Network architecture of signaling from uncoupled helicase-polymerase to cell cycle checkpoints and trans-lesion DNA synthesis

Abstract

When replication is blocked by a template lesion or polymerase inhibitor while helicase continues unwinding the DNA, single stranded DNA (ssDNA) accumulates and becomes coated with RPA, which then initiates signals via PCNA mono-ubiquitination to activate trans-lesion polymerases and via ATR and Chk1 to inhibit Cdk2-dependent cell cycle progression. The signals are conveyed by way of a complex network of molecular interactions. To clarify those complexities, we have constructed a molecular interaction map (MIM) using a novel hierarchical assembly procedure. Molecules were arranged on the map in hierarchical levels according to interaction step distance from the DNA region of stalled replication. The hierarchical MIM allows us to disentangle the network’s interlocking pathways and loops and to suggest functionally significant features of network architecture. The MIM shows how parallel pathways and multiple feedback loops can provide failsafe and robust switch-like responses to replication stress. Within the central level of hierarchy ATR and Claspin together appear to function as a nexus that conveys signals from many sources to many destinations. We noted a division of labor between those two molecules, separating enzymatic and structural roles. In addition, the network architecture disclosed by the hierarchical map, suggested a speculative model for how molecular crowding and the granular localization of network components in the cell nucleus can facilitate function.

Introduction

Much has been learned about the complex molecular machinery that controls and implements cell functions. It is becoming increasingly important however to organize that information in a way that sheds light on those functions at the system level. For that purpose, graphical portrayals or network maps are needed to serve in much the same way as circuit diagrams do in electronics. To that end, we previously developed a graphical notation for Molecular Interaction Maps (MIMs) designed to deal with the complexities presented by bioregulatory networks that include such components as multi-molecular complexes, protein modifications and enzyme-enzyme interactions.^1–3 The notation allows the interactions of the modified forms of a given molecular species, as well as its reactive sites and domains, to be represented in one place on the map. In addition, the notation does not require detailed knowledge of the combinatorial influences among the interactions. The MIM notation has been used to generate compact and comprehensive maps of mammalian cell cycle control, DNA replication, and DNA repair to serve as information organizers and as hypothesis generators for experiment and computer simulation (http://discover.nci.nih.gov/mim/).^1–4

For optimal utility, however, it is important to arrange the molecular components of a map in a manner that best reveals relationships and pathways (discussed, for example, by Hu et al.⁵). Also, the high connectivity and cross-talk among functional portions of the global cell regulatory network makes it hard to decide where to limit the scope of a network diagram, so as to keep it coherent and of manageable size. Excessive map size makes a map difficult to create and to read. Moreover, biological network diagrams have usually been arranged according to preconceptions of function or modularity.

Here we use a more objective method, based on interaction hierarchy, to show important relationships and to focus a map in a manner that keeps its size within a coherent scope. We develop an orderly hierarchical MIM focused on signaling from stalled replication forks by way of RPA, ATR and Chk1, including the processes of helicase-polymerase uncoupling, trans-lesion synthesis, and initiation of checkpoint signals. Mammalian replicative polymerases stall when encountering a lesion on the template strand or when the enzyme is inhibited. While the polymerase is blocked, the helicase at the head of the fork continues to unwind the DNA. The consequent helicase-polymerase uncoupling generates long stretches of single-stranded DNA (ssDNA) that are the source of signals to cell cycle checkpoints, trans-lesion polymerase switching, and other responses.^6–8 The hierarchical MIM of this network revealed aspects of architecture and pathways that might not otherwise have been obvious and that led us to propose a speculative model of how the signals may be generated and transmitted within the molecularly crowded cell medium.

Results

Hierarchical levels of signaling from helicase-polymerase uncoupling

A network map of signaling from stalled replication (Fig. 3) was assembled according to the hierarchical procedure outlined in Figure 2 (See Methods). The sequence of addition of molecular species to the various levels of the hierarchy is given in Table 1. The MIM symbols used are defined in Figure 1. The successive levels of hierarchy are shown in bands of alternating background color (Fig. 3). References supporting the interactions displayed on the map are cited in the Annotations (Appendix 2).

Figure 3.

MIM of signaling from helicase-polymerase uncoupling to trans-lesion DNA synthesis and cell cycle inhibition, assembled according to the scheme shown in Figure 2, which generated 5 levels of hierarchy. The order of assembly at each hierarchic Level is listed in Table 1. The color scheme for interactions is arbitrary and does not affect the meaning of the notation. We generally use green for stimulations and red for inhibitions. (Highlighting of interactions or pathways in subsequent figures superimposes different color schemes.)

Figure 2.

Generic schema for hierarchical MIM assembly.

Table 1.

Hierarchical assembly of MIM events at a stalled replication fork, according to the processes marked in Figure 2 [The interaction numbers are as marked in the MIM figures and in the Annotations (Appendix 2)]

Step	Process	Interaction number	Description	Species added
1	A	-	DNA and its interactive sites in the region of a stalled replication fork.
2	B	1	MCM binds to dsDNA at replication fork.	MCM
2	B	2	Helicase action of MCM separates DNA strands.
2	B	3	GINS binds to ssDNA at replication fork	GINS
2	B	4	RPA binds ssDNA accumulated during helicase-polymerase uncoupling.	RPA
2	B	5	Polα binds at accumulated ssDNA	Polα
2	B	6	Polα initiates a new DNA strand (primase and replicative polymerase actions)
2	B	7	Rad17-RFC binds at ss/ds DNA junctions.	Rad17-RFC
2	B	8, 9	DNA-bound Rad17-RFC loads ring-shaped complex Rad9:Rad1:Hus1 (“9-1-1”).	Rad9:Rad1:Hus1 (“9-1-1”)
2	B	10	9-1-1 enhances replicative DNA synthesis (?)
2	B	11	PCNA ring binds DNA near the growing end of daughter strand.	PCNA
2	B	12, 13	Polδ catalyses DNA leading strand synthesis, promoted by DNA-bound PCNA.	Pol δ
2	B	14	Template lesion blocks progress of replicative polymerase.	Template lesion
2	B	15	Trans-lesion polymerases can add nucleotides despite template lesion.	Pol η, ι, κ, Rev1
3	C	16	RPA assists the DNA-binding of Polα
3	C	17, 18	Ubiquitin ligase Rad6:Rad18 binds RPA at accumulated ssDNA.	Rad6:Rad18
3	C	19, 20	PCNA can bind replicative polymerases or trans-lesion polymerases.
3	C	21	Binding of TLS polymerases to PCNA promotes trans-lesion synthesis.
3	C	22	PCNA binds Rad18 in Rad6:Rad18 dimer.	PCNA:Rad18
3	C	23	Rad6:Rad18 monoubiquinates PCNA.	PCNA-Ub1
3	C	24	Rad18 binds Polη.	Rad18:Polη
3	C	25	MCM binds GINS via interaction between subunits MCM5 and Psf2.	MCM:GINS
4	D2	26	Tipin in TIM:Tipin binds RPA2 in RPA.	TIM:Tipin
4	D2	27	Tipin in TIM:Tipin binds MCM.	Tipin:MCM
4	D2	28	TIM:Tipin co-localize with PCNA (indirect?)
4	D2	29	Atrip in ATR:Atrip binds RPA2 in RPA.	ATR:Atrip
4	D2	30	Atrip in ATR:Atrip binds MCM7 in MCM.	ATR:MCM
4	D2	31	ATR phosphorylates RPA2.	RPA2-P
4	D2	32	ATR phosphorylates Rad17.	Rad17-P
4	D2	33	ATR phosphorylates Rad9.	Rad9-P
4	D2	34	ATR phosphorylates MCM2, 3, and 4.	MCM-P
4	D2	35	Claspin binds PCNA.	Claspin:PCNA
4	D2	36	Claspin avidly binds TIM in TIM:Tipin.	Claspin:TIM
4	D2	37	Cdc45 binds MCM:GINS.	Cdc45:MCM:GINS
4	D2	38	Cdc45 binds Polδ or Polε.	Cdc45:Polδ
5	C	39	Claspin binds phosphorylated Rad17.	Claspin:Rad17-P
5	C	40	Phosphorylated Rad17 stabilizes binding of 9-1-1 to ssDNA.
5	C	41	ATR phosphorylates Claspin T916 indirectly.	Claspin-P
5	C	42	Claspin binding to Rad17-P facilitates ATR-dependent Claspin phosphorylation.
5	C	43	Phosphorylation of RPA2 may enhance RPA binding to sites of replication stress.
5	C	44	Claspin and TIM:Tipin enhance binding between PCNA and Rad18.
5	C	45	ATR activation is enhanced by its association with RPA and thereby with ssDNA.
5	C	46	Phosphorylation of MCM4 reduces the affinity of MCM for chromatin.
5	C	47	Cdc45:Psf2:MCM5 interaction supports MCM helicase activity.
5	C	48	Monoubiquitnation of PCNA enables its binding to trans-lesion polymerases.
6	D2	49	TopBP1 binds phosphorylated Rad9.	TopBP1
6	D2	50	TopBP1 binds Atrip in the ATR:Atrip complex.
6	D2	51	ATR phosphorylates TopBP1.	TopBP1-P
6	D2	52	Chk1 binds Claspin.	Chk1
6	D2	53	Chk1 is phosphorylated by ATR.	Chk1-P
6	D2	54	Cdc7 binds Claspin	Cdc7
6	D2	55	Cdc7 phosphorylates Claspin	Claspin-P
6	D2	56	Brca1 binds Claspin.	Brca1
6	D2	57	Brca1 is phosphorylated by ATR.	Brca1-P
6	D2	58	βTrCP ubiquitinates Claspin	βTrCP, Claspin-Ub
6	D2	59	USP7 and/or USP28 deubiquinate Claspin	USP7, USP28
6	D2	60	ATR phosphorylates ATM	ATM, ATM-P
6	D2	61	ATR phorphorylates p53	p53, p53-P
7	C	62	TopBP1 binding to phosphorylated Rad9 stimulates TopBP1 binding to Atrip.
7	C	63	TopBP1 binding to ATR:Atrip is required for kinase activity of ATR.
7	C	64	Polyubiquinated Claspin is degraded via proteasomes.
7	C	65	Claspin T916-phosphorylation enhances binding to Chk1
7	C	66	Claspin T916-phosphorylation enhances binding to Brca1
7	C	67	Chk1 binding to Claspin stabilizes Claspin.
7	C	68	Cdc7 stabilizes Claspin
7	C	69	Chk1 binding to Claspin stimulates ATR-induced Chk1 phosphorylation.
7	C	70	Brca1 binding to Claspin stimulates Brca1 phosphorylation by ATR
7	C	71	Chk1 binding to Claspin stabilizes Chk1
7	C	72	Phosphorylation of Brca1 stimulates Chk1 phosphorylation by ATR
7	C	73	Phosphorylated ATM phosphorylates p53
8	D2	74	53BP1 binds USP28	53BP1
8	D2	75	ATR phosphoryltes 53BP1	53BP1-P
8	D2	76	53BP1-P may stimulate Claspin deubiquitination by USP28 (?)
8	D2	77	Chk1 phosphorylates Cdc25A & B	Cdc25A & B
8	D2	78	bTrCP ubiquinates and stimulates degradation of Cdc25A & B	Cdc25A & B
8	D2	79	ATM-P phosphorylates and activates Chk2	Chk2, Chk2-P
8	D2	80	p53 stimulates transcription of p21cip1	p21cip1
8	D2	81	p53-stimulated p21 transciption is enhanced by p53 S15 phosphorylation.
8	D2	82	hCLK2 binds ATR and stimulates binding of ATR to TopBP1	hCLK2
8	D2	83	HYD binds TopBP1	HYD/EDD
8	D2	84	HYD stimulates TopBP1 degradation.
8	D2	85	HYD binding to TopBP1 may be inhibited by ATR-induced TopBP1 phosphorylation.
9	C	86	Chk2 phosphorylates Cdc25A & B at S76.
9	C	87	ATM-induced Chk2 phosphorylation stimulates Chk2-induced phosphorylation of Cdc25A & B
9	C	88	Cdc25 phosphorylation by Chk1 is stimulated by ATR-induced Chk1 phosphorylation.
9	C	89	Cdc25 phosphorylation at S76 stimulates Cdc25 ubiquitination.
10	D2	90	p21 binds Cdk2	Cdk2
10	D2	91	p21:Cdk2 binding inhibits the kinase activity of Cdk2
10	D2	92	Phosphorylation of Cdk2 at T14 and T15 inhibits Cdk2 kinase activity
10	D2	93	Cdc25A & B remove the inhibitory phosphorylation from Cdk2.
10	D2	94	Cdk2 phosphorylates several proteins that promote cell cycle progression.

