Regulation and Function of Cyclin D2 in B Lymphocyte Subsets¹

The cell cycle is divided into an interphase, in which cell growth and DNA replication occurs, and a mitotic phase, in which cell division occurs. The point in the G₁ phase that represents an irreversible commitment to replicate the genome and undergo cell division has been termed the “restriction point” (1). Growth factors function by signaling passage through the restriction point, at which time continued cell cycle progression proceeds independent of extracellular cues (Fig. 1). Proper control of restriction point progression is essential for maintaining normal levels of cell growth and proliferation (2). Accumulating evidence indicates that passage through the restriction point is regulated by the retinoblastoma tumor suppressor gene product pRb and by the related proteins p107 and p130 (3). pRb functions to repress transcription of genes whose products are required for transition through the restriction point and S-phase progression (3). This repression is achieved through the binding of numerous proteins, notably members of the E2F/DP-1 family of transcription factors, and by recruiting histone deacetylases and chromatin remodeling SWI/SNF complexes to E2F-responsive gene promoters (reviewed in Ref.4). pRb is inactivated by sequential phosphorylation initiated early in the G₁ phase by a subset of cyclin-dependent kinases (cdk4 and cdk6),³ followed by cdk2 in the late G₁ phase (Fig. 1; Refs.4 and 5). The proper timing and duration of cdk4/6 activation are controlled by: 1) the binding of regulatory proteins, including activators (i.e., D-type cyclins, composed of D1, D2, and D3) and inhibitors (composed of the INK4 and Cip/Kip families); and 2) phosphorylation of the cyclin and cdk4/6 subunits (6, 7, 8, 9, 10, 11, 12, 13, 14).

Much functional data support the view that D-type cyclins play a role in G₁-S progression. Notably, forced expression of cyclin D1 can shorten the G₁ phase period, and the overexpression of cyclin D1-cdk4 can suppress a pRb-induced G₁ block (15, 16). Even more compelling are experiments demonstrating that cyclin D1 is dispensable in vertebrate cells lacking pRb (17). However, this view of the function of cyclin D-cdk4 has been challenged by reports that cyclin D1, cyclin D2, and cdk4 knockout mice are viable (18, 19, 20, 21). Notably, cdk4-deficient mice are smaller and have smaller cells where examined, indicating a role for cdk4 in cellular growth (18). An important and unresolved question concerns whether D-type cyclins possess distinct biological functions. The high degree of amino acid homology (50–60% identity throughout the coding region) between the mammalian D-type cyclins suggests functional redundancy. Indeed, gene targeting of individual D-type cyclins in mice reveals only limited tissue-specific abnormalities, particularly in the retina and breast for cyclin D1-deficient mice and in the testes and ovaries for cyclin D2-deficient mice (19, 20, 21). Like cdk4-null mice, cyclin D1-deficient mice exhibit dwarfism-like phenotype, further implicating cyclin D-cdk4 in the regulation of animal growth. Mice expressing a single D-type cyclin manifest narrowly restricted abnormalities, suggesting that the functions of the three D-type cyclins during development, proliferation, and differentiation of the majority of tissues are interchangeable (22).

Early investigations of mouse splenic B cells implicated cyclins D2- and D3-, but not cyclin D1-containing cdk4 complexes as crucial links between mitogenic stimuli (e.g., anti-Ig, LPS, or CD40 agonists) and pRb phosphorylation (23, 24, 25, 26, 27). In general, both cyclins D2 and D3 are capable of forming active pRb kinase complexes comprised of cdk4 and, to a lesser extent, cdk6 (23). Many of these studies also document for the first time changes in Ink4 and Cip/Kip protein levels following Ag receptor cross-linking; however, the biological significance of these observations to the humoral immune response remains to be investigated. The notable exception is p18^INK4c, which has been shown to be essential for cell cycle arrest and differentiation of nonsecreting plasmacytoid cells to Ab-secreting plasma cells (reviewed in Ref.28).

The idea that cyclin D2 is necessary for BCR-induced G₁-to-S phase progression was initially formulated on the basis of correlative data gathered from studies comparing the induction of cyclin D2 following the stimulation of B cells with partial vs complete mitogens. These studies demonstrated an absence of cyclin D2 induction in B cells stimulated with partial mitogens that act to promote G₁-phase, but not S-phase entry, whereas CD40 or BCR agonists induced sustained expression of cyclin D2 (23, 24, 25). In keeping with these earlier studies, cyclin D2 induction has been recently observed in transitional-2 B cells (IgM^highIgD^highCD21^highCD24^highCD23^high) stimulated to proliferate in response to BCR cross-linking, but not induced in transitional-1 B cells (IgM^highIgD^lowCD21^lowCD24^highCD23^low), which fail to proliferate following anti-Ig stimulation (29, 30). An important caveat to these studies concerns the proliferative response of transitional B cells. Although there is good agreement that T1 cells do not proliferate following BCR ligation, disagreement exists over whether T2 cells actually proliferate (31, 32, 33). That said, a cautionary note is warranted insofar as these results collectively might be interpreted to mean that cyclin D2 induction alone is sufficient to promote S-phase entry in B cells. Important in this regard is the observation that BCR cross-linking of primary immature B lymphocytes leads to G₁-phase entry, but not proliferation, despite induction of cyclin D2 that is comparable to that of anti-Ig-stimulated mature B cells (34); significantly, cyclin E is not induced by anti-Ig. It is probable that, in immature B cells, cyclin D2-cdk4 initiates pRb phosphorylation, yet in the absence of cyclin E, subsequent phosphorylation and inactivation of pRb by cdk2 are not achieved, thereby accounting for impaired S-phase entry.

