o'3o3'o3'o3'o3t Uo3W^3u3d_C> D^D^D^"31o)D^17^)3duduD^"31o)d3D3DI^)D EsESaSGSEGESESá EiáEi Ei oaaaSaElaa EsEEEs £000000*00*oooíiooícdooícdooo5^ooí^oo, ooo_o*^o£^oooo£^o£^íí£5oioioioi --ÖCÖcöCCcaCCCeaCCeöCCeoCCCcdCCcöCCCCCCcaCcaCCCCRjCcöcöeaCCCC OACOdni ^ nutaes« cdcduc„d.cd . „. - ~ ~ ' c~dcdc—dc—d - - c—dc-dc~d c~d - UUwran¡ Cy Ii 33 ag E 6"¿a! 0600.. •a o t— r- Or-Hó<<CN*NCV*NHD —H;H—;'SrHñTt'"r-ij ^^~*^ ' ^<^<^n^—¡—iH—¡**i*—*ji**! < 8001-J © Biodiversity Heritage Library, http://www.biodiversitylibrary.org/; www.zoologicalbulletin.de; www.biologiezentrum.at European populations of striped field mouse 209 ofvariables on this component describe the relative contribution of each variable to change in general size, thus are proportional to allometric coefficients ofthe characters with respect to size (Bookstein et al. 1985). The first principal component of the variance-covariance matrix from log-transformed data should therefore represent some kind of an isometric size vector that can be rescaled to the length of one (Somers 1986) if covariation between the variables approach equality. Where allometry exists, it thus provides a standard measure against whichgrowth trajectories ofindividual cranial characters can be compared (Smith & Patton 1988). To "normalize" thefirstprincipalcomponenttounitywedivideditsrawloadingsbyavalue £pki2 ,/2 where k = raw loadings and p = number of cranial variables) such that their ] , squared elements sum up to unity. Then the normalized loadings were divided by 1/Vp to rescale the loadings to be expected if all dimensions (p = 14) have grown at the same rate > (Shea 1985). Resulting positive allometric variables with multivariate adjusted coefficients 1 are those that are relatively larger in large individuals than in smaller ones; negative allo- metric variables (coefficients <1) are those with the opposite relationship (Strauss & Book- stein 1982). Allometriccoefficientswereusedasnewvariablesinadiscriminantanalysis (employingthe MGLH routine of SYSTAT) that treats population samples separately. Canonical variable plots (Fig. 5) give insight into grouping patterns. Electrophoresis A total number of 53 animals were caught with snap traps at four localities (no. 1, 2, 9, 10 in Fig. 1). Tissue samples (muscle, liver, heart) were taken in the field and stored in liquid nitrogen until being returned to the ZFMK biochemical laboratory, where theywere cut into smallpieces andmaintainedinanultracold freezer(—85 °C) forlongterm storage (tissuecol- lection). M Prior to electrophoretic analysis a fivefold volume of 0.1 Tris/HCl homogenate buffer (pH 7.0) containing 0.002M EDTA and 0.05M NADP was added to the weight ofportioned tissue, eitherpureorgan specificprobesormixes frombothliver andmuscle, whichwerethen homogenized with a motor-driven homogenizer (Polytron dispenser with 12mm shaft, Kine- m—atica, Switzerland) keeping samples cool in an ice-bath. Homogenates were shaken with 0.1 0.2ml Toluene and immediately centrifuged for 10 minutes at 13.000g (Biofuge 13, Her- aeus-Sepatech, Germany). Theclearsupernatant(25ju\persample)wastransferredonto—Micro TestTissueCulture Plates (COSTAR, Cambridge; Greiner, Germany) andrefrozenina 20 0 freezer until electrophoretically processed. WeemployedtheproceduresofverticalstarchgelelectrophoresisfirstdescribedbySmithies (1955) and recently reviewed in Geiger (1990), who also gave details due to technical novelties and apparative equipments. Starch gels are made in concentration of 12 °/o and 12.5 °7o (w/v) starchingelbufferusing BIOMOL starch(Hamburg, Germany; Tab. 1). Handling andprepa- ration of gels follows the outlines made by Murphy et al. (1990). Sample application in the ve—rtical apparatus is done by means of an Eppendorf comforpette