Figure 1.

MIM symbols used in the current manuscript. For a detailed description of the MIM notation, see reference ² and/or http://discover.nci.nih.gov/mim/. New symbols added for the current manuscript are marked with an asterisk (*). A small ellipse surrounding a DNA single or double strand, represents binding as a DNA-encircling ring molecule that can slide along the DNA. New symbols also were added for DNA actions: polymerase, primase, helicase and trans-lesion synthesis.

The map describes a central part of the network that signals the presence of stalled replicative polymerase to the controls of trans-lesion DNA synthesis and Chk1-mediated cell cycle checkpoints. It depicts our current understanding of how signals from single-stranded DNA (ssDNA), accumulated during helicase-polymerase uncoupling, facilitates trans-lesion synthesis and inhibit cell cycle progression.

Eukaryotic replication forks stall when the replicative DNA polymerase is unable to proceed, either because it has encountered a blocking lesion on the DNA template strand or because its enzymatic activity is suppressed by a chemical inhibitor or by lack of nucleotide precursors. In this scenario, the helicase that unwinds the DNA double helix ahead of the polymerase continues its unwinding action while the polymerase is blocked, thereby generating long stretches of conformationally single-stranded DNA between the unwinding site and the polymerization site.^6,7 That circumstance, known as “helicase-polymerase uncoupling,” generates the ssDNA that serves as input to the network. The depicted outputs are recruitment of trans-lesion polymerases and inhibition of cell cycle progression via the checkpoint kinase Chk1.

The overall network map (Fig. 3) shows 5 interaction levels in the hierarchy from the proximal (DNA) level to the distal (cell cycle checkpoint) level. Level 1 contains the DNA structure in the region of helicase-polymerase uncoupling. A lesion on the template strand blocks the replicative polymerase (interactions 12 and 14), whereas helicase action continues to unwind the DNA strands (interactions 1 and 2), as a result of which ssDNA accumulates in the region between polymerase and helicase. (The small filled circles within a rectangle denote the binding sites for coating of ssDNA by RPA). Only events on the leading strand are shown, but analogous events are likely on the lagging strand.

Direct interactions with DNA in the region of helicase-polymerase uncoupling

Level 2 contains the molecular species that interact directly with the DNA in the region of helicase-polymerase uncoupling. Included are the molecules responsible for the helicase and polymerase actions, as well as the proximal machinery that signals the presence of ssDNA accumulated due to helicase-polymerase uncoupling.

The progress of the replicative polymerase (Pol δ or ε) is assisted by PCNA, a trimeric ring that clamps around the DNA double helix and remains associated with the blocked polymerase (interactions 11–13).

The ssDNA generated by helicase-polymerase uncoupling becomes coated with RPA (interaction 4), which recruits Polα, the primase-polymerase that initiates the synthesis of a new DNA strand (interactions 16, 5 and 6). It is uncertain whether new strand initiation occurs on the leading strand in mammalian cells; the Polα-related steps may occur mainly on the lagging strand (see Annotations).

Rad17-RFC binds to a ss/dsDNA junction, which may form when the RNA priming segment deposited by Polα is degraded in the course of new strand intitiation. Rad17-RFC loads a clamp, Rad9:Rad1:Hus1 (“9-1-1”), onto the replicating new DNA strand. The details of 9-1-1 function are uncertain, but it may assist the polymerase in a manner akin to PCNA (interactions 8–10).

At the replication fork, the MCM:GINS combination unwinds the DNA, a process that in the current scenario is not blocked [interactions 1–4]. GINS is a ring protein complex associated with the MCM helicase. The GINS ring may accommodate a DNA single strand (see Annotations).

Polymerase switching for trans-lesion DNA synthesis

Intimately associated with components at Level 2 are the pathways that recruit trans-lesion polymerases when the replicative polymerase is unable to proceed. If the block is due to chemically altered bases on the template strand, these polymerases are able to insert one or a few bases to the replicating strand so as to bypass the lesion. The trans-lesion polymerases (e.g., Pol η, ι, κ, Rev1) each have their own specificities for the kinds of template defects they can bypass, but at the cost of a high error rate. Monoubiquitination of PCNA allows the polymerase binding sites on the PCNA trimer to switch between the different polymerases (interaction 48).

The molecular interactions that make polymerase switching permissive are highlighted in Figure 4. The steps leading to PCNA monoubiquitnation are highlighted red; the subsequent steps leading to polymerase switching are highlighted blue. The process begins at ssDNA-bound RPA generated by helicase-polymerase uncoupling, often indicative of stalled replicative polymerase. RPA binds ubiquitin transferase complex Rad6:Rad18 (interaction 17), which then binds PCNA (interaction 22) and monoubiquitinates it (interaction 23). Monoubiquination allows the polymerase binding sites on PCNA to switch between replicative polymerase Pol™ or Σ (interaction 19) and one or another of the trans-lesion polymerases (interaction 20). The recruited trans-lesion polymerases can then copy (or add arbitrary nucleotides, usually A) across the template defect (interactions 21 and 15).

Figure 4.

Pathways leading from ssDNA-bound RPA to the recruitment of an error-prone polymerase η, ι, κ or Rev1 for bypass of a potential template lesion. The pathway leading from RPA to monoubiquitination of PCNA is highlighted red, and the subsequent steps leading to polymerase switching and trans-lesion synthesis are highlighted blue. Possible influences from ATR and Claspin are also shown. (See text and Annotations for more detailed descriptions).

Possible influences on this process by ATR and Claspin from Level 3 are noted at the end of the next section.

Constriction of signaling pathways at the level of ATR and claspin, with segregation of enzymatic and binding actions

In level 3, we note that there are relatively few components, suggesting a constriction in the signaling path (Fig. 3). The constriction may be akin to a “transmission line” that gathers proximal signals from many sources and sends signals out to many distal effectors. The major components at this level are ATR and Claspin.

ATR sends signals distally through phosphorylation of proteins in levels 4 and 5: 53BP1, Brca1, Chk1, ATM, p53 and TopBP1 (Fig. 5, red highlights to levels 4 an 5). In addition, however, ATR signals proximally through phosphorylation of proteins in level 2: RPA, Rad17, Rad9 and MCM (Fig. 5, red highlights to level 2).

