DEVELOPMENTAL GENETICS 13:126-132 (1992)

Rate of Phenotypic Assortment in Tetrahymena thermophila F.P. DOERDER, J.C. DEAK, AND J.H. LIEF Department of Biology, Cleveland State University, Cleveland, Ohio

ABSTRACT During vegetative, asexual reproduction in heterozygous Tetrahymena fhermophila, the macronucleus divides amitotically to produce clonal lineages that express either one or the other allele but not both. Because such phenotypic assortment has been described for every locus studied, its mechanism has important implications concerning the development and structure of the macronucleus. The primary tools to study assortment are R,, the rate at which subclones come to express a single allele stably, and the output ratio, the ratio of assortee classes. Because Rf is related to the number of assorting units, a constant Rf for all loci suggests that all genes are maintained at the same copy number. Output ratios reflect the input ratio of assorting units, with a 1 : 1 output ratio implying equal numbers of alleles at the end of macronuclear development. Because different outcomes would suggest a different macronuclear structure, it is crucial that these parameters be accurately measured. Although published Rf values are similar for all loci measured, there has been no commonly accepted form of presentation and analysis. Here we examine the experimental determination of Rf. First, we use computer simulation to describe how the variability inherent in the assortment process affects experimental determination of R,. Second, we describe a simple method of plotting assortment data that permits the uniform calculation of R,, and we describe how to measure Rf accurately in instances when it is possible to score only the recessive allele. Using this method to produce truly comparable Rfs for all published data, we find that most, if not all, loci assort at Rfs consistent with -45 assorting units, as has been asserted. o 1992 WiIey-Liss, Inc. Key words: Macronucleus, macronuclear development,

R,

INTRODUCTION For the ciliate protist Tetrahymena thermophila, phenotypic assortment refers to the appearance, in heterozygotes, of subclones expressing only a single allele. This phenomenon, also called macronuclear assort-

0 1992 WILEY-LISS. INC.

ment, occurs regardless of dominance relationships and is apparently due to the elimination of one member of the allele pair. The process results in a functionally haploid (hemizygous) macronucleus which is stable for expression of the single allele throughout the remainder of the life cycle. The germinal micronucleus, on the other hand, remains heterozygous, as shown by breeding analysis. Macronuclear assortment has been found to occur for every allele combination so far examined and is, apparently, a random, stochastic process involving many independently assorting units [Schensted, 1958; Orias and Flacks, 19751. Such assortment has important implications concerning the genetic organization of the macronucleus: any model for the molecular organization of the macronucleus must also account for the random assortment of alleles. Like most ciliates, T . thermophila cells possess a germinal, diploid micronucleus and a compound, somatic macronucleus. Because only the macronucleus is transcriptionally active, it is the macronucleus that determines the phenotype of the cell. The macronucleus develops from a micronuclear primordium at conjugation in a process that includes chromosome fragmentation, elimination of germinal DNA sequences, de novo addition of telomeres to remaining chromosome fragments, and amplification of the newly formed macronuclear chromosomes [for review, see Blackburn and Karrer, 19861. For most, if not all, loci, assortment appears to begin with the first macronuclear division after development [Doerder et al., 19771. There are two experimentally determined assortment parameters relevant to the development and structure of the macronucleus, the rate of assortment, Rf, and the output ratio. Although the output ratio is easily measured, the determination of Rf is less straightforward. Indeed, authors have used a variety of methods to calculate RP It is critical that Rf be accurately measured because it is related to the number of assorting alleles. A constant R,for all loci suggests that

Received for publication September 30, 1991; accepted November 13, 1991. ~

~~

Address reprint requests to Dr. F. Paul Doerder, Department of Biology, 1983 E 24th Street, Cleveland, OH 44115-2403.

