Individual selection experiments

Three different “sets” of libraries (A, B, and C) were constructed by shuffling combinations of C2H2 ZFs derived from various subsets of different wild-type DZFs (Table 3.1). Three independent selections were then performed, each using one of the three library “sets” (A, B, and C). For each of these three selections, all possible pairwise combinations of the four libraries in a “set” were analyzed for potential interactions.

3.3.2.1 Libraries

For constructing library set A we decided to use six subdomains derived from the human Ikaros, human Eos, human Pegasus, Hunchback D.m., Hunchback L.m., and Hunchback H.t.

DZFs. Various subdomains from DZFs which are all know to interact with each other (e.g.

Ikaros and Eos) were used as internal positive controls to test if it is generally possible to create chimeric DZFs that can mediate interaction. The expectation was to re-isolate several shuffled combinations of these domains.

Since the initial characterization of the ten DZF domains showed interesting specificity patterns for Pegasus, Hunchback D.m. and Hunchback L.m. (Hunchback D.m. can interact with both Pegasus and Hunchback L.m. while these two do not interact with each other) another library set B was constructed consisting of shuffled combinations of these three

DZFs. We were interested in testing if we could enrich for chimeric proteins containing these subdomains.

To create chimeric DZF domains with additional novel interaction specificities, a third library that only contained DZF domains that do not interact with each other was designed. We reasoned that chimeric proteins consisting of these subdomains would be more likely to possess novel specificities. Therefore DZF domains from human Eos, human TRPS1, Hunchback D.m., Hunchback H.t. and Hunchback C.e. were applied to create shuffled library set C.

3.3.2.2 Selections

For library set A each pool of DNA fragments encoded 30 different shuffled DZF domains since reassembly of wild-type DZFs was precluded (6 x 6 = 36 - 6). Note that for each library, the number of transformants (10³-10⁴) exceeded the theoretical number of shuffled DZFs by ~100-fold. The potential number of possible pairwise combinations of shuffled DZFs for each selection experiment is 30 x 30 = 900. ~10⁶ transformants were plated which outnumbered the theoretical number of potential combinations by at least 1000-fold (Table 3.1). Interestingly, a large number (~1000) of surviving colonies were obtained for each selection experiment.

Theoretical # of potential DZFs in each library

Actual # of transformants

for each library

Theoretical # of combinations for each selection (library x library)

Actual # of combinations tested for each

selection (library x library) Set A:

Ik, Eo, Pe, Hd, Hl, Hh

30 10³-10⁴ 900 ~ 10⁶

Set B:

Pe, Hd, Hl 6 10⁴-10⁵ 36 ~ 5 x 10⁶

Set C:

Eo, Tr, Hl, Hh, Hc

20 10³-10⁴ 400 ~ 10⁶

Table 3.1 Library sizes and B2H selection statistics. Data in columns 1 and 2 describe the theoretical and actual sizes of each library in “Sets” A, B, and C. Data in columns 3 and 4 describe the theoretical and actual number (#) of combinations tested in each selection performed for “sets” A, B, and C. Each “set” consists of four libraries (two Zif268 fusion libraries and two RNAP alpha-subunit fusion libraries, depicted in Figure 3.5) which are used to perform a corresponding “set” of four selections that tested all pairwise combinations of the four libraries. Abbreviations are as in Figure 3.4. See text for details.

In library set B each pool of DNA fragments encoded six shuffled DZF domains precluding the reformation of the wild-type DZFs (3 x 3 = 9 – 3). The number of transformants for each library was 10⁴-10⁵ which exceeded the theoretical number by ~1000-fold. We plated ~10⁶ transformants, outnumbering the potential number of combinations (36) by at least 10000-fold (Table 3.1). A large number of colonies (~1000) survived the selections.

For library set C each fragment pool contained 20 (5 x 5 = 25 – 5) shuffled DZFs, again precluding the reassembly of wild-type domains. The number of transformants (10³-10⁴) for these libraries outnumbered the theoretical number of shuffled DZFs by about 100-fold. ~10⁶ transformants were plated which exceeded the theoretical number of potential combinations (400) by at least 1000-fold (Table 3.1). Interestingly, less surviving colonies (~100) compared to the other selection sets were obtained suggesting that this selection yields only a couple of interacting DZF domains.

3.3.2.3 Plasmid linkage

For all three independent selection sets (A, B and C) 2 to 12 colonies were picked from each of the four selection plates. Plasmids encoding the interacting pairs of DZF variants were then isolated from these colonies.