Figure 5.

“Transmission line” architecture of the network at Level 3. ATR phosphorylates components at both higher and lower levels of the hierarchy (red highlight), while Claspin and TIM:Tipin appear to bind the components together (blue highlight). ATR phosphorylates and stabilizes Claspin (green highlight). Thus ATR and Claspin appear to enhance each other’s actions.

Claspin, on the other hand, lacks enzymatic activity and appears to function through binding to proteins in the distal region (Brca1, Cdc7 and Chk1), as well as proteins in the proximal region directly (PCNA and Rad17) or indirectly via the TIM:Tipin complex (RPA and MCM) (Fig. 5, blue highlights). Claspin therefore may hold ATR target proteins together at the DNA, thereby facilitating their phosphorylation by ATR. Moreover, ATR-induced phosphorylation stabilizes Claspin (Fig. 5, green highlight). The enzymatic (ATR) and structural (Claspin and TIM:Tipin) components therefore appear to cooperate to mutually enhance their actions.

In addition, ATR and Claspin may facilitate the recruitment of trans-lesion polymerases through PCNA monoubiquination by Rad18:Rad6 (see Fig. 4). ATR may do so through binding and phosphorylation of RPA (interactions 29, 45 and 31), which facilitates recruitment of Rad18:Rad6 (interaction 17). Claspin may do so through its binding to PCNA (interaction 35), which also facilitates recruitment of Rad18:Rad6 (interactions 44 and 22).

Failsafe activation of ATR by way of a 2-man rule

Optimal activation of ATR is thought to require concurrent signaling using a “two-man rule” to verify signal validity and avoid inappropriate activation of ATR-dependent checkpoints.⁸ Both signals are initiated by the DNA structural changes caused by helicase-polymerase uncoupling. One signaling path originates from ssDNA-bound RPA (Fig. 6, blue highlights), the other from 9-1-1 complexes loaded by Rad17-RFC at or near the ssDNA (Fig. 6, red highlights). The latter signaling path may be particularly important, because it proceeds via TopBP1, a critical element in ATR activation. ⁸ The binding of TopBP1 to phosphorylated Rad9 in the 9-1-1 complex (interaction 49) enhances TopBP1 binding to Atrip in the ATR:Atrip complex (interactions 62 and 50), which in turn stimulates the kinase activity of ATR (interaction 63).

Figure 6.

Pathways to fail-safe activation of ATR. Full activation of ATR requires concurrent signaling from two pathways: (1) Rad17-RFC via 9-1-1 and TopBP1 (red), and (2) ssDNA via RPA (blue). Also highlighted is positive feedback by ATR-induced phosphorylation of Rad9, which provides more TopBP1-binding sites on the 9-1-1 complex (green).

Feedback interactions of ATR

Four positive feedback loops for ATR activation can be discerned (Fig. 7). Three of them go by way of proximally-directed ATR-induced phosphorylations, as follows: (1) ATR phosphorylates the Rad9 component of the 9-1-1 complex, thereby increasing the availability of binding sites for TopBP1 in that ATR-activation path. Figure 6, green highlight, shows the ATR-induced 9-1-1 phosphorylation segment; Figure 7, red highlight, shows the entire loop. (2) ATR phosphorylates Rad17, thereby stimulating the binding of 9-1-1 to DNA (interaction 40, detailed mechanism unknown) (Fig. 7, green highlight). (3) ATR phosphorylates the Rpa2 subunit of RPA, which may enhance the positive signal leading to ATR activation (Fig. 7, purple highlight). (4) ATR phosphorylates TopBP1 with consequent inhibition of TopBP1 degradation (Fig. 7, blue highlight; see Discussion).

Figure 7.

Feedback interactions of ATR. Red: ATR phosphorylates Rad9, which provides binding sites for TopBP1; that binding enhances the binding of TopBP1 to the ATR:Atrip complex; the latter binding strongly stimulates the kinase activity of ATR. Blue: ATR phosphorylates TopBP1, thereby inhibiting TopBP1 degradation and making more TopBP1 available for stimulation of ATR. Green: ATR phosphorylates Rad17, thereby stabilizing the association of the 9-1-1 complex with sites of replication stress. Purple: ATR phosphorylates RPA, thereby enhancing the binding of RPA to ssDNA in regions of replication stress, where it can bind the ATR:Atrip complex and stimulate ATR. Yellow: ATR phosphorylates MCM4, thereby reducing the affinity of the MCM complex for chromatin and reducing helicase action. Thus activation of ATR in response to helicase-polymerase uncoupling may feed back to inhibit helicase activity. The phosphorylaton of MCM4 by ATR may be facilitated by the binding of MCM7 to Atrip.

In a potential feedback that limits helicase-polymerase uncoupling (Fig. 7, yellow highlight), ATR interacts with the MCM complex by binding MCM7 and phosphorylating MCM4. The consequence of ATR-induced MCM4 phosphorylation is to inhibit the MCM helicase activity, suggesting that, as ATR is activated in response to helicase-polymerase uncoupling, ATR may feed back to limit the extent of uncoupling.

Another effect of ATR-induced Rad17 phosphorylation is to increase the availability of binding sites for Claspin (Fig. 5, interaction 39), which facilitates ATR-induced Claspin phosphorylation and consequent Claspin stabilization (Fig. 5, interaction sequence 42, 41, 65, 67, 64).

MIM representation of a model of ATR and Chk1 activation

The 2-man rule, as well as the feedback loops involving ATR-Rad17-Rad9-TopBP1, are incorporated in a model of ATR and Chk1 activation proposed by Medhurst et al.¹⁰ The model suggests how ATR and Rad17 cooperate to localize Rad9 to sites of DNA damage and how that leads to Chk1 activation. To show how their model is represented in the MIM, we quote verbatim their description of the model in the legend to their Figure 5C, and indicate our corresponding interaction identification numbers (see Fig. 8).

ATR and Rad17 collaborate in modulating Rad9 localization at sites of DNA damage. ssDNA generated as a result of DNA damage or replication stress is recognized by RPA (interaction 4). Recognition of ssDNA by RPA leads to the independent recruitment of ATR and Rad17 to DNA lesions (interactions 29 and 7, respectively). Rad17 loads the 9-1-1 complex at sites of ssDNA (interactions 8 and 9) and facilitates activation of ATR through an interaction with TopBP1 (interaction 49). ATR subsequently phosphorylates Claspin (interaction sequence 39, 42, 41), which acts to recruit Chk1 and promote its phosphorylation by ATR (interaction sequence 65, 52, 69, 53). ATR also phosphorylates Rad17 (interaction 32). ATR-mediated phosphorylation of Rad17 stabilizes the 9-1-1 complex at sites of DNA lesions (interaction 40). This could in turn result in the maintenance of activated ATR and continued checkpoint signaling until DNA damage is repaired (interaction sequence 49, 62, 50, 63).¹⁰

Figure 8.

ATR and Rad17 collaborate to localize Rad9 to sites of DNA damage and lead to activation of Chk1, according to the model proposed by Medhurst et al¹⁰ in their Figure 5C (see text). The relevant part of the MIM in Figure 3 is shown.

From ATR and Chk1 to inhibition of Cdk2-dependent cell cycle progression

The hierarchical network map shows also how signals are transmitted through Chk1 to the cell cycle control machinery. ATR and Claspin in Level 3 cooperate to phosphorylate and activate Chk1 in Level 4, whereupon Chk1 phosphorylates cell cycle controllers such as Cdc25A in Level 5 (Fig. 9). ATR phosphorylates Chk1 with assist from Chk1:Claspin binding (Fig. 9, green highlights; interactions 53 and 69). Binding between Chk1 and Claspin may also enhance the level of phosphorylated Chk1 by mutual stabilization of Chk1 and Claspin (interactions 71 and 67). Phosphorylation of Chk1 is also assisted by ATR-phosphorylated Brca1, and Brca1 phosphorylation by ATR is assisted by Brca1:Claspin binding (Fig. 9, red highlights). The interactions leading from phosphorylated Chk1 to cell cycle controllers are shown in Figure 9, blue highlights.

Figure 9.

Pathways from ATR and Claspin to inhibition of cell cycle progression. The MIM was extracted from Figure 3. Sections of the network are highlighted as follows. Green: ATR phosphorylates Claspin, thereby permitting Claspin to bind Chk1 and enhancing the ability of ATR to phosphorylate Chk1. Red: ATR-induced phosphorylation also allows Claspin to bind Brca1, thereby allowing Brca1 to be phosphorylated by ATR. That phosphorylation further enhances the ability of ATR to phosphorylate Chk1. All of those processes may be facilitated by Claspin as a potential scaffold holding the reactants together. Blue: phosphorylated Chk1 phosphorylates Cdc25A at a site that stimulates its degradation, thereby removing the phosphophatase required for activation of Cdk1. Purple: ATR phosphorylates p53, thereby activating the transcription of Cdk2-inhibitor p21^Cip1. The kinetic diversity of the interactions in these pathways may provide robustness to a switch that activates a cell cycle checkpoint.