MACRONUCLEAR ASSORTMENT all genes are maintained a t the same copy number, whereas different Rp suggest different copy numbers. In the first description of phenotypic assortment [Allen and Nanney, 19581, R, was calculated to be 0.0113 stable lines per fission. Subsequently, various authors, using different methods of calculation, reported similar & values for other loci (see references in Table 2). Schensted [1958], in a mathematical analysis of assortment, showed that Rf is related to the number (N) of assorting units by the formula R, = 1/(2N - 1). For R, = 0.0113, N is 45 assorting alleles in a G1 macronucleus. Since the average G1 macronucleus has 45C DNA, it has been assumed that the average copy number is indeed 45. However, because there has been no uniform method of calculating Rf, the hypothesis that Rf is a constant has not been adequately tested. In this paper, we replot all available assortment data and recalculate Rf in a uniform manner. Schensted [1958] also showed that the input ratio determines the output ratio of assorted classes. The input ratio is the ratio of alleles a t the commencement of assortment. Thus, for a lA:44a input ratio, 1/45, or 2.22%,of assortees are expected to express A and 44/45, or 97.78%, a. Similarly, for a 22A:23a input ratio, a 1:l output ratio of A and a expressing classes is expected. In assortment experiments, the input ratio must be inferred from the output ratio. Output ratios of 1:l suggest equal numbers of alleles a t the end of macronuclear development, whereas eccentric output ratios imply unequal gene amplification. In this paper, we reexamine the experimental measurement of R, We present a simple way to plot assortment data from which Rf can be calculated as a slope; the method is equally applicable to combinations of codominant and dominantlrecessive alleles, although the latter requires the use of the output ratio. We use this method to analyze previously published assortment data. We also present the results of computer simulation assessing the effect of sample size on the measurement of the output ratio. These results provide guidelines for future assortment experiments.

MATERIALS AND METHODS Schensted Matrix Schensted [ 19581showed that, for N assorting units, an N + 1by N + 1 matrix contains all relevant information regarding the outcome of assortment. The elements of this matrix are determined by a combinatorial function that describes the outcome of assortment for each input ratio. Assortment outcomes for any fission, n, are determined by raising the matrix to the nth power. A table of assortment outcomes based on the Schensted matrix for selected fissions has been published [Doerder et al., 19751. For this paper, values for each fission were recalculated with a FORTRAN program (using the Schensted [1958] formula and plotted as shown in Figure 1.

rn a- 5 2 3 z -1-6

127

-

-7

0

1

I

50

100

I

I

1

150

200

250

Ll

300

FISSIONS Fig. 1. Kinetics of assortment according to Schensted matrix. The natural logarithm of the fraction unassorted (ordinant) is plotted against fission (abscissa). Solid curves show assortment for both alleles considered together. Upper solid curve; 22A:23a input ratio; lower solid curve; lA:44a input ratio. Solid curves were calculated by the formula ln(1 - r) where r = fraction of stably assorted clones. In actual experiments, r = (i + j)/t where i = number of stable A clones, j = number of stable a clones, and t = total number of experimental clones. Dotted curve (22:23)and dashed curve (1:44) show alleles considered singly as a function of the output ratio (see text and legend to Fig. 8);the two dotted curves overlap such that they are indistinguishahle.

TABLE 1. R, and N Values Determined by Computer Simulations Using 250 and 30 Clones With Input Ratios of 22A:23a* No. ofclones 250 250 250 Pooled (750) 30 30 30 30 30

Rf" 0.0103 0.0094 0.0141 0.0111

Nb 49 54 36 45

0.0129 0.0103 0.0109 0.0059 0.0106

39 49 46 85 48

0.0086 59 30 0.0142 36 30 0.0067 75 Pooled(240) 0.0095 53

30

Output' A:a Partiald Rf 113:lll 117:109 117:117 347:337 A allele 0.0118 a allele 0.0105 13:15 Partial 0.0300 12:16 Partial 0.0157 18:12 Partial 0.0128 11:13 Partial 0.0095 Partial 18:9 0.0174 A allele 0.01295 a allele 0.00627 12:14 Partial 0.0118 12:16 Partial 0.0160 14:15 Partial 0.0070 1 1 O : l l O Average 0.0150

N

43 48 17 32 40 53 29 39 80 43 32 72 34

*Simulations were run with ASSORT (see Materials and Methods) for 250 fissions. These are the same simulations as are shown in Figures 2 and 3. "R, calculated from slope (regression) on fissions 80-200. bN calculated from equation R, = 1/(2N - 1). 'Number of A and a assortees a t 250 fissions when simulation was terminated. dCalculations either based on single alleles (A or a) or from region of curve between 80 fissions and first region of no change in the number of stable clones for 30 consecutive fissions.