pipetting amounts of 5 10//1 per individual into a preformed slot (20 in total) in the gel, which is then sealed by moltenvaseline. Gels wereelectrophoresed overnight(16h) at 3—4V/cmina4°Ctemperated freezer, the gels additionly connected to an cooling system with cooling plates. Each gel was mm then sliced into 1.2 thick slabs for histochemical overlay-staining adopting the visualiza- tiontechniquesasdescribedbyAyalaetal. (1972), Catzeflisetal. (1982), Filipuccietal. (1987), Harris & Hopkinson (1978), Hartl & Höger (1986), Seiander et al. (1971) and Shaw & Jain (1970). 44enzymesandgeneralproteinsencodedby57presumptivestructuralgenelociwereexami- ned for all populations. Electrophoretic running conditions, separation buffer systems used, enzymes assayed and theirtissue sources are listed inTab. 1; although no progeny testing was routinelydone (withtheexceptionoftheKaliningradareasampleKAL) to confirmthe mode of inheritance of allozyme variants, resulting zymograms generally conformed with simple patterns ofcodominant Mendelian inheritance, so that genetic interpretation ofbanding pat- ternscouldeasilybedonebasedonprinciplespublishedbyCsaikl(1985), Harris&Hopkinson (1978), Hartl et al. (1988), Richardson et al. (1986) and Seiander et al. (1971). Designation of © Biodiversity Heritage Library, http://www.biodiversitylibrary.org/; www.zoologicalbulletin.de; www.biologiezentrum.at 210 A. Hille & H. Meinig encoding loci and allelic variation of the allozymes are as follows: Genes are symbolized by italicizing the enzyme and protein abbreviation of Table Í; numerical suffixes distinguish amongmultiple zones ofcathodal or anodal or bothactivities on certainzymograms inorder ofdecreasing mobility from the most anodal one considering anodal migration first; electro- morphs (interpreted as alleles)weregivenlettersinalphabeticalorder, arbitrarilystartingwith the one that migrated the least to the anode (anodal migration) or the least to the cathode (in case of cathodal migrating) under standard electrophoretic conditions as described here (Tab. 6). General statistical tests ModifiedMantel's(1967)randomizationtestinamultipleregressionandcorrelationextension was used to test for matrix associations between genetic, geographical and morphological distances among the four populations KAL, AJD, BRZ and PRE, wherethe distances in one matrixareregressedonthedistancesintheothermatrices (Manly 1991). Significanceofcorre- lations betweengeographicand morphometric distance for all 13 populations inthemorpho- metric study were tested with ordinary Mantel analysis (1967). Results and discussion Craniometric analyses Variation of single variables Coefficients of variation evidence very low intra-populational differences. The banding diagram (Fig. 3) shows values as low as0.018 for Cbl in sample Lake Dojran (doj) and a higher value of 0.052 in Osthessen (ohe). As a representative of tooth variables M3Br ranges from 0.052 in Kaliningrad (KAL) to 0.099 in Osthessen. The diagram shows no disruptive geographical trend due to a characterization of certain populations. Variation in size Condylobasal length (Cbl) and zygomatic breath (zBr) can be considered the most useful single indicators of overall cranial size among the variables examined. They © Biodiversity Heritage Library, http://www.biodiversitylibrary.org/; www.zoologicalbulletin.de; www.biologiezentrum.at European populations of striped field mouse 211 are highly correlated with the other skull measurements (less with dentition varia- bles; Tab. 2) and have low within-population coefficients of variation (Fig. 3). For mm example, condylobasal length means range from 21.46 in population Harz, Ger- mm many (har) to 24.75 in population Lake Dojran, Macedonia (doj), representing a 13.3 % difference among localities. Although our study is faced with a relatively low degree of variability (Tab. 2), multiple group principal component analysis was effective enough to discriminate between minor morphometrically mensurable differences in cranial size and shape. In order to analyze size variation among populations in a multivariate treatment, the first two multiple group principal components from the pooled within-group characterrelationships canbeconsidered as general size factors, since allvector coef- ficients are positive (tooth variables excluded) and show correlations with the origi- nal log-transformed character values (Strauss 1985). The correlation between Cbl, for example and MGPC-1 is 0.936. Communalities ofthe variables that are the pro- portions ofvarianceaccounted forbythetwomain factorsaregiveninTab. 2. Linear skull measurements and tooth variables show almost complete loadings on both components. To investigate the relationships in the craniometric variables on their own, Table 3 gives the loadings for the three vectors, together with the percentage variation they express (cf. Thorpe & Leamy 1982). The first multiple-group principal component % accounts for 36.48 of the within-group variation across the entire sampled range of A. agrarius in Europe, the first three components account for 69.82 °7o of total variance. MGPC-1 is the largest (36 °7o) and is equallyloadedin magnitude with con- tributions of the cranial variables ID, D, APF, RoM, NL, Cbl and zBr, but inverse Table 2: Pearson product-moment correlation coefficients between the mean log -transfor- 10 med cranial variables for the 13 population samples ofA. agrarius, their scores on the first three Principal Component axes extracted by a multiple-group PCA and communality of variables on the first two components (see text for details). character PC-1 PC-2 PC-3 communality log Cbl 0.936 0.695 0.176 0.863 log zBR 0.885 0.674 0.232 0.627 log Ioc 0.554 0.554 0.457 0.054 RoM log 0.854 0.658 0.163 0.621 NL log 0.861 0.643 0.055 0.616 log MBr 0.736 0.708 0.296 0.288 log APF 0.850 0.624 0.128 0.569 log MxT 0.508 0.786 0.505 0.362 D log 0.908 0.615 0.012 0.727 log MIL 0.299 0.563 0.846 0.088 log MlBr 0.349 0.649 0.524 0.228 log M2Br 0.102 0.779 0.393 0.625 log M3Br 0.015 0.825 -0.003 0.904 eigenvalue 0.002 0.001 0.001 °7o explained variance 36.48 23.26 10.08 1 1 © Biodiversity Heritage Library, http://www.biodiversitylibrary.org/; www.zoologicalbulletin.de; www.biologiezentrum.at 212 A. Hille & H. Meinig — o o o Q rt tvdio- Oi/o cOo ovq es eqs OOvso Ovo q VO o ts i _*cp m rs 03 P * O O O o 3Br ooooo cOOo rOma O<</Sso o oOvoo 0r-0- ceso cOo 0ees0s Oo eOsS o 5 w),»-k +eOo1sS +eOo1sS +vVo1Oo +1 P -GO o 2Br CoO oo VoO co vo oo Ovo VO oo Os o o o O oVO oON d or- Ooo es Ovo iBr CooO or- oot-- Ö cÓ«oo VocOo ot(-S- ccOoo o o00 o«qO Ó O©s co r- <ON roa o Ö Ö Ö 1§— oooo eos vrVaOo <Oe/so VOvOo «o oVO VoOO oe«so' o«ooo oq«o VTOT eOs Itu I +rOO1aS +o1 +Ori1-n O Ö Ó o o Q Ocso ^VoOo vq ö rco- Orr-- q OOs OrO-s eOs VO ÖOOs Ocso '"H rt cOo 0<ON0 Ovo 0~0* eOoso coOoo oeoss qvo q,—' cOOos OoS eO0s0 cvooq Noo o^o mh rNí M H H Ö Ö O O o P<F rv--<q <fN- V0o0O ccooo o OO rq-- ce^os es Ö c^o VOcOo co &1 O O [Br o(CNO VoqO o 0oV00 ovOo Ooo Ovcoo Övo cOI/oO OOuso 0eO0s VoqO o NL ra ra OeosS r- ra OÓn teó-s- VÓO oOOs oro-~ 0cr0ot +Oe—1s( +Öe-1sH +OÖe1s +Öe1s O RoM COSs1 ,o— cOo CeOSs r-1 OOOss TOti>- Ov00o oeos1 coo ccoo OVO *-< O O o oc OO ro"Taí O«¿o o Oeoss cOo cOo es eos VOO Oco O 3h o Om ^O- ^o -Osf o fi +1 +1 +1 +1 BNr cOooS o«o c«Ooo eOs ooo Or0-0- Ovosq OOVs0 Oos eos eios> Oeoss oOoos T</To rosn 4^j! (oSs o ui—l IT) co >rO/-o OOco r- O vqo oOo o OóoS 0Oo0s oooo q ñ03 (U Ö .Na pop. AJD rov BRZ ber har PRE goerKAL <OD Í°Prad bba £CU P0Q) ¡ 1 i¡