Mouse models of PAK function.

p21-activated kinases are a family of highly conserved protein serine/threonine kinases that are increasingly recognized as playing essential roles in a variety of key signaling processes. Genetic analyses in mice, using constitutive or regulated gene disruption, have provided important new insights into PAK function. In this paper, we review the genetic analysis of all six PAK genes in mice. These data address the singular and redundant functions of the various PAK genes and suggest therapeutic possibilities for small molecule PAK inhibitors or activators.

Introduction

In mammals, the p21-activated kinase (PAK) family consists of two clearly defined subgroups, termed I and II., Group I PAKs are closely related by structure and sequence, whereas the group II PAKs are more divergent, suggesting a more ancient evolutionary origin. In man and mouse, both subgroups contain three members with distinct expression patterns (Table 1). The functions of these genes have been extensively probed in a variety of experimental settings, including gene knockdown and gene knockout in cells or organisms ranging from yeast to mice. In this chapter, we will discuss the genetic analysis of group I and II PAKs in mammals, as much new has been learned over the last few years with the increasing availability of genetically engineered mice bearing constitutive or regulated loss of PAK alleles.

Table 1. PAK tissue distribution and knockout phenotype

PAK isoform	Tissue expression	Knockout phenotype
PAK1	Brain, muscle and spleen	Immune defects, glucose homeostasis defects, neuronal defects, cardiac hypertrophy upon pressure overload
PAK2	Ubiquitous	Lethal E8.0
PAK3	Brain	Abnormalities in synaptic plasticity, defects in learning and memory
PAK4	Ubiquitous, high levels in prostate, testis and colon	Lethal E11.5, neuronal defects, improperly formed vessels
PAK5	Brain	Viable, healthy
PAK6	Testis, prostate, brain, kidney and placenta	Viable, healthy
PAK1/3		Loss of brain volume, impaired learning and memory, hyperactive behavior
PAK5/6		Deficits in learning and locomotion

In general, both group I and group II PAK gene expression can be divided into one of two types: ubiquitous or restricted. PAK2 and 4 are both ubiquitously expressed, and, as will be discussed, these genes are required for embryonic development in mice. PAK1 is highly expressed in the brain, muscle and spleen, whereas PAK3 is expressed predominantly in the brain. Expression of PAK5 is also restricted to the brain, while PAK6 is expressed mainly in the testis, prostate, brain, kidney and placenta. The wide range of tissue distribution speaks to the expansive, and likely non-overlapping, biological functions of the PAK family members.

To understand the specific function of each member of the PAK family, individual and, on occasion, compound mutations have been constructed in the mouse. In the sections below, we will review what is known of the genetics of each of the six PAK genes in mice, and, where data are available, in man.

Group I PAKs

PAK1

PAK1 mice are fertile, viable and have a normal life span. However, these mice display subtle immune deficiencies, mainly manifesting in bone-marrow-derived granulated cells, including mast cells, eosinophils and basophils.PAK1 mice have normal peripheral blood indices and, in culture, bone marrow-derived mast cells (BMMCs) from PAK1 knockouts differentiate normally. However, PAK1 BMMCs fail to respond to stimulation with either IgE or Kit ligand., When antigen-sensitized wild-type BMMCs are challenged with IgE, the activated Fcγ receptor triggers rapid release of histamine containing granules. This response is greatly blunted in PAK1 BMMCs, which show a 3-fold reduction in release of granule contents as compared with control. Similarly, PAK1 BMMCs show an inability to disassemble F-actin after allergen stimulation, a function required for degranulation. These data place PAK1 in a critical position for mast cell degranulation, through regulation of F-actin disassembly. Importantly, the cellular defects also manifest in vivo, as Allen et al. showed that passive cutaneous anaphylaxis is greatly reduced in PAK1 knockout mice, consistent with a defect in mast cell degranulation.

The signaling defects that underlie these phenomena are not completely understood, but are likely to include attenuated activation of the ERK, JNK and p38 pathways, as these pathways are known to affect mast cell function and PAK1 is required for their activation., Mast cell motility is also affected by PAK1, as demonstrated in the context of heterozygosity of the Nf1 gene. Such Nf1 mast cells are normally hyperresponsive to kit ligand due to dysregulated Ras activation, and this augmented response is thought to play a pathogenic role in NF1 syndrome. Loss of PAK1 in this setting inhibits the motility of mast cells, raising the possibility that PAK1-directed small molecule inhibitors might be of benefit in NF1-related pathologies.