128

DOERDER ET AL.

Computer Simulation ASSORT is a simple FORTRAN program that replicates and randomly divides assorting units as independent entities. The composition of daughter macronuclei is determined in a lottery fashion by calculating the frequency of each type of unit (i.e., expressing an allele) and generating a random number to determine which subunit to draw for the next generation macronucleus. As is shown in Figure 2, for large N, ASSORT yields assortment kinetics identical to those of the Schensted matrix. ASSORT allows the user to specify the input ratio, the number of clones, and the total number of fissions. Unlike MACREC [Doerder and DiBlasi, 19841, which incorporates unequal macronuclear division, chromatin extrusion, and recombination, all of which can affect assortment parameters, ASSORT does none of these things. With appropriate parameters, MACREC simulations closely approximate Schensted kinetics. We used ASSORT because it is simpler and faster to use.

this is shown in Figures 1 and 8. The precision of either method depends on the accuracy with which the output ratio is determined. This graphic method can be used to assess the intrinsic variation in assortment outcomes as based on simulation. A search of the assortment literature reveals that authors have used sample sizes ranging from 200 assorting clones. We used the simulation program ASSORT to demonstrate the effect of sample size on assortment outcomes of Rf and output ratio. Two sample sizes, n = 30 and n = 250 were chosen. The

0.0

-o.

5

L 8-1.5 I,

U RESULTS AND DISCUSSION 0 g -2.0 Lief (unpublished PhD thesis, University Pitts- 9 burgh, 1979) showed that, by plotting the natural log- 3 arithm of the fraction of unassorted clones against fis- f -2 . sions, a straight line with a slope equal to -R, is obtained. For a given number of assorting units, a fam- -3.0 ily of lines is obtained. Figure 1 shows two such lines 0 50 100 150 200 250 for inputs of 1144 and 22:23 as based on the Schensted FISSIONS matrix (see Materials and Methods). The equilibrium 0.0 regions (>80 fissions) of these curves are parallel straight lines with a slope of -0.0113 as determined by regression. Rf is the instantaneous rate of change of the 8 - 0 . 5 unassorted fraction, hence the straight line in a loga0 rithmic plot. *-1.0 For both of the solid curves shown in Figure 1, assortment of both alleles is taken into consideration. fi Experimentally, such curves would be obtained when & codominant alleles are assorting. However, in assort-2. ment experiments involving dominant and recessive 5 A alleles, only assortment for the recessive allele can be -2, positively scored as stable. In this instance, the output -I AAAAAAAAA ratio of stable assortees must be used in calculating Rf. -3 This can be done in two wavs. In the first. a calculated number of dominant assortees proportional to the out0 50 100 150 200 250 put ratio is added to the number of stable recessive FISSIONS assortees. For example, if the output ratio is 2A:la a t Fig. 2. Simulated assortment of 250 clones. The program ASSORT the completion of assortment, and if at a particular fission 20 of 200 clones are assorted to a, then 2 x 20 (see Materials and Methods) was used to assort 250 clones with a n ratio of 22A:23a for >200 fissions. Three different, consecutive clones would be assumed to be stable for A . The frac- input simulations are shown. Solid line is derived from Schensted matrix for tion unstable would be 1 - ([2 x 20 + 201/200). This 22A:23a input ratio. method was used to obtain points for curves shown in Fig. 3. Simulated assortment of 30 clones. The program ASSORT Figures, 4-7 and 9. The second method is to calculate a curve for only the recessive allele. To continue with (see Materials and Methods) was used to assort 30 clones with an ratio of 22A:23a for >200 fissions. Five of eight consecutive the same example, for the a curve, the output fraction input simulations are shown (three are omitted for clarity); two of those of 33%a is used to calculate the fraction unstable for a shown were selected to show extremes of assortment behavior. Solid by the expression (0.33 - >[20/200]). An example of line is derived from Schensted matrix for 22A:23a input ratio.