Cross-talk from ATM

Also at Levels 4 and 5, Figure 3 shows cross-talk from ATM and Chk2 in parallel with ATR and Chk1, as follows:

1. ATR phosphorylates and activates ATM (interaction 60);

2. ATM phosphorylates and activates Chk2 (interaction 79);

3. Chk2 phosphorylates Cdc25A (in parallel with Chk1) and marks it for ubiquitin-dependent degradation (interaction 86).

Discussion

Hierarchical assembly of a MIM of the network that signals blocked replication

A common form of replication stress is characterized by the accumulation of ssDNA due to continued helicase action at the replication fork while the polymerase is blocked. Such helicase-polymerase uncoupling may be due to damage of the template strand or to inhibition of the polymerase by a toxic molecule, such as aphidicolin, or by imbalance of precursor nucleotides. Fundamental to this type of replication stress is signaling from the ssDNA between the progressing helicase and the blocked polymerase. The accumulated ssDNA is detected by RPA, which initiates signals that are transmitted through interlocking pathways in a network of molecular interactions that control polymerase switching for trans-lesion synthesis and Cdk2 activity for cell cycle progression or arrest.

To untangle the complexities of the network, we assembled an orderly molecular interaction map (MIM) using a novel hierarchical procedure (Fig. 2, Table 1). The procedure places the components of the network in levels of interaction hierarchy that are layered from DNA-proximal at the bottom to DNA-distal at the top (Fig. 3). The DNA-proximal layers include the relevant DNA structures (Layer 1) and the components that interact directly with those DNA structures (Layer 2). Components are added at each successively higher level as they are brought in by interactions with components at lower levels. The successively higher levels thus contain components successively further removed from direct interaction with the replication-stressed DNA.

The resulting map is sharply focused through well-defined input and outputs. The input is helicase-polymerase uncoupling; the outputs are switching in of translesion polymerases and inhibition of Cdk2-dependent cell cycle progression. The scope of the map is thus limited to the interactions that contribute to particular network functions, thereby keeping the map as uncomplicated as possible.

Although the input focus of the current map is placed at the bottom, it could in other maps be at the top or in the middle with components layered according to the number of interaction steps required to reach that focus.

Helicase and polymerase are shown on opposite sides of the map, stretched out in a linear representation. Despite their uncoupled actions, however, helicase and polymerase may remain linked in the same “replication factory” while the accumulating single-strand regions loop out between them. That linkage could correspond to Claspin and TIM:Tipin binding of PCNA and MCM (interactions 35, 36 and 27). The hierarchical ordering of events however is unaffected by the geometric arrangement of the diagram.

The assembled hierarchical map revealed several architectural network features that will be discussed next.

Transmission line signaling

The architectural feature that we noticed first was an apparent constriction of the middle layer (Layer 3) of the map, which is occupied mainly by ATR and Claspin (Figs. 3 and 5). ATR and Claspin (together with the TIM:Tipin complex) appear to constitute a principal nexus in the signal transmission mechanism. ATR receives input from lower levels and transmits output through phosphorylations to both higher and lower levels (Fig. 5, red highlight). Claspin (with an assist from TIM:Tipin) binds critical components at both higher and lower levels (Fig. 5, blue highlight), and may serve to hold these components together to facilitate interactions between them. Moreover, ATR and Claspin seem to work together, but with a division of labor. ATR phosphorylates and stabilizes Claspin, and Claspin provides a scaffold that facilitates the actions of ATR. In that division of labor, ATR functions mainly enzymatically, whereas Claspin functions exclusively through binding interactions. Separation of enzymatic and scaffolding functions may be common in networks to increase functional efficiency.

We can think of the architectural constriction as a kind of transmission network that receives signals from many sources and sends them by way of few intermediaries (e.g., via ATR and Claspin) out to many places. A transmission line architecture of this kind seems to exist in many signaling pathways.¹¹ The functional or differentiation state of a cell may be determined in part by the linkage patterns between transmission line inputs and outputs, which may be determined by transcriptional co-selection.

Checkpoint proteins have been divided into “sensor”, “adapter” and “effector” categories.¹² The hierarchical map (Fig. 3), automatically generated a similar division of network proteins into DNA-proximal (Level 2), middle (Level 3) and distal (Levels 4 and 5) zones, respectively (Fig. 3).

Signal initiation by RPA

In regions of replication stress, ssDNA-bound RPA initiates signals to switch between replicative and translesion polymerases and to inhibit cell cycle progression. The signaling path from ssDNA-bound RPA to PCNA monoubiquination and consequent facilitation of polymerase switching is highlighted in Figure 4. ATR appears to be involved only indirectly through its phosphorylation or Rpa2, which may increase the affinity of RPA for ssDNA in regions of replication stress (Fig. 4, interactions 31 and 43). Claspin, on the other hand, has a more direct effect through its binding of PCNA, which facilitates the PCNA monoubiquitation by Rad6:Rad18 (Fig. 4, interactions 44 and 22).

Signals from ssDNA-bound RPA to Cdk2-dependent cell cycle control proceed through at least 3 RPA actions: (1) recruitment of ATR:Atrip with priming (but not full activation) of ATR kinase activity (Fig. 6, blue highlights); (2) binding of TIM:Tipin (interaction 26), through which Claspin is recruited to the site (Fig. 5, blue highlight, interaction 36); (3) recruitment to the ssDNA of Polα, the primase-polymerase that initiates the synthesis of a new DNA strand (interactions 16, 5 and 6), providing binding sites in the DNA primer regions for Rad17, which then load 9-1-1 and recruit TopBP1 for ATR activation (Fig. 6, red highlights). (It is uncertain whether new strand initiation occurs on the leading strand in mammalian cells; the Polα-related steps may occur mainly on the lagging strand, but the signaling consequences may be the same).

This part of the MIM thus describes the RPA functions that recognize and stabilize long ssDNA segments, and recruit to those regions the key molecules required for the next step in signaling replication stress.

Failsafe stimulation of ATR kinase activity

ATR-activating signals come from two proximal sources that act in concert as a “2-man rule” for confirmed action.⁸ This fail-safe requirement for optimal ATR activation depends on concurrent signals from RPA and 9-1-1 complexes, which bind independently to the ssDNA in the region of helicase-polymerase uncoupling (Fig. 6, blue and red highlights, respectively). The pathway from 9-1-1, which activates ATR by way of TopBP1, may have the major direct activating effect, whereas binding of the ATR:Atrip complex to RPA may be required to permit that activation. The requirement for two independent signals to activate a response may guard against deleterious consequences from transient fluctuations.

Kinetic diversity and robustness

The ATR activation mechanism in the hierarchical network map comprises steps of different kinetic character, including (1) phosphorylations or bindings that have direct positive effects and (2) phosphorylations that have indirect effects by inhibiting the degradation of positive components (e.g., inhibited degradation of Claspin and TopBP1). A particularly noteworthy case is an interlocking pair of ATR-activating feedback loops that have TopBP1 as the central component. The direct member of the pair is made up of a sequence of stimulatory binding and phosphorylation steps (Fig. 7, red highlights). The indirect path involves TopBP1 phosphorylation by ATR with consequent TopBP1 stabilization (Fig. 7, blue highlights). Such kinetic diversity may confer robustness on a stimulus-response network.

Inhibition of cell cycle progression by ATR

The multiple pathways by which ATR could inhibit cell cycle progression (Fig. 9) constitutes another case in which kinetic diversity may confer robustness of response that may also be switch-like. Figure 9 shows 4 paths to inhibition of Cdk2-dependent cell cycle progression via ATR, marked by color highlights as follows:

1. ATR-induced Chk1 phosphorylation

   1. with an assist from Claspin (green).
   2. with assists from both Claspin and Brca1 (red).
2. ATR-induced inhibition of Cdk2-dependent cell cycle progression

   1. by way of Chk1 and Cdc25A (blue).
   2. by way of p53 and p21^cip1 (purple).

Here again we see a combination of stimulator and inhibitory actions leading to the same effect, in addition to which there is a transcription-dependent step (induction of p21 by p53). Such kinetic diversity may assure execution of the response over a wide range of concentrations and rate constants.

A possibly relevant case is the intra-S-phase checkpoint response to DNA damage caused by camptothecin-induced blockage of topoisomerase 1, in which the checkpoint response is mainly due to Chk1-induced phosphoryation and degradation of Cdc25A (reviewed in ref. 13). ATR and Chk1 are activated within minutes after DNA damage and are crucial to the events that promptly arrest cell cycle progression.

In the p53 limb, the transcription-dependence of p21 must involve a time delay. That delayed action may extend the duration of cell cycle arrest. Drugs that inhibit Chk1 thus may be selectively toxic to p53-deficient cancer cells treated with DNA damaging agents. Studies combining DNA damaging and Chk1 inhibiting drugs however have given conflicting results, possibly due to the critical importance of scheduling of the relative times of drug administration (reviewed in refs. ¹⁴ and ¹⁵). A therapeutic strategy in which a Chk1 inhibitor is used to pre-treat a p53-deficient tumor (to block the intra-S checkpoint fully) may produce selective tumor cell killing by a subsequently administered DNA-damaging drug. Camptothecin may be suitable for such drug combination therapy, because it produces DNA damage mainly in S-phase cells. This example suggests how an orderly molecular network map can help to clarify issues in the design of therapeutic manipulations and put them in a cogent context within the overall interaction network.

