© Naturwiss.-med Ver Innsbruck; download unter www.biologiezentrum.at Ber nat.-med Verein Innsbruck Band 86 S 271 - Innsbruck Okt 1999 Inheritance of Opérant Learning Performance in Mice *) by Johannes H SCHRODER Michael BORNHAUSEN & Renate SIEGMUND **) Die Vererbung der Fähigkeit von operanten Lernen bei Mäusen S y n o p s i s : To assess the inheritance of differences of opérant learning ability in two inbred mouse strains (CBA = "bad learners" versus 102 = "good learners"!, groups often male or of ten female subjects of altogether 12 generations were tested simultaneously in individual standardized lever boxes The animals were required to press the lever for food reward in an automated nocturnal 15-hour contingency subdivided into alternating 30 minutes ON-phases an 60 minutes OFF-phases An additional food pellet was presented at the start of each ON-phase Data on the number of lever presses, rewards and errors from both sexes of the two strains, their Fj and F hybrids, and respective baekcross generations were obtained in eight consecutive test sessions These results and their ratio (z-values) demonstrated a striking pattern resemblance of performance levels in the parental generations of CBA- and 102-strains as well as in the subsequent generations Hence, it is concluded that long-term opérant performance studies yield reliable data for a quantitative approach in behavioral genetics Since a linear model (Y - a + bX) seemed to fit best the experimental results, the intercept (a) and the slope (b) of the performace curves in all 12 generations groups were determined for both sexes The slope (b) of the learning curve revealed a rather good measure for the opérant learning performance and therefore was analysed in more detail for both learners and nonlearners A unidirectional additive-dominance model with no non-allelic interactions was found to describe best the inheritance of b Herilabililies varying between 0.0118 and 0.76 favour ihe view thaï opérant learning lielungs to ine c h a r t e r * with a medium to relatively high reproductive fitness With respect to the inheritance of opérant learning performance, the two parental strains under investigation differed by only independently segregating genetic unit The genotpic variance abounted to 60 and the environmental variance to 40 9r of the phenotypic variance Introduction: There are few experimental results concerning the inheritance of learning ability and learning performance in mammals (FESTING 1973a b, LASSALLE et al 1979 SCHRODER & SUND 1984) As to our knowledge, nothing has been done hitherto with respect to the genetic analysis of opérant behavior performance in vertebrates This paper describes an atlempt to analyse the genetic basis of opérant learning in both sexes of the laboratory mouse {Mus musculus domesticus) Material and Methods: 2.1 Animals: To ascertain the feasibility of an analysis of the genetic basis of opérant learning, after a pilot study with six Neuherberg mouse strains (C57BL/6J/Han, BALB/c, C3H/Hc, 102 CBA NMRI) two inbred mouse strains *) In memoriam Prof Dr Dr h.c mult Konrad Lorenz on the occasion of the 25th anniversary of the Nobel pri7e award B *) Author's addresses: Profs Drs J.H Schröder M Bornhausen, GSF-Forschungszentrum München Oberschleißheim and Priv Doz Dr R Siegmund Humboldt-Universität zu Berlin Medizinische Fakultät (Charité), BRD 271 © Naturwiss.