;

'

MACRONUCLEAR ASSORTMENT

129

TABLE 2. Comparison of Published and Recalculated R, Values Published Locus mat

TAT P-1 SerH SerT rseAl rseB

MPr

Ch3c gal rdnA (Prm) tsA

Reference" (1)Subclone a (1)Subclone b (1)Subclone c (1)Subclone d (1)Subclone e (1)Subclone f (1) (2) (3) (3) (4) (5) (6) (6) (7) (7)

(7) (7) (7)

R, nr' nr nr nr nr nr 0.0113d 0.0113 0.0113 0.0153 0.013, 0.0112 0.0114, 0.0118 0.0105 0.0087 0.0116 0.0123 0.0056 0.0095 0.0048

Recalculated RP 0.0106 0.0102 0.0168 0.0159 0.0103 0.0124 Mean a-f: 0.0127 (0.0261)" (0.0070)" 0.0101 ndf 0.0089 0.0108 0.0033 0.0127 (0.0141)g 0.0127 (0.0091)g 0.0106 (0.0084)g 0.0094 0.0063

Implied N 48 50 30 32 49 41 40 (20)

(71) 50 nd 57 47 151 40 (36) 40 (56) 48 (60) 54 79

"References: 1,Figure l a 4 in Allen and Nanney [19581; 2, Borden et al. 119731;3, Allen [1971]; 4, Nanney and Dubert [1960]; 5 , Phillips [19671; 6, Doerder 119731; 7, Merriam and Bruns [1988]. Published Rg are those calculated by the authors, except those from Merriam and Bruns [1988], which were calculated from their Table 4. bCalculated as slope (regression; all points >70-80 fissions) from plotted values as shown in Figures 4-9. "Value not reported in original source. dR, from Table 1 and Figure 2 of Allen and Nanney [19581. "Value parentheses is based on insufficient points for reliable R, fNo data in original source from which to calculate R, Walues in parentheses calculated from assortment of recessive allele.

results are shown in Figures 2 and 3 and Table 1. As expected, variation was greater with n = 30 (Fig. 3). Although n = 30 curves tended to parallel the Schensted expectation, there was notable departure from theoretical as the unassorted fraction decreased. We calculated R, in two ways (Table l). In the first, we used all data points from 80-200 fissions. This gave Rf values yielding Ns ranging from 36 to 85,with a pooled value of 53. In the second, we used data points in a n interval beginning a t 80 fissions and ending at the first instance where the unassorted fraction showed no change for 30 fissions; in assortment experiments, this criterion was often used to indicate completion of assortment. In this instance, calculated R, values yielded Ns ranging from 17 to 72,with a mean of 34. Clearly, use of all data points, regardless of the scatter, is more accurate. The output ratios were also variable, ranging from 1:l to 2:l.To show the effect of inaccurate output ratio on the calculation of assortment parameters, the 2:l ratio was used to calculate R, and N (Table 1). Whereas the use of both alleles gave a n acceptable N = 48,separate calculations for A and a yielded N = 39 and N = 80, respectively. Clearly, a n inaccurate output ratio yields a n erroneous N. As expected, assortment parameters with 250 assorting clones were considerably less variable. In this in-

stance (Fig. 21, congruence with expectation from the Schensted matrix was excellent, except near the end, when the number of unassorted clones became smaller. N was less variable, and output ratios were close to 1:l. When the three assortments were pooled to yield a single simulation involving 750 clones, &, N, and output ratio were in excellent agreement with Schensted predictions. A search of the assortment literature reveals that authors have used various methods to calculate R, In a few instances, the procedure is clearly in error, yet nearly all authors report similar R, values, implying that N is a constant, -45. Because graphic representation of the type shown in Figure 1 permits calculation of Rf as the slope (regression) of a straignt line, we have plotted published assortment data (Figs. 4-9) and recalculated R, values (Table 2). In this way there is a uniform basis for comparison to determine whether N is indeed a constant. Figure 4 shows data from the first described instance of macronuclear assortment, Allen and Nanney's [ 19581 analysis of selfers. Reasonably straight lines are delineated, particularly in regions between 75 and 163 fissions, where sample sizes are adequate. The Rfi calculated by linear regression for this fission interval are shown in Table 2. The mean