Speculation about segregation of protein tool sets in foci

A myriad of interactions of large numbers of signaling molecules are engaged in diverse functions in the cell medium. If that medium were homogeneous, it might be difficult for molecules required for a particular interaction to find each other. Structural localization of relevant sets of molecules could dramatically enhance efficiency of function. It mkes sense, therefore, that molecules related, for example, to particular DNA repair function do localize in granular structures or “foci”.^16–19

The segregation of binding and enzymatic functions noted in Figure 5 (blue versus red highlights) suggests that relevant molecules could interact in and around such foci. Molecules engaged in multiple bindings would hold the constituents of foci together, while the weaker-binding enzymatic components could dissociate reversibly and have local mobility to act upon each other. Enzymatic interactions could take place in a penumbra region around the foci, where large proteins or complexes would tend to concentrate, due to limited diffusion imposed by molecular crowding.^20,21 The efficiency of network function would be enhanced by the co-localization of the interacting components within the penumbra. The ability of molecules to diffuse through the crowded molecular environment would depend on molecular size. Small molecules, such as Chk1 and Chk2, would be able to diffuse between foci, perhaps coordinate the function of different foci, and transmit signals to effectors.

Based on those considerations, we propose a simplified model of Chk1 signaling from stalled replication forks through a molecularly crowded medium (Fig. 10). ssDNA that accumulates after helicase-polymerase uncoupling becomes coated with RPA (bottom of Fig. 10). The RPA molecules have bound to them either ATR or a complex of TIM:Tipin bearing Claspin, to which Brca1 may be tethered. Those molecules are arrayed in a cluster along the ssDNA. We assume that ATR has already been activated and that RPA has already coated the ssDNA, perhaps enhanced by ATR-dependent RPA phosphorylation (interactions 31 and 43). Inactive Chk1 molecules (small black circles) are small enough to diffuse unimpeded through the molecularly crowded medium toward the molecular cluster around the ssDNA. In order for ATR to phosphorylate and activate Chk1 (interaction 53), Chk1 must make contact with both ATR and Claspin (interaction 69), with or without the participation of Brca1 (Fig. 3). That may take place in a penumbra region near the ssDNA cluster. The molecular crowd confines the large ATR-containing complexes to the penumbra region, while the small activated Chk1 molecules (red-filled circles) can diffuse away. Thus a counter-flow is established; inactive Chk1 molecules diffuse into the foci and activated Chk1 molecules diffuse out.

Figure 10.

(A) Speculative model of activation and signaling in a molecularly crowded environment. The molecules are scaled approximately according to size (except for the proteins of the molecular crowd, which are of fixed but arbitrary size). The molecular crowd confines large molecules, such as ATR, within the penumbra region near the DNA, while small proteins, such as Chk1, can diffuse through the crowd. Unphosphorylated Chk1 (small black circles) diffuses toward the cluster around the ssDNA, becomes activated by phosphorylation within the penumbra region, and diffuses out to convey signals to other parts of the cell. The activated Chk1 molecules are shown as small red circles. (B) MIM of the interactions involved in the model depicted in (A).

To maintain the concentration differences between the penumbra region and the exterior, thermodynamics demands a source of energy. The energy may come from ATR-dependent phosphorylation of Claspin (interaction 41), which facilitates Claspin:Chk1 binding (interaction 65), which in turn inhibits ubiquitin-dependent Claspin degradation (interaction 67) and makes Claspin available to promote ATR-dependent Chk1 phosphorylation (interaction 42). These phosphorylations promote binding of those proteins to form larger complexes that would be more confined by the molecular crowd.

In addition, the multiple binding capability may link multiple cluster assemblies, so as to produce large multi-centric foci within or around which molecules that act on those assemblies would be concentrated.

Molecular crowding may localize “tool-box” components that are required for a particular function in other cases: for example, the localization within γH2Ax foci of molecules engaged in the repair of DNA double-strand breaks, where molecular crowding within and/or around the foci may help also to prevent the ends of the severed DNA from diffusing far away from each other.^18,19

Methods

The hierarchical network map described here uses the symbols defined in Figure 1. The rules of the MIM notation are summarized in Appendix 1. For more detail, see reference ² or http://discover.nci.nih.gov.

The hierarchical procedure (Fig. 2) begins by placing on the map structures or highly connected molecular species (A in Fig. 2) that become the focal start of the map. In the current map, the focal starting structure is the DNA in the region of helicase-polymerase uncoupling, shown in Level 1 of Figure 3. The next step (B in Fig. 2) is to add the molecular species that interact directly with the focal structure; those entities are placed in Level 2 of Figure 3. The procedure then loops between adding interactions between components already on the map (C in Fig. 2) and adding molecular species that interact directly with what is already on the map (D in Fig. 2). The components added at each iteration of the loop are placed in successively higher levels of the map.

In addition to one or more starting structures, the hierarchical map requires one or more focal endpoints. In the current map, a focal endpoint is cell cycle progression and its inhibition by cell cycle checkpoint actions, which are placed in the upper level of the map. The path leading to trans-lesion synthesis ends at the DNA in Level 1.

The loop between processes C and D in Figure 2 continues until all of the components and interactions that may influence the effect of the focal starting point on the focal endpoints have been added. Finally, any components and interactions that do not have known or plausible effects on those influences are removed (E in Fig. 2).

The order in which molecular species and interactions were added to the helicase-polymerase/checkpoint map in the course of the hierarchical procedure is shown in Table 1. Each interaction, as numbered in Figure 3, is described in the Annotations (Appendix 2), together with other salient information and references. Each description explains the meaning of the corresponding interaction symbol in accord with the MIM notation as defined in reference ². A brief description of each numbered interaction is also included in Table 1.

The MIMs presented here are to be read using the “heuristic” interpretation, as defined in reference ⁹; that is, unless otherwise indicated by contingency symbols, multiple interactions (or modifications) shown for a given entity may exist at the same time. Multiprotein entities that remain tightly associated throughout the network (such as the MCM’s, the RPA subunits, or the Rad9:Rad1:Hus1 complex) are grouped within an outer round-cornered box. If an interaction involves an entire content as a group (as far as is known) then the interaction line terminates at the outer (complex-defining) box; if the interaction (or protein modification) is known to be with a particular component or subunit, then the line proceeds to that unit within the box.

Appendix 1: Summary of MIM Notation Rules

The MIM symbols used here are shown in Figure 1. Full details, examples, and applications are presented in reference ² and ³ and at http://discover.nci.nih.gov/mim/.

1. A named molecular species generally appears in only one place on a map. (Exempt from this rule are molecules, such as GTP and ubiquitin, that act in a similar manner in a large number of different reactions.)

2. Interactions between molecular species are shown by different types of connecting lines, distinguished by different types of arrowheads or other terminal symbols (Fig. 1).

3. Interaction lines can change direction (but not by more than 90° at a corner; that restriction prevents ambiguities at branch points).

4. When lines cross, it is as if they did not touch.

5. Symbol definitions are not affected by color, which can be used optionally to emphasize particular features of a network.

6. A small filled circle (“node”) on an interaction line indicates the product of the interaction. Thus the product of binding between two molecules, i.e., a dimer, is represented by a node on the binding interaction line; similarly for multimers in extensible fashion (see reference ²). The product of a modification (e.g., phosphorylation) is the modified (e.g., phosphorylated) molecule; therefore, the phosphorylated product is represented by a node placed on the modification line. For convenience and legibility, multiple nodes can be placed on the same line and represent the same entity.

Appendix 2: Annotations

(Interaction numbers are indicated in brackets.)