-med Ver Innsbruck; download unter www.biologiezentrum.at (102 -formerly 101-, and CBA) were selected for the hybridization experiments With the exception of the NMR1 albino strain, all others were inbred and propagated exclusively by brother x sister mating since more than 50 generations Although some of the six strains exhibited larger differences in opérant learning behavior ihan did 102 and CBA, these two strains were chosen because they did not segregate imo different color phenotypes in subsequent F2 and backcross generations The mice were bred and kept as specific pathogen free mouse strains as described previously (SCHRÖDER 1977) At weaning (21 - 28 days old) the mice were separaied by sex and housed in groups, five or fewer per standard Macrolon cage (25 x 20 x 14 cm-') Water and Altromin|R) food pellets were available ad libitum 2.2 Apparatus: To assess the learning ability in opérant learning tests, isolated test mice were required to press a lever in a Skinner box for food reward A nocturnal test session lasted 15 hours (16:00 - 07:00) with alternating 30 of light and 90 of dark phases Only in light phases, a bar press was rewarded by a food pellet, while lever actions in the dark phases remained unrewarded The mice had unlimited access to tap water Each test cage was equipped with individual cuslom-built electronic circuitry which controlled the test program and the record of data automatically Accordingly, interactions belween experimenter and subjects could be minimized during ihe tesi sessions Ten mice could be tested simultaneously (fig 1) Thus 10 males and 10 females were tested for each of the 12 different generations {cf table 1) A total of 240 mice within years were analysed in this study as described elsewhere in more detail (BORNHAUSEN et al 1980) Table I: Percent genomic composition of the 12 different generations with respect to the parental (maternal, mat., and paternal, pat.) genetic material Genome (%) Gen e ration Origin P| P, F, F,* F, F;* B^ B,* B; B3 B3* B, 102 inbread strain CBA inbread strain (102d"xCBA9)-F l (CBAtfx 102 ) - F , (M)2tfxCBAQ)-F (CBAtfx 102 9) - F , (IO2ö"xCBAQ|-F| cf x 102 ? (CBAtfx 102 ) - F * a " x l Ç (102c/xCBA9)-F x 102 a1 (102cfxCBA9)-F| tfxCBAÇ (CBAcf x 102 ) - F , *a"xCBA9 (!02 tf x C B A ) - F , x CBA cf 102 CBA 100 (50 pat + 50 mat): Y 50 pat: Y 50 mat 50 pat: Y 50 mal 75 (50 mal+ 25 pal); Y 75 mai 75 pal: Y 25 pat; Y 25 mat 25 pal 100 (50 pat+ 50 mal); Y 50 mai 50 pat: Y 50 mat 50 pal; Y 25 mal 25 pat: Y 25 mat 75 mal 75 (50 mat + 25 pali: Y 75 (50 pal+ 25 mal); Y 2.3 Dala analysis: The toial number of bar presses in light (rewarded) and dark (unrewarded) phases were compared to each other using the approximation of the binomial distribution by the normal distribution with P = 0.25 and Q = 0.75 According to a formula given by SACHS (1973) viz z = (x - nP) / (VnPQ), where x = the number of rewarded bar presses and n = ihe sum of the number of rewarded bar presses + that of unrewarded bar presses, a inouse was designate a learner when the number of rewarded bar presses significantly exceeded that of the unrewarded ones i.e z > 1.65 Accordingly, all mice were distinguished.into learners (z > 1.65) and nonleamers z < 1.65) 272 © Naturwiss.-med Ver Innsbruck; download unter www.biologiezentrum.at Fig la: Skinner-box as used for the present experiments Fig b: A set of ten Skinner-boxes as used tor the simultaneous record of opérant learning performance 273 © Naturwiss.-med Ver Innsbruck; download unter www.biologiezentrum.at To score the intercept (a) and the slope (b) of the learning curve, each test mouse had to run through successive test sessions which were at least 24 hs apart Accordingly, each animal under investigation had to pass 120 hours of testing within the Skinner box Using a regression analysis computer program provided by STATGRAPHICS, version 2.1 the learning curve was computed from the individual z-values of the successive sessions Because the linear model ( Y = a + bX) best fitted the experimental data, all other models (e.g reciprocal, exponential and multiplicative ones) were rejected From the indiv idual parameters (a and b) of each of the 10 mice of either sex inanvofthe 12 different generations, arithmetic mean (xj, standard error (SE), standard deviation (SD), and