130

DOERDER ET AL.

value, 0.0127, is not much different from 0.0113 and implies 40 assorting units. The results of additional assortment experiments are shown in Figures 5-9. In those instances in which experiments were carried out over a sufficient fission interval, & and N were calculated (Table 2). In keeping with the results shown in Table 1, we calculated Rf using all data points beginning with 70-80 fissions. With the exceptions of those discussed below, resulting and implied Ns are consistent with -45 assorting units. For gal, Mpr, and ChxA loci, we also calculated Rf based on assortment for only the recessive allele (Fig. 8).The resulting Q (Table 2) are similar, but not identical, to those calculated from curves for both alleles together. Arguing from the results of simulation (Table l),this suggests that the output ratios (used to calculate R,) were not indicative of the true input ratio. In any event, the differences among calculated Ns are not large (cf. differences for mat). Most of the exceptions to N - 45 are more apparent than real. For TAT and P-1 (Fig. 5), experiments were terminated before equilibrium was reached, so there are insufficient data points for a valid calculation. For tsa, initially reported not to assort [McCoy, 19731,there may be some selection against the temperature-sensitive allele at the nonrestrictive temperature [Merriam and Bruns, 19881, and any type of selection would affect R, For rseB, there was no apparent selection [Doerder, 19731, and recalculation using only the recessive (mutant) allele (not shown) yielded no change in &. However, although assortment was slow for most of the experimental interval, there was a precipitous drop from 62% to 14% unassorted between fissions 195 and 221. This suggests that some factor affecting assortment, perhaps macronuclear DNA content (see below), was operative. In any event, the slower assortment of rseB is the only major exception to N = 45.

Merriam and Bruns 119881 reported the curious observation that rdnA alleles assort a t an & similar to that for other loci. The observation is odd because these genes, coding for rRNA, are present a t -10,000 copies per macronucleus [reviewed by Gorovsky, 19801. Re-

w

+ IT

0

ffl ffl -2

a 3

t

L

50

0

150

100

200

250

FISSIONS 0.0

z

0

H

a a

-0.5

LL D

w

-1.0

0

rn rn a

z

=2

-1.5

J

-2.0 0

I

I

I

I

50

100

150

200

250

FISSIONS Fig. 4. Phenotypic assortment among mating-type selfers. Assortment data from six assortment experiments are plotted. Data were derived from Figure 1 of Allen and Nanney [1958]. Solid line is derived from Schensted matrix for 22A:23a input ratio.

0.0

c

z 0 U

Fig. 5. Phenotypic assortment for enzyme and drug resistance loci. Square, TAT (tyrosine aminotransaminase), both codominant alleles scored [Borden et al., 19731. Triangle, P-l (phosphatase), both alleles scored [Allen, 19711. Plus sign, gal (recessive allele confers resistance to 2-deoxygalactose), both alleles as corrected by output ratio [Merriam and Bruns, 19881. Solid line is derived from Schensted matrix for 22A:23a input ratio. Dotted line is derived from Schensted matrix for 32A:32a input ratio. Fig. 6. Assortment for serotype genes. Square, SerH (H cell surface protein), both codominant alleles scored [Allen, 19711. Triangle, SerT (Tcell surface protein), both alleles scored [Phillips, 19671. Plus sign, rseA (recessive allele inhibits expression of SerH), both alleles as corrected by cutput ratio [Doerder, 19731. Diamond, rseB (recessive allele inhibits expression of SerH), both alleles as corrected by output ratio [Doerder, 19731. Solid line is derived from Schensted matrix for 22A:23a input ratio.