Certain other immune cells are also known to be affected by loss of PAK1. For example, PAK1 macrophages display numerous, unstable lamellipodia upon adhesion, accompanied by attenuated ERK activation in response to CSF1.

In addition to the immune defects, PAK1 mice have defects in glucose homeostasis. Such mice display significant impairment of insulin secretion, in the absence of detectable defects in islet architecture or insulin content. These mice also display abnormal glucose clearance, with significantly higher glucose levels at 30 and 60 min in a glucose tolerance test, with fasting glucose levels comparable to those of wild-type mice. An insulin tolerance test also showed significantly higher levels of glucose at all time points suggestive of peripheral insulin resistance. Consistent with this idea, Wang et al. showed that, in skeletal muscle, GLUT4 does not translocate to the plasma membrane following insulin stimulation of PAK1 mice. Interesting, loss of PAK1 was found to be associated with distinct signaling defects in islet cells as compared with skeletal muscle cells. In islet cells, loss of PAK1 did not affect phosphorylation of cofilin (a substrate of the PAK1 target, LIMK), but did lead to loss of ERK activation, whereas the opposite pattern was seen in skeletal muscle. In this latter case, defective cofilin activation, with resultant F-actin misregulation, might explain the observed failure of GLUT4 translocation in PAK1-null mice. This hypothesis also is consistent with the previously noted role of PAK1 in mast cell degranulation and F-actin depolymerization in mast cells, and might indicate a general defect with vesicular traffic in PAK1-null cells.

Interestingly, in humans, PAK1 expression levels in type 2 diabetic donor islets have been shown to be significantly reduced, while Cdc42 and RhoGDI expression remain unchanged. These data, together with the observed defects in insulin secretion and glucose tolerance in PAK1-null mice, suggest that PAK1 may play a role in glucose metabolism in man, and might represent a therapeutic target in certain forms of diabetes.

In the brain, the PAK1 activator Rac1 is known to be involved in dendritic spine formation, raising the possibility that PAK1 is involved in neuron structure and function. Asrar et al. noted that PAK1 gross brain structure is normal and, at the cellular level, PAK1 cells show normal synaptic and spine structures, suggesting that PAK1 is not involved with the formation of synaptic structures. However, despite their normal appearance, neuron function, or synaptic plasticity, which can be described by both long-term potentiation (LTP) and long-term depression (LTD), is notably altered in PAK1-null mice. Such mice are deficient in LTP, but not LTD, suggesting a selective role for PAK1 in synapse function. Further analysis showed that, compared with wild-type mice, whose dendritic spines are highly enriched in F-actin, polymerized actin is scarce in the spines of PAK1-null mice. In addition, phosphorylated cofilin is misregulated at stimulated synapses in PAK1 cells, potentially causing the deficits in F-actin at the spine.

In contrast to an ablation of PAK1 in the entire organism, Liu et al. used a conditional gene deletion system to specifically knock out PAK1 in cardiomyocytes (PAK1). These mice were also shown to be healthy and fertile, as expected, since constitutive PAK1-null mice show no gross defects in health or fertility. However, when PAK1 mice were subjected to pressure overload by transverse aortic constriction (TAC), a significant increase in heart weight/tibia length ratio was noted. Additionally, there was a significant increase in the cross-sectional area of cardiomyocytes in the PAK1-TAC mice compared with the control-TAC mice. It was also noted that more interstitial fibrosis developed in the PAK1-TAC myocardium. Importantly, atrial and brain natriuretic peptides (ANP and BNP), as well as fetal contractile protein isoforms such as α-myosin heavy chain 7 mRNA, was found to be significantly elevated in hypertrophied PAK1 myocardium. The increase in these genes suggests that the progression of ventricular hypertrophy to heart failure in this mouse model represents a re-activation of the fetal heart program in cardiomyocytes. These results also suggest that PAK1 normally acts as an anti-hypertrophic agent. Indeed, when treated with FTY720, a sphingosine-like analog that activates PAK1, wild-type mice resisted developing pressure overload-induced hypertrophy. In addition to these studies, Liu et al. assessed the effects of prolonged TAC stress to determine if PAK1 mice became more sensitized to heart failure. With extensive TAC treatment, they found that these mice displayed higher lung weight/tibia length ratio suggestive of pulmonary edema, a significant reduction in fractional shortening, indicating decreased contraction, and further enlargement of the heart compared with the control group. These data show that extensive load stress renders PAK1 knockouts more susceptible to heart failure as compared with controls. Similar data were obtained using a separate heart failure model, in which hypertrophy is induced by neuroendocrine stimuli. In terms of signaling, Liu et al. found that pressure overload induces an increase in MKK7, MKK4 and JNK in wild-type animals, but that this pathway is attenuated in cardiomyocyes from PAK1-null mice. These data implicate the JNK cascade, as opposed to the ERK cascade, as the major MAPK pathway linking PAK1 to ventricular hypertrophy.