• [1]The MCM helicase, consisting of the subunits MCM2 through MCM7, binds as a ring encircling the DNA double helix at the replication fork.¹ The MCM complex is shown with an outer border, within which only those subunits are shown (in this case MCM4, 5 and 7) that will have interactions shown in the map. The MCM helicase is an essential part of the machinery that unwinds the DNA at the replication fork.²
• [2]The helicase action is represented by the arrow labelled ‘H’, and the stimulation of the helicase action by DNA-bound MCM is represented by the open-triangle arrow. The helicase action separates the DNA double helix into two single strands: the upper one shown in the map is the template for the forwardly synthesized (“leading”) DNA strand. The lower is the template for “lagging” strand synthesis, which takes place discontinuously in a direction opposite to the direction of helix unwinding, producing “Okazaki fragments” that become ligated to from the intact lagging daughter strand. The DNA unwinding mechanism is still not completely understood, and several possible mechanisms have been considered. ³ The current map focuses on leading-strand events, but similar polymerase blockage events may occur on the lagging strand. For a MIM of the assembly and activation of the MCM complex in the context of the initiation of replication origins, see http://discover.nci.nih.gov/mim/.⁷⁷
• [3]The GINS complex, consisting of subunits Sld5, Psf1, Psf2 and Psf3, binds DNA at the replication fork, closely associated with the MCM complex. Only subunit Psf2 is shown explicitly; it has a specific interaction that will be added later. Human GINS is a ring or horseshoe-shaped DNA-binding structure.^4–6 The crystallographic GINS ring may accommodate ssDNA, but not dsDNA,^5,7 and GINS binds preferentially to DNA that has an ssDNA component.⁴ Accordingly, we show GINS as a ring around a single strand as it emerges from the MCM helicase, similar to model A proposed by Boskovic et al.⁴ GINS possibly directs one of the unwound strands in a manner that would prevent re-annealing (our speculation).
• [4]RPA is a single-strand DNA binding protein complex that binds to long stretches of ssDNA, such as accumulate during helicase-polymerase uncoupling. RPA binds to ssDNA that accumulates when replicative polymerase is blocked by a template lesion, inhibitor, or metabolic perturbation while the helicase machinery forges ahead leaving in its wake long stretches of ssDNA (reviewed in ref. 8). The ssDNA accumulated as a result of such helicase-polymerase uncoupling becomes coated with RPA.⁹ RPA consists of 3 subunits: RPA1, 2 and 3, also known as RPA70, 32 and 14, respectively. Only subunit RPA2 is shown explicitly in the MIM.
• [5,6]Polymerase-primase Polα may bind to the accumulated ssDNA, assisted by the ssDNA-bound RPA, and may initiate the synthesis of a new DNA strand segment.¹⁰ RPA accumulation on long stretches of ssDNA is required for priming a new strand by Polα (reviewed in ref. 11). Polα first synthesizes a short stretch of complementary RNA (not shown), after which the polymerization switches to DNA synthesis. The RNA segment is then degraded (not shown), leaving a ss/ds DNA junction. The question mark under the Polα binding site signifies uncertainty of whether, in mammalian cells, priming of new DNA strands can occur on the leading strand, or whether it occurs only on the lagging strand (see [19]).¹²
• [7]Clamp-loader Rad17-RFC binds at ss/dsDNA junctions. ^11,12 An ss/dsDNA junction forms when the RNA primer segment (synthesized in the first stage of Polα action) is degraded. The incipient ss/ds junction (not shown explicitly in the MIM) is at the solid circle to which the Rad17-RFC binding line points. That is also where Polα-catalysed synthesis (boxed Prm in the MIM) switches to Polδ/ε-catalysed synthesis (boxed P). Production of ss/ds junctions by Polα may be required for checkpoint activation via ATR.¹¹ Initiation of new DNA strands on ssDNA downstream from the lesion has been demonstrated in yeast and may generate the ss/dsDNA junctions required for Rad17-RFC binding (discussed by Jansen et al.¹³).
• [8,9]DNA-bound Rad17-RFC loads a ring-shaped complex consisting of subunits Rad9, Rad1 and Hus1 (commonly called “9-1-1” complex) at ss/dsDNA junctions.^11,14–16 Rad17 is required for recruitment of 9-1-1 to chromatin in response to DNA damage.¹⁷ Recruitment (loading) of 9-1-1 is independent of ATR, suggesting that Rad17 phosphorylation is not required.¹⁷ (Previous reports of the essentiality of phosphorylation for all Rad17 functions were not confirmed.) The 9-1-1 DNA clamp is structurally homologous to PCNA.¹⁸
• [10]The 9-1-1 ring may glide along the DNA in association with a replicative DNA polymerase, and, like PCNA, keep the polymerase bound to the template as replication proceeds.¹⁰
• [11]PCNA (consisting of 3 PCNA monomers) binds as a mobile ring surrounding the DNA near the growing end of the daughter strand.¹⁹ PCNA keeps the replicative polymerases, PolΣ and Pol™, bound to the DNA, thereby allowing the synthesis of long DNA strands [14, 15] (“high processivity”).
• [12]Synthesis of the DNA strand growing in the direction of the replication fork (“leading strand” synthesis) is thought to be catalyzed by Polδ. (Synthesis of the strand growing in the opposite direction (“lagging strand”) is probably catalyzed by Polε. The MIM shows in detail only events relating to the leading strand).
• [13]PCNA keeps the replicative polymerase Polδ bound to the DNA at the site of leading strand synthesis, thereby allowing the synthesis of long DNA strands (“high processivity”). (A similar process occurs on the lagging strand, where Polε may act in place of Polδ, shown in the MIM in abbreviated fashion). PCNA is loaded onto DNA at replication origins (not shown), closely associated with DNA polymerases, and remains associated with the polymerase as replication progresses.
• [14]When the replicative polymerase encounters a template lesion, such as some types of chemical damage to DNA bases, the progress of the polymerase is blocked. This is shown in the MIM as an inhibition of the Polδ-catalyzed leading-strand synthesis. The DNA lesion is represented by the encircled × on the template strand.
• [15]Trans-lesion polymerases, such as Polη, Polι, Polκ and Rev1, are able to progress over particular types of template lesions. Trans-lesion synthesis is denoted by “TLS” attached to the replication arrow in the MIM. Trans-lesion polymerases have low fidelity and are error-prone. When the replicative polymerase is unable to proceed, its place on PCNA can switch to a trans-lesion polymerase. When the progress of replicative polymerase Polε or Polδ is blocked by a template lesion, a trans-lesion polymerase is recruited via PCNA, and inserts one or more nucleotides across the template lesion, thereby overcoming the replication block (reviewed by Lehmann et al.²⁰). The trans-lesion polymerases differ with respect to the types of lesions they can surmount. The reduced specificity of trans-lesion polymerases increases the error rate; hence it is not surprising that trans-lesion polymerases insert only a few bases before being replaced by a high fidelity replicative polymerase.
• [16]ssDNA-bound RPA assists polymerase-primase Polα to bind ssDNA and initiate the synthesis of a new DNA strand segment.¹⁰ We found it convenient to show this on the leading strand. In metazoan cells, however, it may occur on the lagging strand, where new strand synthesis is normally initiated periodically. ¹² Therefore, we have placed a question mark under the Polα binding site on the DNA. The model requires that the a strand initiation site with a ss/dsDNA junction and a 5' DNA terminus occur within molecular complex range of the site of Atrip:ATR binding to ssDNA:RPA.
• [17,18]

  Rad6:Rad18 (a ubiquitin ligase E2 and E3 combination) binds to RPA in regions of accumulated ssDNA.²¹ Rad18 accumulates rapidly and persists in DNA damage-induced regions of ssDNA.²²

• [19–21]

  PCNA can bind replicative polymerases (Polδ or Polε) [19] or trans-lesion polymerases (Polη, Polι, Polκ or Rev1) [20]. Binding of TLS polymerases to PCNA promotes trans-lesion synthesis (TLS) [21]. Each subunit of the PCNA homotrimer has a binding site that can be occupied by polymerases (or by other proteins); the binding site is located in a groove buried under the PCNA inter-domain connecting loop.¹⁹ The branched binding notation indicates that the different polymerases may bind as alternatives at a given PCNA site. On the leading strand, PCNA enhances the DNA polymerization catalyzed by Polε, the presumed replicative polymerase for the leading strand. PCNA has an analogous action on Polδ in the lagging strand.

