Global l structure t of sensory transcription networks 02/7/2012
Counting possible graph patterns in an n-node graph One 1-node Three 2-node graph pattern graph patterns Thirteen 3-node graph patterns 1999 4-node graph patterns
The network motifs of 4 nodes in sensory transcription networks Although there are a total of 1999 possible graph patterns formed by 4-nodes, only two of them are network motifs in known sensory transcription networks: 1. One is called the two-output FFL, in which the output gene Z and its edges in a simple FFL is duplicated. 2. The other is called the bi-fan, in which two input transcription factors X 1 and X 2 jointly regulate two output genes Z 1 and Z 2. X X 1 X 2 The two-output FFL motif Y The Bifan motif Z 1 Z 2 Z 1 Z 2
A family of graphs that have the same theme architecture There are 3, 13, 1,999, 9,364 and more than a million graph patterns in 2-, 3-, 4-, 5- and 6-node graphs, respectively. Thus, the number of graph patters in a digraph increase extremely rapidly, however, the unique patterns of n-node graph do not increase that fast. A large number of similar patters can be grouped into a family with the same architectural theme as a smaller basic pattern. Each of these families can be considered to be generated by a repeated topological generation process starting from a simple basic pattern. In a simple topological generation process, one nodes and its edges in the basic pattern are duplicated, and are added back to the original pattern. This results in a family of graphs that all contain the original basic pattern.
Types of network motifs in sensory transcription networks There are a total of 5 types of network motifs found in sensory transcription networks so far. These include small motifs with fixed number of nodes: 1. Autoregulation; 2. Feed forward loop; And motif families with varying number of nodes: 3. Single-input module (SIM) family; 4. Multi-output FFL family, formed by multiple FFLs. 5. Densely overlapping regulon (DOR) family. Each member in the family shares the same architectural pattern.
Simple topological generalizations of the FFL From Alon Fig 5.6
Simple topological generalizations of the FFL If the basic graph pattern is a motif, then all the members of the generated graph is likely to be a motif, because the chance for them to occur in a random network is even lower than the basic graph pattern. But it turns out that not all of them can be a true motif, because their occurrences in a real network are as low as in a random network. In the case of the families generated by a FFL, only members of the multi-output family are network motifs. From Alon Fig 5.6
Dynamics of multi-output FFL: E. coli flagella motor as an example E. coli flagella motor is a 50nm molecular machine, assembled by about 30 proteins in sequential stages. Robert.Macn Ann. Rev. Microbiol. 2003, 57:77-100
The multi-output FFL that regulates the flagella motor genes in E. coli The 30 motor proteins are encoded in 6 operons, each is regulated by two transcription activators X=flhDC, and Y=fliA, and Y is regulated by X, thus forming a multi-output FFL. X Z1=fliL, Z2=fliE etc. K xy K 1 K 2 K n Y K 1 K 2 K n OR OR OR Z 1 Z 2 Z n From Alon Fig 5.8
The multi-output FFL that regulates the flagella motor genes in E. coli Each operons is activated by X in the absence of Y, and by Y in the absence of X. Thus, the input function is approximated by an OR gate. When X is activated, it will activate Y as well as Zs one by one with a interval of about 0.1 cell generation. X Z1=fliL, Z2=fliE etc. K xy K 1 K 2 K n Y K 1 K 2 K n OR OR OR Z 1 Z 2 Z n From Alon Fig 5.8
Temporal order in the flagella system of E. coli The temporal order matches the assembly order of the flagella, in which proteins are added to the existing structure going from the intra-cellular to the extra-cellular sides. This is achieved by the design of ordered strength of promoters: k xz1 < k xz2 < k xzn From Alon Fig 5.9
First-in first out order (FIFO) in the multi-output FFL with an OR-logic input function Genes Z 1, Z 2, and Z n are sequentially turned on when X crosses their respective activation thresholds: k XZ1 < k XZ2 < k XZn. When X starts decaying, Y takes over to activate Zs. Genes Zs are turned off sequentially when Y decays below their activation thresholds: Since k XZ1< k XZ2< k XZn, and k YZ1 > k YZ2 > k YZn. k YZ1 > k YZ2 > k YZn are designed in opposite ordering, a first-in first out order (FIFO) is obtained in a multi-output FFL with an OR-logic input function.