PAK2

Though much is known about PAK2 biochemistry and its role in cellular signaling pathways, much less is known about its function in the developing embryo. PAK2-null mice have been produced, but contrary to the other group I PAK members, loss of the PAK2 gene leads to embryonic lethality at ~E8.0., Embryogenesis fails due to multiple developmental abnormalities, most prominently involving defective vascularization. PAK2 mice have been produced by our laboratory and are currently under study.

PAK3

In man, lesions in PAK3, a gene whose expression is predominantly restricted to the brain, are associated with X-linked, nonsyndromic mental retardation (MRX)., To date, five separate MRX-30 kindreds have been reported, each with distinct missense or nonsense mutations in PAK3 that are predicted to result in loss of function in the PAK3 protein. Interestingly, one of these mutations, R67C, does not affect kinase activity per se but does diminish the binding of PAK3 to its activator, Cdc42. It is also interesting to note that the various PAK mutants have distinct effects on neurons when expressed in cultured cells.

To model MRX-30, Meng et al. produced PAK3-null mice. These mice are healthy, fertile, have normal locomotor activities and a normal life span. Additionally, these mice display normal brain structure and no abnormalities in the CNS, including hippocampus and cerebellum, much like patients with nonsyndromic mental retardation. Furthermore, PAK3-null mice display normal dendrite and spine morphology, and normal basal synaptic strength, indicating that basal synaptic strength and presynaptic neurotransmitter release are not affected in these knockout mice. The group determined that while LTD and early phase LTP are normal between genotypes, late phase LTP (L-LTP) is significantly reduced in PAK3-null mice, a phenotype mimicking the PAK1-null mice. However, unlike PAK1 knockout mice, phosphorylated CREB is reduced in the PAK3 knockouts, whereas phosphorylated cofilin is unaffected. Together with the molecular analysis, the group showed that these mice display deficient memory retention through taste aversion tests. As CREB is known to be required for L-LTP and memory formation, these finding provide a potential molecular pathway linking PAK3 mutations to mental retardation. These findings also suggest that PAK1 and PAK3 are redundant in synapse function.

To test for such functional redundancy, Huang et al. crossed PAK1- and PAK3-null mice. The double-knockout mice were normal at birth but soon showed major loss of brain volume compared with that of wild-type mice, despite normal brain organization. PAK1/PAK3 double-knockout mice also had deficits in learning and memory and displayed a hyperactive behavior, reminiscent of that of humans with PAK3 mutations. The double knockout mice also displayed a much less complex neuronal morphology, with reduced dendrite length and number of dendritic tips, suggesting that PAK1 and 3 are involved in branch formation. As in PAK1-null mice, phosphorylated cofilin levels were significantly reduced in the double knockouts, suggesting a role for both PAK1 and 3 in the regulation of cofilin and thereby in neuron structure and function.

Group II PAKs

PAK4

PAK4, as with PAK2, is expressed ubiquitously and its expression begins early in development. In cells, PAK4 is thought to mediate Cdc42 induced filopodia formation, induce changes in cytoskeletal organization and cell adhesion and stimulate cell survival pathways. The PAK4 gene has been disrupted in vivo and causes embryonic lethality by day E11.5, most likely from a heart defect.PAK4-null embryos show a thinning of the myocardial walls of the bulbous cortis and ventricle, likely leading to impaired ventricular functioning and pooling of blood within the heart, causing its distended appearance. The PAK4-null mice also display other impairments, chiefly noted within the nervous system and in vessel formation.,PAK4-null embryos display a thin neuroepithelia in the hindbrain and forebrain, causing an unusual translucent head and neural tube in the intact embryo. These embryos have defects in neuronal differentiation and axonal outgrowth. While neuronal progenitors form properly, differentiation and migration of both motor neurons and ventral interneurons are defective. These findings are consistent with in vitro data showing that PAK4 is involved in the formation of filopodia, a structure known to play key role in the guidance of neuronal growth cones. Equally important, defects have been noted in the caudal portion of the neural tube of PAK4-null mice, consistent with misfolding. Such misfolding might cause the formation of two separate lumens, a process regulated by the cytoskeleton.