• [22]DNA-bound PCNA can bind the Rad18 component of a Rad18:Rad6 dimer.² Rad18 is an E3 ubiquitin ligase, while Rad6 is an E2 Ub conjugating enzyme.²³
• [23]When bound to ssDNA, the Rad18:Rad6 dimer adds a single ubiquitin unit to K164 at each PCNA site.²⁰ This reaction is facilitated by the binding of Rad18:Rad6 to ssDNA near its PCNA target. Rad18:Rad6 can monoubiquitnate all 3 monomers of DNA-bound PCNA.²³ The functional unit is a dimer of Rad18:Rad6 heterodimers (not shown).²³ Rad18 knockout cells have normal growth rate, but have increased sensitivity to UV, HU, cisplatin, mitomycin and MMS, but not to IR.^20,24 PCNA monoubiquitination occurs in response to UV, MMS, mitomycin, cisplatin, hydrogen peroxide and benzo[a]pyrene diolepoxide, but not after IR, bleomycin or neocarcinostatin (which do not slow down replication forks), nor after actinomycin, daunorubicin, or nocodazole.²⁰
• [24]Rad18 may recruit Polη to stalled replication forks.²²
• [25]GINS binds tightly to MCM:⁷ the MCM5 subunit of the MCM complex binds the Psf2 subunit of the GINS complex.¹⁹
• [26]The Tipin subunit of TIM:Tipin binds RPA via a site on the RPA2 subunit,^9,25 consistent with the observation that TIM:Tipin associates with chromatin only during S phase.²⁶ TIM and Tipin form a tight complex involving amino acids 67–143 of Tipin.²⁶ TIM and Tipin are required for the intra-S checkpoint response to DNA damage, but the TIM:Tipin complex exists whether or not DNA damage is present.⁹ The TIM:Tipin interaction involves amino acids 67–143 of Tipin and amino acids 1–267 or 267–673 of TIM.²⁶
• [27]TIM:Tipin, by way of Tipin, binds the MCM complex in the presence or absence of replication stress.^9,27 We do not know which subunit(s) of MCM constitute the binding site for Tipin, nor whether Tipin binds RPA and MCM independently or in competition. TIM and Tipin are associated with the mammalian replication complex at progressing replication forks,²⁵ perhaps by way of their binding to MCM and/or RPA.
• [28]TIM and Tipin co-localize with PCNA especially during late S phase (but direct binding not established).²⁶ Knockdown of Tipin or TIM reduces DNA synthesis rate (H3/C14 in undamaged cells) (ref. ⁹; Fig. 3). Tipin appears to be required for efficient S phase progression.²⁶ Knockdown of TIM or Tipin reduces the percentage reduction of DNA synthesis rate that occurs after UV (ref. ⁹; Fig. 3B). Both TIM and Tipin are required for UV to block initiation of new origins (ref. ⁹; Table 1)
• [29]ATR:Atrip (a tight complex between ATR and Atrip) binds, by way of its Atrip subunit, to the RPA2 subunit of RPA.²⁸ RPA enables the binding of ATR:Atrip (via Atrip) to ssDNA.²⁹ A conserved acidic domain in the N-terminal region of Atrip engages a basic cleft in the RPA70 subunit.³⁰
• [30]ATR:Atrip, by way of its Atrip subunit, can bind the MCM complex by way of its MCM7 subunit.^28,31 We do not know whether Atrip can bind RPA2 and MCM7 simultaneously or in competition. The minimal Atrip-binding region of MCM7 is 577–719; ATRIP antibodies immunoprecipitated both MCM7 and ATR; MCM7 antibodies immunoprecipitated ATR; hence MCM7 can bind ATR via ATRIP.³¹
• [31]ATR phosphorylates RPA2 at multiple sites.³²
• [32]ATR phosphorylates Rad17 (at Ser635 and Ser645); this is required for response to genotoxic stress, although not required for binding of Rad17 to chromatin.^33,34 In response to DNA damage, Rad17 is phosphorylated by ATR at S635 and S645, sites that are in SQ motifs known to be strong targets of ATR.¹⁷ Rad17 phosphorylation is not required for the survival of undamaged cells, but is required to avoid spontaneous chromosome breaks.¹⁷ Cells lacking Rad17 phosphorylation have a shortened S-phase, suggesting that Rad17 phosphorylation is part of a checkpoint that delays entry into mitosis until all of the genomic DNA has been properly replicated, thereby preventing chromosome damage.¹⁷
• [33]ATR may phosphorylate Rad9 in the 9-1-1 complex.¹⁰ Phosphorylation of 9-1-1 is not required for 9-1-1 binding to chromatin.
• [34]ATR phosphorylates MCM2, MCM3 and MCM4.^32,35 MCM2 is phosphorylated by ATR on Ser108 in response to DNA damage or stalled replication forks.³¹ In response to UV or HU, ATR phosphorylates MCM2 at S108 and MCM3 at S535. The functional consequences of the MCM2 and MCM3 phosphorylations is not clear.
• [35]Claspin binds PCNA.³⁶
• [36]Claspin associates avidly with TIM, suggestive of direct binding.²⁵ TIM:Tipin may recruit Claspin to chromatin in response to replication stress.²⁶
• [37]Cdc45 binds to the MCM5 subunit of MCM and the Psf2 subunit of GINS, perhaps binding to the MCM:GINS combination, as shown in the MIM.³⁷ GINS appears to be required for the association of Cdc45 with MCM.⁷
• [38]Cdc45 may associate with Polε or Polδ. MCM, GINS and Cdc45 may all be part of the multimolecular machinery at elongating replication forks.³⁷
• [39]Claspin binds phosphorylated Rad17.¹⁷ We do not know whether this binding and binding of Claspin to PCNA [35] can occur at the same time.
• [40]Phosphorylated Rad17 (presumably still bound to 9-1-1 after clamp-loading) stabilizes the association of 9-1-1 in foci associated with sites of replication stress.³⁸
• [41]ATR phosphorylates Claspin at T916 and S945 in the Chk1-binding domain, probably indirectly.^12,39 The identity of the ATR-dependent kinase is not known, but it is not Chk1.⁴⁰ Human Claspin contains 3 ATR/ATM-dependent phosphorylation sites (within 3 copies of consensus ExLxLC(S/T)GxF in the region 910–985, the Chk1 binding domain); at least 2 of these phosphorylation sites are required for Chk1 binding and consequent Chk1 phosphorylation (see below), demonstrated in a human cell-free system.⁴¹ These sequences however are not consensus phosphorylation sites for ATR/ATM. These sites can be phosphorylated by Chk1 in cells.⁴² According to Chini and Chen⁴² human Claspin is phosphorylated at T916 in response to UV or replication stress; however, Bennett et al⁴⁰ doubt that Chk1 is the major responsible kinase.
• [42]Claspin binding to Rad17 facilitates the ATR-dependent phosphorylation of Claspin, enhances the S-phase checkpoint response to DNA damage, and improves survival of cells subjected to hydroxyurea or UV.¹⁷ Claspin (as well as Chk1) are required for normal replication fork progression, especially in fast-growing cell lines.⁴³
• [43]Phosphorylation of RPA2 may enhance RPA binding to sites of replication stress at the expense of binding to normal replication sites.³²
• [44]Claspin enhances binding between PCNA and Rad18,³⁶ (see Discussion). Claspin and TIM are required for efficient ubiquitination of PCNA (mechanism unknown).³⁶ PCNA monoubiquitination (a consequence of PCNA binding to Rad18) is dependent on Chk1 (perhaps by stabilizing Claspin), but not on Chk1 kinase activity nor on ATR.³⁶
• [45]ATR activation is enhanced by its association with RPA and thereby with ssDNA.¹²
• [46]Phosphorylation of MCM4 reduces the affinity of the MCM complex to chromatin and thereby reduces the helicase activity of the MCM complex.⁴⁴
• [47]Cdc45 in complex with the GINS:MCM complex may support the helicase action of MCM at elongating replication forks.³⁷
• [48]Monoubiquitination of PCNA enables its binding to Polη and switches this error-prone trans-lesion polymerase into action. Other trans-lesion polymerases, such as Polι, can also bind to monoubiquitinated PCNA and may be switched into action one after another.^19,45,46 Monoubiquitination increases PCNA binding of Y-family TLS polymerases (Polη, Polι, Polκ and Rev1, all of which have ubiquitin-binding domains), stimulates polymerase switching, and stimulates trans-lesion synthesis (TLS).²⁰ These polymerases have low processivity, and are error-prone partly due to lack of 3'–5' exonuclease proofreading function. PCNA normally prefers to bind the replicative polymerases Polε or Polδ. But when modified by the addition of a single ubiquitin unit (Ub¹), PCNA’s preference switches to favoring trans-lesion polymerases.⁸ The ubiquitin moiety on PCNA is recognized by all Y-family translesion polymerases: Polη, Polι, Polκ and Rev1, all of which have ubiquitin-binding domains (reviewed in ref. 20). PCNA ubiquitination has been observed after UV, MMS, HU and aphidicolin, and these treatments (together with cofactor ATRIP) are known to activate ATR (ref. ⁸).
• [49]The Rad9:Rad1:Hus1 (“9-1-1”) complex recruits TopBP1 to newly initiated strands on ssDNA accumulated during helicase-polymerase uncoupling. TopBP1 binds the 9-1-1 complex via phosphorylated Rad9. The Rad9 phospho-serine-387 site binds avidly to TopBP1 at the tandem Brca1-domains 1 & 2 in the TopBP1 N-terminal region.¹⁰ S387 is conserved among vertebrates and constitutively phosphorylated (perhaps due to the modest level of ATR kinase activity normally present during S phase).
• [50]TopBP1 (via its activation domain in the C-terminal region) binds ATR:Atrip. TopBP1 and ATR co-localize tightly with RPA at ssDNA.³⁹ Atrip contains a conserved TopBP1-interacting region that is required for the association of TopBP1 with ATR and for the full activation of ATR by TopBP1.⁴⁷ This interaction is required for cells to survive and recover DNA synthesis after replication stress.⁴⁷ TopBP1 can bind and activate ATR:Atrip without ATR:Atrip binding to RPA; nevertheless the interaction between TopBP1 and ATR:Atrip would be facilitated by their juxtaposition at DNA structures due to RPA-dependent bindings.³⁰ The binding of ATR:Atrip to ssDNA-associated Rad9 (in the 9-1-1 complex) may serve the same purpose.⁴⁷
• [51]ATR phosphorylates TopBP1 at S1159.⁴⁸ ATR may phosphorylate TopBP1 in response to replication stress, thereby increasing the availability of TopBP1 for activation of ATR and possibly providing positive feedback activation of ATR.¹²