coefficient of variation (CV = SD/x) were calculated by the help of a respective STATGRA-PHICS program (cf, table 2) The mean parameters were compared to each other both between males and females within the same generation and between the 12 different generations of either sex using a standard (-test (SACHS 1973) Apart from P2 no significant sex differences were found within the same generation Because of this, the weighed means of males and females of the same generation could be pooled to compare them with those of other generations (table 2b : table 4) Altogether 256 double (dual) comparisons were adjusted by the means of multiple comparisons (Bonferroni-Holm procedure; HOLM 1979) Table 2a: Arithmetic mean (x) standard error (SE), standard deviation (SD) and percent coefficient of variation (CV] of the parameter b of the linear model (Y = a + bX] for males and females of the 12 different generations Generation SE SD CV (%) 1.02 0.28 0.88 86.27 0.65 0.15 0.49 0.95 0.32 0.97 ' 102.11 114.29 0.67 0.29 0.92 137.31 29.91 1.60 0.27 0.85 53.13 19.35 1.47 0.49 1.56 106.12 0.77 50.00 1.01 0.21 0.64 63.37 1.00 108.70 1.24 0.22 0.71 57.26 0.34 1.09 134.57 2.09 0.28 0.80 38.28 0.39 1.16 81.12 0.43 0.13 0.41 95.35 1.66 0.52 1.64 98-80 1.05 0.18 0.58 55.24 1.19 0.27 0.86 72.27 1.72 0.18 0.55 31.98 X SE SD CV ( * ) Pi P; 1.50 0.29 0.92 61.33 1.20 0.27 0.87 22.50 Hi F, * 0.92 0.39 1.24 134.78 0.84 0.30 0.96 F: 1.61 0.15 0.48 F;* 2.17 0.13 0.42 B, B,* 1.54 0.26 0.92 0.32 B-, 0.81 B3 B} * 1.43 B4 Table 2b: Variance« of b Vp Vc VF = - phenotypic variance gcnotypic variance environmental variance Vp VF = - VG + V E VP| + VPI + VF, + VF, * (only environmental variance possible) - 37.37 8.S3 22.45 5.32 vG - V E in % of V P : V GG in 274 (if VP: (Learners; + Nunlearners) (Learners) (Learners +• Nonlearners) (Learners) 39.93 (Learner« + Nonlearners) 60.07 (Learners + Nonlearners) 39.75 (Learners) 60.25 (Learners! 0.75 © Naturwiss.-med Ver Innsbruck; download unter www.biologiezentrum.at To determine whether or noi an additive-dominance model fits the results to be presented, the scaling test as proposed by MATHER (1949) and MATHER & JINKS (1970 pp 71 - 73) was applied The number of independently segrgating units was estimated according to the formula n = V 1/2 (P, - P2) / '/2 (P, - P : ) - e in which P, and P\ are the arithmetic means of the parental generations and e is the mean standard error of the non-segregating generations P| Pi, and F, (cf table 3) Estimates of the additive, dominance, and interaction parameters (cf table 4) were carried out according to the formalism given by JINKS & JONES (1958) and MAI HER & JINKS (1970 p 70) The heritability of b(cf table 5) was estimated using the formulae provided by MCCLEARN & DF.FRIES (1973) and WHITNEY el al (1970) Table 3: The scaling tesi according to MATHER (1949) and MATHFR & JINKS (1970) and the number of independently segregating uni[>; Nonlearners + learners Parameter *) ± Standard deviation A Learners' -0.13 ±3.14 0.08 ± U 0.42 t 2.95 -0.08+1.64 3.03 + 5.76 I.I9 + 2.93 * Formulas were used as follows: ^ - i gỗ p p B = BC) - Pi - F,; C - FT - F | - P7 - P7: VA VB = VBJ + Vpj + VF,; 4VËc;-t-Vp;; - V F : 16VFT+4V F, +VP| + \/pT in which P7- P^.FJ" FT, BC^ and BC^are the arithmetic means of P, P2 F, F, BC, and BC3 while Vp^ VPT VF[ V h VËq and VBÖJ are the respective variances F, and F, values are the weighed means of o" + ? of F, + F,* and of Fi + Fi* respectively Number (n) of independently segregating units where e = mean standard error of the non-segregating generations (P, P2 F,) n _ 0.97 (Nonlearners + Learners) n = 0.84 (Learners) Resulls: 3.1 Genomic composition of the 12 different generations: Table concerns the percent genomic composition of the 12 different generations with respect to the paternal (pat) and maternal (mat) genetic material of the original parental strains 102 and CBA inlcuding the origin of the Y-chromosome Summing up the different backcross generations B, B,* B , B ? B,* and B we have to distinguish between two different types: Backcrosses with an average of 75 % 102 + 25 % CBA genomic material (BC,) and those with 25 % 102 + 75 % CBA (BC2) The pooled data of males and females of BC, contain the weighed means of B,, B,* and B2, while those of BC2 comprise the weighed means of B ì ( B,* and B (figs and 2) 275 © Naturwiss.