In addition to neuronal and heart defects, PAK4-null mice also show defects in vascularization. In a developing embryo, fetal blood vessels must integrate with maternal blood vessels in the placenta for proper nutrient delivery, specifically the labyrinth layer. PAK4-null mice display defects within this layer including the lack of proper vessel branching. Additionally, the yolk sac of the developing PAK4-null embryo shows poorly developed vessel branching. Since vessel formation is initiated in both the placenta and yolk sac, PAK4 most likely is involved with the formation or extension of branching vessels and therefore angiogenesis. Both angiogenesis and the heart defect could be explained by a disruption in endothelial cell migration, a process known to be regulated by PAK4. However, further evidence is needed to support this hypothesis. Correspondingly, it is likely that these defects contribute to the embryonic lethality caused by the genetic disruption of PAK4.

Due to embryonic lethality of PAK4-null mice and further interest in neuron development, conditional knockout PAK4 mice have been developed, in which PAK4 is deleted in the nervous system. Neuronal defects were seen in PAK4 as early as day E16.5, with normal neuron development beginning around day E11; however, striking differences in brain anatomy were observed at birth. Gross morphology of the brain was normal, specifically all layers of the cortex were present, but significantly thinner than that of the controls, with the outer most layers being the thinnest. This thinning was attributed to a decrease in proliferation of neural progenitor cells. In addition, these mice displayed severe hydrocephalous, possibly contributing to their early demise between two and four weeks after birth. Furthermore, adherens junctions in the neuroepithelial were lost at the ventricular interface at day E18.5. Since these junctions have been associated with cell proliferation, it is possible that their absence causes the observed decrease in cell proliferation. To further investigate the role of PAK4 in the developing brain, PAK4 substrates were examined for activity. No significant changes were detected in the phosphorylation or total expression levels of ERK, A-Raf, B-Raf, Raf-1, Stat3, c-Myc or p120-catenin. Interestingly, however, neuroepithelial adherens junctions were found to lack β-catenin, a known substrate of PAK4., As disruption in β-catenin signaling has been shown to cause reduced levels of cell proliferation as well as hydrocephalous and post-natal death,, these data support a significant role for PAK4 and β-catenin in the developing mammalian brain.

PAK5 and PAK6

Interestingly, PAK5-null mice are viable and healthy and through in depth examinations appear to be comparable to that of the wild-type. The morphology of the brain, hippocampus, eyes, pancreas and other tissue such as the testes, prostate, epididymis and adrenal gland all appear to be normal. Similarly, PAK6-null mice are viable, healthy and fertile. The lack of abnormal phenotypes for these single knockouts could be due to redundancy between PAK5 and 6, due to their overlapping expression patterns. To further examine this possibility, PAK5 and 6 were knocked out in tandem. The combined knockout mice were healthy and viable without any obvious signs of defects. Unlike PAK4, PAK5 and 6 have a restricted pattern of expression and are expressed later in development, perhaps explaining the lack of overt phenotype. The double knockouts did display deficits in learning as well as locomotor activity, consistent with the fact that PAK5 and 6 are expressed in brain structures that are strongly involved in cognitive function. Further analysis in these mice showed normal motor neuron progenitor cells; however, they were reduced in numbers compared with that of wild-type mice. The authors also noted that motor neurons in these mice do not migrate to their proper location. In vitro data of cultured motor neurons suggest a lack in proper growth cone size and structure, possibly linking PAK5 and 6 to cytoskeletal changes in the brain, affecting neuronal plasticity.

Summary

The use of genetic engineering, particularly constitutive and conditional gene knockouts in mice, has been essential in determining the specific functions of each PAK isoform in mammals. In the near future, such models should be useful in determining the role of PAK isoforms in various mouse tumor models as well as other disease models. For example, PAK1 and PAK4 have been implicated in a number of human malignancies; breeding PAK-deficient mice to the appropriate mouse cancer models could provide useful preclinical evidence for future therapeutic interventions. Also, knock-in models could help further define PAK function. For example, knock-in of “analog-sensitive” kinase alleles could be used to test the effects of small-molecule mediated PAK inhibition in vivo., As genetic engineering methods are become increasingly facile, it is not unreasonable to expect that we will have a sophisticated understanding of the function of all six PAKs in the near future.