• [52]Chk1 binds Claspin directly.⁴⁹ Claspin improves colony survival of irradiated U2OS cells and helps to moderate DNA synthesis after IR.⁵⁰
• [53]ATR phophorylates Chk1 at Ser317 and Ser345, and thereby activates Chk1.⁴⁹ This phosphorylation occurs in a manner that is dependent on TopBP1, as well as Claspin.^39,43 Chk1 activation requires binding of ATR:Atrip to ssDNA via RPA (via interactions 5 and 28).²⁹ Chk1 phosphorylation is facilitated by phosphorylation of 9-1-1,¹⁰ presumably by way of Rad9 phosphorylation, TopBP1 and ATR. Consistent with that presumption is that, when Rad9 is physically linked to TopBP1 (or fragments thereof ), a TopBP1 region between Brca1 domains 6 and 7 (a region dubbed “activation domain”) is necessary and sufficient for HU-initiated Chk1 phosphorylation.¹⁰ Rad17 is required for Chk1 activation in response to HU and UV (Rad17 phosphorylation was required for survival after HU, but not UV).¹⁷
• [54]Cdc7 binds Claspin (directly or indirectly), even in an unperturbed cell cycle.⁵¹
• [55]Cdc7 phosphorylates Claspin.⁵¹ Depletion of Cdc7 impaired hydroxyurea-induced Claspin phosphorylation and chromatin association.⁵¹ The relevant phosphorylation site(s) have not been determined.
• [56]Brca1 binds Claspin, especially after DNA damage by ionizing radiation.⁵⁰
• [57]Claspin and the ATR pathway control Brca1 phosphorylation at S1524.⁵⁰
• [58]βTRCP-SCF ubiqitinates Claspin and marks it for degradation. ⁵² Claspin is normally degraded at the onset of mitosis.⁵² The ubiquitin-proteasome pathway has an essential role in recovery from the S-phase checkpoint, with Claspin as the primary target.³⁴ Other targets of this degradation pathway are Wee1 and Cdc25A.³⁴ βTRCP-dependent Claspin degradation facilitate recovery from DNA cell cycle arrest due to DNA damage or replication stress.^52,53
• [59]USP28 is a deubiquinating enzyme that removes ubiquitin chains from several checkpoint proteins, including Claspin and Chk2; in the absence of USP28, Claspin and Chk2 are destabilized.³⁴ USP28 siRNA diminishes Claspin after IR.⁵⁴ Another deubiquitinating enzyme, USP7/HAUSP, however is more efficient than USP28 in the control of βTRCP-dependent ubiquitnation of Claspin.⁵³
• [60]ATR phosphorylates ATM at S1981 in response to hydroxyurea or UV and thereby activates the kinase activity of ATM.⁵⁵ In contrast to the activation of ATM by ionizing radiation, which requires ATM autophosphorylation and binding to Nbs1, the activation of ATM by hydroxyurea of UV does not have these requirements and proceeds instead via ATR.⁵⁵
• [61]ATR phosphorylates p53 at multiple sites.⁵⁶
• [62]Binding between TopBP1 and phosphorylated Rad9 enhances the binding between TopBP1 and Atrip in the ATR:Atrip complex. 9-1-1 (via Rad9) recruits TopBP1, which then triggers ATR-mediated phosphorylation and activation of Chk1.¹⁰ More precisely: the binding of TopBP1 (via its activation domain in the C-terminal region) to ATR:Atrip requires the binding of TopBP1 (via its Brca1-2 region near the N-terminus) to the Rad9:Rad1:Hus1 (9-1-1) complex.¹⁰ TopBP1 bridges between 9-1-1 and ATR. In this way, TopBP1 and 9-1-1 serve both to recruit ATR to newly primed DNA strands adjacent to ssDNA and to stimulate ATR kinase activity.
• [63]TopBP1 binding to ATR:Atrip is required for kinase activity of ATR. ATR has a basal kinase activity that is enhanced when ATR is bound to Atrip.⁴⁷ Activation of ATR by TopBP1 does not require RPA.³⁰ ATR phosphorylates many proteins in diverse signaling pathways.⁵⁷ Matsuoka et al. reported over 700 proteins that become phosphorylated at (S/T)Q sites in response to ionizing radiation. The (S/T)Q sequence is required for ATR and ATM target sites. Much work remains to be done to determine which of these proteins actually function in ATR or ATM response pathways. Many proteins may happen to have sites that fit the rather general sequence requirements for ATR/ATM targets, but phosphorylation of these sites may do little or no harm and may have been tolerated in evolution.
• [64]Polybiquitnation of Claspin stimulates proteasome-dependent Claspin degradation.⁵⁸ Claspin is continually degraded via polyubiquitination and proteasomes. Claspin degradation is a key event in checkpoint recovery.⁵⁸
• [65]Phosphorylation of Claspin enhances its binding to Chk1, although weak transient binding may occur without phosphorylation—the Chk1-inhibitor UCN01 inhibits this phosphorylation in HU-treated cells.⁴² Binding of Claspin to Chk1 requires phosphorylation of at least 2 highly conserved Claspin sites: Thr916 and Ser945.^12,39,42 Binding of Chk1 to Claspin requires TopBP1,³⁹ presumably because TopBP1 is needed to activate ATR, which can then phosphorylate Claspin. Claspin and Chk1 are both required for normal replication fork progression, especially in fast-growing cell lines.⁴³
• [66]Claspin phosphorylation enhances binding to Brca1.⁵⁰
• [67]Chk1:Claspin binding stabilizes Claspin (dependent neither on ATR nor on Chk1 kinase activity).^36,58,59 Claspin is transiently stabilized soon after DNA damage (UV, hydroxyurea, etoposide, camptothecin).⁵⁸ (The MIM shows this stabilization as an inhibition of ubiquitin-dependent Claspin degradation).
• [68]Cdc7 stabilizes Claspin, presumably by phosphorylating Claspin.⁵¹ Depletion of Cdc7 reduced Chk1 activation (presumably via decreased Claspin function) early after hydroxyurea, while DNA synthesis was only moderately suppressed. Cdc7 is required for the acute response to replication block, where DSB were not significant.⁵¹
• [69]Chk1:Claspin binding is required for efficient phosphorylation of Chk1 at Ser317 and Ser345 by ATR.^43,50 Claspin is required for Chk1 activation in reponse to hydroxyurea.⁶⁰ Claspin was required for the intra-S and G₂/M checkpoint responses to ionizing radiation.⁵⁰ Chk1 S317 phosphorylation in response to replication stress is enhanced by Rad17,¹⁷ perhaps by recruiting Chk1 to ss/ds DNA junction sites via Claspin and phospho-Rad17 [interactions 39 and 52], thereby bringing Chk1 close to ATR at those sites. Claspin stimulates Chk1 phosphorylation at S317 and S345 in response to IR, UV or HU.³⁹
• [70]Brca1 binding to Claspin stimulates Brca1 phosphorylation by ATR. Claspin is required for IR-dependent Brca1 phosphorylation.⁵⁰
• [71]Chk1:Claspin binding stabilizes Chk1.³⁶
• [72]Brca1 and the ATR pathway stimulate Chk1 phosphorylation. ⁵⁰ Both Brca1 and Claspin are needed for enhancement of IR-induced Chk1 phosphorylation.
• [73]Phosphorylated ATM phosphorylates p53 (reviewed in ref. 61).
• [74]USP28 binds 53BP1; this binding is stimulated by UV, IR or hydroxyurea.⁵⁴
• [75]53BP1 is phosphorylated at multiple sites in an ATR and/or ATM-dependent manner in response to IR or UV.^34,62
• [76]We do not know how the binding of 53BP1 to USP28 or the phosphorylation of 53BP1 by ATR affect the deubiquination of Claspin by USP28. Since ATR enhances signaling mediated by Claspin, it seems plausible to suggest that phosphorylation of 53BP1 by ATR may stimulate USP28-induced Claspin deubiqutination.
• [77]Chk1 phosphorylates Cdc25A & B. Chk1 phosphorylates and downregulates several replication factors, particularly the Cdc25’s that remove inhibitory phosphorylations from the replication and G₂/M facilitators, Cdk2 and Cdk1.⁶³ The Chk1-dependent S-phase checkpoint involves inhibition of both the initiation and elongation of DNA replication units.⁶⁴ Chk1 (as well as Claspin) is required for normal replication fork progression, especially in fast-growing cell lines.⁴³ Chk1 facilitates homologous recombination repair after replication stress by phosphorylating and thereby activating Rad51.⁶⁵
• [78]bTrCP ubiquinates and stimulates degradation of Cdc25A & B.⁶⁶
• [79]ATM that has been phosphorylated and activated by ATR then phosphorylates and activates the kinase Chk2, which then cooperates with Chk1 to initiate cell cycle checkpoints.⁵⁵
• [80]p53 stimulates transcription of p21^cip1.67
• [81]p53-stimulated p21 transciption is enhanced by p53 S15 phosphorylation.⁶⁷
• [82]HCLK2 facilitates the binding between ATR and TopBP1, thereby enhancing the kinase activity of ATR.⁶⁸ HCLK2 (residues 292–603) binds ATR and stimulates (is required for) ATR autophosphorylation (not shown) and activation. Both HCLK2 and TopBP1 are required to activate the kinase activity of ATR, although they may function at different steps of the activation porcess.^68,69
• [83]TopBP1 binds human HYD (also known as EDD).⁷⁰
• [84]Human HYD ubiquitinates TopBP1, leading to its degradation. ⁷⁰ Presumably this occurs subsequent to HYD binding to TopBP1.
• [85]HYD binding to TopBP1 may be inhibited by ATR-induced TopBP1 phosphorylation. We presume that TopBP1 phosphorylation by ATR may inhibit the binding of TopBP1 to HYD, since DNA damage inhibits TopBP1 ubiquitnation and degradation?⁷⁰
• [86]Although Chk1 and Chk2 phosphorylate many of the same sites of Cdc25, the role of Chk2 in Cdc25A degradation and S76 phosphorylation has recently been questioned.⁷¹
• [87]ATM-induced Chk2 phosphorylation stimulates Chk2-induced phosphorylation of Cdc25A & B.⁷²
• [88]ATR-induced Chk1 phosphorylation stimulates Chk1-induced phosphorylation of Cdc25.^73,74
• [89]Cdc25 phosphorylation at S76 by Chk1 stimulates Cdc25 ubiquitination; however other phosphorylation sites and other kinases may also be involved.⁷⁵
• [90]p21 binds Cdk2.⁷⁶
• [91]p21:Cdk2 binding inhibits the kinase activity of Cdk2.⁷⁶
• [92]Phosphorylation of Cdk2 at T14 and Y15 inhibits the kinase activity of Cdk2 and blocks its ability to promote S phase.⁷⁶ These phosphorylations are attributed to the dual-specificity kinase, Wee1 (not shown).
• [93]Cdc25A and B are dual-specificity phosphatases that remove the inhibitory T14-Y15 phosphates from Cdk2.⁷⁶
• [94]Cdk2 phosphorylates several proteins that promote cell cycle progression.⁷⁶ Cdk2 facilitates progress of cells through S phase by way of Rb-E2F pathways (not shown).