0

Miller (β-Galactosidase) units Miller (β-Galactosidase) units

Miller (β-Galactosidase) units

Miller (β-Galactosidase) units Miller (β-Galactosidase) units

Miller (β-Galactosidase) units

Selection A Selection B

Selection C

Figure 3.6 Plasmid linkage analysis for selected interacting pairs. Plasmids isolated from colonies that survived the selection were transformed into the B2H reporter strain and β-galactosidase assays were performed. Each graph corresponds to one selection “set” (selection A, B and C) which consists of 4 individual selection experiments (S1-S4, S5-S8, S9-S12). Note that each bar shown for an individual selection in a particular “set” corresponds to one colony picked from the respective selection plate.

For plasmid linkage analysis, each of these pairs of plasmids was transformed into the B2H reporter strain and β-galactosidase assays were performed. The result (Figure 3.6) of these assays demonstrated that for each selection experiment most (average of ~ 85 %) of the selected pairs were able to activate transcription in the B2H system. This confirms that the survival phenotype of the cells on the selection plates is linked to their ability to activate expression of the selection gene. The activation obtained for most of the selected pairs of DZF domains is similar to the activation for the wild-type DZFs. Interestingly, some of the novel DZF variants displayed higher β-galactosidase units than the wild-type proteins, implying that these novel pairs might interact more strongly than wild-type DZF domains (Figure 3.6 and data not shown).

3.3.2.4 Sequencing

Plasmid pairs that resulted in elevated level of β-galactosidase expression were sequenced to determine the identities of the individual C2H2 ZFs and linkers encoded. Sequencing revealed that the selection identified several novel heterodimeric pairs of interacting DZF domains which represent combinations of C2H2 ZFs from the parental DZFs used to construct the libraries (Table 3.2). Many of the interacting pairs contain C2H2 ZFs derived from wild-type DZFs which are known to interact with each other. For example, the majority of selected DZFs from library A are various shuffled combinations of the subdomains from Ikaros, Eos and Pegasus. It seemed likely that these shuffled DZFs would all interact with the parental DZF domains from which they are derived and would therefore not exhibit novel interaction specificities. Similarly, library B only harbored shuffled ZF combinations consisting of subdomains from either Pegasus and Hunchback D.m. or from Hunchback D.m.

and Hunchback L.m.. We were not able to pull out chimeric proteins containing ZFs from Pegasus and Hunchback L.m. together in a single DZF domain. However, three of the identified pairs from Selection A consisted of DZF domains composed of subdomains from parental DZFs that do not interact with each other. Interestingly, selection C harbored only one pair of interacting DZF variants also composed of C2H2 ZF subdomains from wild-type DZFs which do not interact with each other (Table 3.2). These four pairs are likely to represent DZF domains with potentially novel specificities and were therefore chosen for additional analysis (see below).

3.3.2.5 Linkage analysis

To further confirm that activation of reporter gene expression in the B2H system is linked to the presence of various shuffled DZF fragments that were isolated from the different selections, fragment analysis was performed. Thus, fragments encoding selected DZF variants were re-cloned into fresh B2H plasmids and were re-tested for their ability to activate reporter gene expression by performing β-galactosidase assays. It was found that all of the shuffled DZFs tested mediated interaction with their selected partner indicating that the activation phenotype in the B2H system is linked to the specific DZF domain (data not shown).

Table 3.2 Interacting pairs of synthetic shuffled DZF domains identified by genetic selection.

Identities of the subdomains present in the synthetic DZF domains are shown. Subdomains are defined as N-terminal ZF, linker, and the C-terminal ZF (shown from left to right). The Selection “set” (A, B and C) from which these pairs were identified as well as the number of times each pair was identified are also indicated. Abbreviations for the subdomains are as defined in the legend to Figure 3.4. Location of the synthetic DZF on the Zif268 or RNAP a-subunit is shown on top of the table. Interacting pairs chosen for further characterization are marked in bold red text.

Library Set A

Library Set B

Library Set C ^{Tr-Eo-Eo Eo-Eo-Hd 3}

6

Zif268 hybrids RNAP α hybrids Frequency

Library Set A

Library Set B

Library Set C ^{Tr-Eo-Eo Eo-Eo-Hd 3}

6

Zif268 hybrids RNAP α hybrids Frequency

3

Zif268 hybrids RNAP α hybrids Frequency