FIFO in the multi-output FFL with an OR-logic input function k 2 K 1 1 0.5 X X K 1 K 2 K 2 0 0 1 2 3 4 5 6 7 8 9 10 1 0.5 Y K 1 Y K 1 K 2 K 2 0 0 1 2 3 4 5 6 7 8 9 10 1 0.5 Z 1 Z 2 Z 1 Z 2 K 1 <K 2 K 1 >K 2 0 0 1 2 3 4 5 6 7 8 9 10 time time From Alon Fig 5.10
The multi-output FFL can act as a persistence detector for each output The multi-output FFL also conveys all of the functions of the single FFL: namely, sign sensitive delay and fluctuation filtering. In the case of the multi-output FFL with OR-logic integration at promoters of output genes Z 1, Z 2, and Z n, the circuit can filter out brief fluctuation of X, thus Z 1, Z 2, and Z n are continuously produced in a brief absence of X. The average delay time for Zs in the E. coli flagella motor synthesis pathway is about one cell generation, that grantees the full set of motor proteins can be produced in the swimming cell that is likely to encounter fluctuating environments. It turns out that the multi-output FFL mostly frequently occurs in It turns out that the multi output FFL mostly frequently occurs in the sensory transcription networks.
Densely overlapping regulons Topological generation of the bi-fan motif gives rise to the socalled densely overlapping regulon (DOR) motif family, which is shaped as a layer of inputs with multiple overlapping connections to a layer of outputs. In real life, a DOR is not fully connected through it is much denser than in a random network. The function of a DOR can be modeled by a multi-dimensional input function that integrates the input at the promoter of each gene----decision making computation, similar to a neural network. 1 1 1 1 1 1 1 2 2 2 2 2 2 2
The osmotic stress response system in E. coli is controlled by a DOR network motif A DOR often involves global regulators to mobilize a wide range of genes, and local regulator to fine tune the regulations. From Alon Figure 5.13
Global structure of transcription networks in E. coli in terms of network motifs Carbon utilization DOR Osmotic stress DOR Stationary phase DOR DNA metabolism DOR Drug and superoxide DOR maltose flagella Heat shock From Alon Figure 5.14
Global structure of transcription networks in E. coli and yeast in terms of network motifs The structures of the transcription networks in E. coli and yeast show that: 1. Sensory transcriptional networks are made of a layer of DORs, and there is no DOR at the output of another DOR. Thus, most of the computations of the networks are done at the cortex of promoters within the DORS. 2. The FFLs and SIMs are integrated within the DORs. 3. Negative autoregulation is often integrated with FFLs and the master regulator of SIMs. Overall, the global l structure t of transcription ti network is organized in a rather simple way. This makes it possible to understand the dynamics of each This makes it possible to understand the dynamics of each motif in isolation, even though it is embedded in a larger network.
Summary of global structure of sensory transcription networks All the known sensory transcription networks in bacteria, yeast, worms, fruit flies and humans are composed of these four network motifs, autoregulation, ti FFLs, SIMs and DORs. Sensory transcription networks are rather shallow, there are not many cascades chain interactions. This is because sensory networks require fast response, but gene transcription is a rather slow process. Cascade chaining will make the response time too slow for the organism to survive. SIMs and DORs tend to operate in systems that require fast signal propagation and coordinated expression. FFLs occur in system that guide progression in stages such as the cycle.
Summary of global structure of sensory transcription networks Negative auto-regulation X Speeds response time, steady-state robust to fluctuations in production Positive auto-regulation Coherent feed-forward loop C1- FFL Incoherent feed-forward forward loop I1-FFL X X Y Z X Y Z Slows response time, possible bi-stability Sign-sensitive delay, filters out brief ON (OFF) input pulses when the Z-input function is AND (OR) logic. Pulse generation, sign-sensitive response acceleration
Summary of global structure of sensory transcription networks Single-input module (SIM) Multi-output feed-forward loop (multi-output FFL) X Coordinated control, temporal (LIFO) order of promoter activity Y 1 Y 2... Y n (LIFO) order of promoter activity X Y Z 1 Z 2 Z n Acts as FFL for each input (sign-sensitive delay, etc), FIFO temporal order of promoter activity Bi-fan Densely overlapping regulons (DOR) X 1 X 2 Y 1 Y 2 X 1 X 2 X n Y 1 Y 2 Y m Combinatorial logic based on multiple inputs, depends on input-function of each gene