-med Ver Innsbruck; download unter www.biologiezentrum.at Table 4: Estinates of the additive, dominance, and interaction parameters according to JINKS & JONES (1958) and MATHER & JINKS (1970) Parameter ± Standard deviation Nonleamers + learners Learners m 4.73 + 7.14 3.92 + 3.84 |d] ±0.88 0.30 ±0.54 [h] - Ì 13.58 - ±7.17 (i] - 2.74 = 5.54 -1.18 ±2.81 Ul - 5 ±3.69 0.16±2.11 |l| 2.45 £ 8.67 I.18±4.66 The magnitude of effects ofnonallelic interactions on the means can be estimated by the parameters m [d], [h] [i], |j] and (I] Because these estimates of the interaction parameters are within the order of magnitude of their standard deviations, there is no evidence ofnonallelic interactions for slope Heterosis will occur only when [h] is negative and greater than [dj [h] is not greater than [d] in the present study The parameters of the six calculated experimental generations P], P2 F, F BC, and BC2 are the formulas given by JINKS and JONES (1958) and JINKS (1970) m = [d] - I /:P7+1/2P2+4F7-2BC,-2BC /2P7-'/2P~7 with V [d | = with Vm = 1/4 Vpj + % V P T + 16 V~2 + V ^ + V E T I/4VP[+'/4VP7 [h| - BC7 + BC^ - F^ - Fj - '/2 Fj - '/2 PT with V |h] - 36 VBq + 36 V l £ + 64 V F J + V F ^ + V P J " + V S [i] - 2B~q + B C j - F |jl = B C l " - P ] - B C j + P J with with V[l( = V i q + B v c ^ + V F J Vfj] = V E ; + V ? ; + V B C ; + V ^ [I] = P + ï ^ + 2F^ + F - B q - B C T with Vm = Vi^ + Vp^ + VF^ + I6VÊ7+ 16VRq"+ T Table 5: Estimates of heritability of b: Vr Vr heritabilityJ (h3) = — = — - — VD V^ + V c l /& (P| -P2) h- = — — Vs (P, - P3) + VE where P } and P are the phenütypic means of the parental generations, (c? + 9) and V E is estimated from the variance within the isogenic generations (P, P2 F], Fj*) h2 - 0.0049 (Learners + Nonlearners) h : ^ 0.01l8(Learners) Characiers with low heribitabilities are related to those most clo^ly connected with reproductive fitness (FALCONER 1960) , [B| - B | where B] and B2 are the phenotypic means of d" + of the backcross generations BC, \B\ - B7) + e (75 % 102 + 25 % CBA) and BO> (25 % 102 + 75 % CBA), and e is the mean standard error of ihe non-segregating generation; h - 0.41 (Learners + Nonlearners) h ! = 0.76 (Learners) According to the second formula the opérant learning performance seems to be a quantitative character of medium fitness 3.2 Significant differences of (he slope (b): The mean values of the parameter b for the linear model (Y = a + bX) as calculated from the individual z-valuesof eveTy 10 males and 10 females, respectively, are given in fig Although Ihe b's between males and females often differed significantly, higher values for males within the same generation were found only in P 276 © Naturwiss.-med Ver Innsbruck; download unter www.biologiezentrum.at P - GENERATIONS 39 — 0F> (102) tf- i f> ( ) P2 (CBA) r/ P2 (CBA) Fi - GENERATIONS J 10- I i F2-GENERATIONS -i 3210- © F; ) { F V X F ; 9) 2 * Fig 2a: Histograms of the slope (b) from (he learning curve (Y= a + bX) of males and females of the Jifferenl 12 generalions distinguished into all individuals (learners + nonlearners) and learners only, a) - tj: significant differences (p £ 0.05) (Continuation on next page) 277 © Naturwiss.-med Ver Innsbruck; download unter www.biologiezentrum.at BACKCROSS GENERATIONS © BC1:75%102 + 25%CBA -i - ja li B-,0" B-,9 BC2: 25% 102 + 75% CBA -1 - - ill 630" B39 II Ï