+ 0 -0.5 a

a LL

n W p 1 . 0 0

rn rn a

z

=-1.5 Z -1

-2.0 0

50

100

150

FISSIONS Figs. 4-4.

200

250

MACRONUCLEAR ASSORTMENT

131

0.0

0.0

z 0

2

0

u I0-0.5

H

L-0.5

a

a

a a

0

0

U.

LL

w

F -1.0

k-1.0

a

0

0 u)

ffl

Ln

cn a

a

z z J

2

3 -1.5

I-1.5 2 -I

-2.0

-2.0 0

50

100

150

200

250

FISSIONS

O

-

I

,

I

100

150

200

250

Fig. 9. Assortment for rdnA and tsa genes. Squares, rdnA (gene for 17s rRNA), both alleles as corrected by output ratio. Triangle, tsA (recessive allele confers temperature sensitivity), both alleles a s corrected by output ratio. Data from Merriam and Bruns [19881. Solid line is derived from Schensted matrix for 22A:23a input ratio.

-I

-3.0

50

FISSIONS

8

2-2.5

0

O I

i

plotting their data, we too calculate an Rf indistinguishable from other R+ (Fig. 9, Table 2). The meaning of this unexpectedly fast rate of assortment is not clear. It has been suggested that nucleoli are the assorting units [Nielsen and Engberg, 1985; Merriam and Bruns, 19881, but the number of nucleoli per cell, 500-1,000 [Nilsson and Leick, 19701, is too large to yield assortment at the observed &. However, since nucleoli have been observed to fuse during the growth cycle [Satir and Dirksen, 19711, the effective number of nucleoli

may be considerably smaller, and there is the attendant possibility of gene conversion. Many assortment curves in Figures 4-9 are not congruent with Schensted expectation. The three reasons for this require brief comment. First, as is shown in Figures 2 and 3, there is variation in assortment outcomes, the magnitude of which is dependent on sample size. The scatter among Chx assortment curves (Fig. 7) is attributable, in part, by the use of 30 clones per assortment experiment [McCoy, 19791. Second, in instances such as SerH (Fig. 6), the curve is consistent with a highly eccentric input ratio (cf. Fig. 1).Eccentric output ratios are observed with certain combinations of SerH alleles [Nanney and Dubert, 19601and imply differential gene amplification during macronuclear development. Third, in instances in which the onset of assortment appears to be delayed (e.g., TAT and P-1 in Fig. 5 ) ,the initially higher DNA content of hybrid macronuclei must be considered. Doerder and DeBault [1978] found that the G1 DNA content of newly developed macronuclei is -64C rather than the usual 45C. In strain B macronuclei, this higher value returns to 45C by 50 fissions, but, in interstrain hybrids (e.g., A x B and B x C2), the higher amount persists for 60100 fissions. As has been argued in detail elsewhere [Doerder et al., 19771, the persistence of 64C of DNA is sufficient to account for the apparently late onset of assortment. Significantly, all instances of “late” assortment involved interstrain crosses. This includes TAT, P - 1 , SerT, rseB, and Chx [Chx crosses of McCoy, 19791. To show the delayed effect of 64C on assortment, an assortment curve for 32A:32a is shown in Figure 5 . In summary, by plotting all assortment data in a uniform way, we have calculated comparable assort-

132

DOERDER ET AL.

ment parameters. Rf is similar for most loci, implying about 45 assorting units in a G1 macronucleus. When it is observed that the loci for which assortment data are available are distributed among all five micronuclear linkage groups [Bruns, 19841, this implies that the majority of macronuclear linkage groups, if not all, are maintained a t the same copy number (rdnA excepted). Although the modal input ratio appears to be 1:1, there are exceptions, which imply differential allele amplification. Instances in which there appears to be a delay in the onset of assortment involve interstrain hybrids with initially higher macronuclear DNA content. There is no compelling case for a developmental program of assortment initiation as was previously thought [Bleyman et al., 1966; Doerder, 1973; McCoy, 19791. Finally, because the assortment process is a stochastic one, there is inherent variation in Rf and the output ratio. This variation can be minimized in assortment experiments by the use of large sample sizes and by accurate measurement of the output ratio.

ACKNOWLEDGMENTS We dedicate this paper to David L. Nanney, who, together with Sally Allen, first described phenotypic assortment. Dave domesticated Tetrahymena as a genetic tool, and the phenomena he described are still rich sources of material for experimental analysis. We thank Virginia Merriam for sharing raw assortment data and Greg Lacrosse for comments on the manuscript. We also thank Dr. Ellen Simon for her help with Table 2.

REFERENCES Allen SL (1971):A late-determined gene in Tetruhymena heterozygotes. Genetics 68:415-433. Allen SL, Nanney DL (1958):An analysis of nuclear differentiation in the selfers of Tetruhymenu. Am Nat 92:139-160. Blackburn EH, Karrer KM (1986):Genomic reorganization in ciliated protozoans. Annu Rev Genet 20501-521.

Bleyman LK, Simon EM, Brosi R (1966):Sequential nuclear differentiation in Tetruhymena. Genetics 54:277-291. Borden D, Miller ET, Nanney DL, Whitt GS (1973):The inheritance of enzyme variants for tyrosine aminotransferase, NADP-dependent malate dehydrogenase, NADP-dependent isocitrate dehydrogenase, and tetrazolium oxidase in Tetrahymena pyriformis, syngen 1. Genetics 74595-603. Bruns PJ (1984):Tetruhymena thermophilu. In O’Brien SJ (ed): “Genetic Maps, Vol 3.”Cold Spring Harbor, pp 211-215. Doerder F P (1973):Regulatory serotype mutations in Tetruhymena pyriformis, syngen 1. Genetics 74:81-106. Doerder FP, DeBault LE (1978):Life cycle variation and regulation of macronuclear DNA content in Tetruhymena thermophila. Chromosoma 69:l-19. Doerder FP, DiBlasi SL (1984):Recombination and assortment in the macronucleus of Tetruhymenu thermophila: A theoretical study by computer simulation. Genetics 1081035-1045. Doerder FP, Lief JH, DeBault LE (1977):Macronuclear subunits of Tetrahymena thermophilu are functionally haploid. Science 198

946-948. Doerder FP, Lief J H , Doerder LE (1975):A corrected table for macronuclear assortment in Tetruhymena pyriformis, syngen 1. Genetics

80:263-265. Gorovsky MA (1980):Genome organization and reorganization in Tetruhymena. Annu Rev Genet 14:203-239. McCoy JW (1973):A temperature-sensitive mutant in Tetruhymenu pyriformis, syngen 1. Genetics 74:107-114. McCoy JW (1979):Variability in the timing and outcome of macronuclear assortment in Tetruhymena thermophila. Genet Res 34:57-

67. Merriam EV, Bruns PJ (1988):Phenotypic assortment in Tetruhymena thermophilar assortment kinetics of antibiotic-resistance markcrs, tsA, dcnth, nnd thc highly nmplificd rDNA locus. Gcnctics 120:

389-395. Nanney DL, Dubert JM (1960):The genetics of the H serotype system in variety 1 of Tetruhymena pyriformis. Genetics 451335-1349. Nielsen H, Engberg J (1985):Functional intron’ and intron- rDNA in the same macronucleus of the ciliate Tetruhymena pigmentosu. Biochim Biophys Acta 82530-38. Nilsson JR, Leick V (1970):Nucleolar organization and ribosome formation in Tetruhymena pyriformis GL. Exp Cell Res 60:361-372. Orias E, Flacks M (1975):Macronuclear genetics of Tetrahymna. I. Random distribution of macronuclear gene copies in Tetruhymena pyriformis, syngen 1. Genetics 79:187-206. Phillips RB (1967):T serotype differentiation in Tetruhymena. Genetics 56583-692. Satir B, Dirksen ER (1971):Nucleolar aging in Tetrahymena during the cultural growth cycle. J Cell Biol 48:143-154. Schensted IV (1958):Model of subnuclear segregation in the macronucleus of ciliates. Am Nat 92:161-170.