The first definitive evidence that induction of cyclin D2 is necessary for BCR-mediated B cell clonal expansion was gleaned from mice homologous for a genetic ablation of cyclin D2 (35). Notably, cyclin D2-deficient splenic B cells exhibit defective proliferation in response to anti-Ig, but not to CD40- or LPS-induced proliferation (35). Along these lines, gene expression profiling of BCR/ABL-transformed BaF3 cells has revealed dysregulated cyclin D2 expression, occurring throughout the cell cycle (36). BCR/ABL is a constitutively active protein tyrosine kinase responsible for chronic myelogenous leukemia (36). Retroviral transduction of the BCR/ABL gene is sufficient to induce proliferation of infected bone marrow cells from wild-type, but not cyclin D2-deficient mice.

Mice with cyclin D2 deficiency do not appear to exhibit hemopoietic abnormalities (20). Initial evaluation of the lymphoid, myeloid, and erythroid progenitor lineages in the bone marrow do not reveal significant differences in the numbers of B220⁺IgM⁻ pre-B cells between cyclin D2-deficient and wild-type mice (35). However, a recent study suggests that cyclin D2-deficient mice exhibit half the number of Sca-1⁺B220⁺ B cell progenitors in comparison to wild-type mice (37). In vitro clonal analysis of bone marrow reveals a preponderance of B220⁺CD19^low pre-B cell colonies derived from cyclin D2-deficient mice, whereas wild-type bone marrow yields a higher proportion of B220⁺, CD19^high pre-B cell colonies. Whether these data point to impaired proliferation or perhaps decreased survival of an early pro-B220⁺CD19⁻ B cell population in cyclin D2-deficient mice remains to be established (Fig. 2). In contrast, analysis of bone marrow and splenic B220⁺IgM⁺ B cell populations reveals near-normal levels in cyclin D2-deficient mice compared with wild type (35). The apparent normal development of IgM⁺ B cells in the absence of cyclin D2 may be an indication of an alternate pathway for the development of IgM⁺ B cells that is dependent upon cyclin D3. The likelihood of this scenario is supported by gene expression profiling of mouse bone marrow B cell subsets, which reveals cyclin D3 expression during early B cell development (38).

In an earlier study, Tanguay et al. (39) demonstrated that peritoneal CD5⁺B-1a cells express a low level of cyclin D2, which is up-regulated in a rapid and transient manner in response to mitogenic stimulation. In contrast, cyclin D3 is expressed near the G₁/S transition, and its expression occurs in the absence of detectable cyclin D2 (40). The nonoverlapping expression of cyclins D2 and D3 suggests that these D-type cyclins may provide distinct functions in B-1a cells, with the latter contributing to restriction point control. Interestingly, assessment of B-lymphoid populations in the peritoneal cavity from cyclin D2^−/− mice reveals diminished numbers of peritoneal CD5⁺B-1a cells, than in wild-type mice, whereas the numbers of conventional IgM^lowIgD^high B cells in the peritoneal compartment are not significantly altered (Fig. 2) (35). The loss of peritoneal CD5⁺B-1a cells in cyclin D2-deficient mice points to an essential nonredundant function of cyclin D2 in this B cell subset.

Gene expression profiling of distinct stages of human and mouse B cell development has confirmed earlier studies of D-type cyclin expression in mature B cell populations, while at the same time arousing speculation that cyclin D2 exhibits new activities unrelated to its function as a cdk4/6 activator and probably independent of its function as a regulator of pRb. Using microarrays to characterize gene expression patterns, Alizadeh et al. (41) analyzed a variety of adult lymphoid malignancies together with normal human lymphocyte subpopulations under a range of activation conditions. This study reveals thatcyclin D2 is expressed at a constitutively high level in diffuse large B cell lymphomas (DLBCLs) (see below). Cyclin D2 is induced in activated human peripheral blood B cells and is not detectable in human blood naive CD27⁻ or memory CD27⁺ human peripheral blood B cells (42). Yet a subsequent report by Klein et al. (43) shows that cyclin D2 is expressed in both human tonsillar naive and memory B cells. At present, it is unknown whether the differences observed in these studies are due to potential artifacts introduced during the B cell isolation or, alternatively, reflect subtle differences in the activation status of the B cell subsets.

One of the most striking observations uncovered by these studies is the apparent absence of cyclin D2 gene expression in germinal center (GC) centroblasts and centrocytes, despite the highly proliferative nature of the former B cell subset (42, 43). How these cells proliferate in the absence of cyclin D2 is not known. One possibility is that cyclin D3 may compensate for the loss of cyclin D2 expression and restore proliferation. In support of this, cyclin D3 gene expression and not cyclin D1, has been identified by expression profiling in human GC B cells (41, 43). Another possibility relates to the finding that several genes known to participate in the G₁/S and G₂/M phases of the cell cycle (e.g., cyclins E1/E2, GADD45, DP-1, cyclin B1-cdc2, pLK, CENP-E/F, Mad2, and survivin) are highly expressed in GC B cells in comparison with other human B cell populations (41, 42, 43). In mouse GC B cells, the c-myc gene, which encodes a transcription factor that regulates G₁-to-S phase progression, is expressed at a high level (44). Thus, the expression of these genes may account for the highly proliferative nature of GC B cells in the absence of cyclin D2. Of note, in many instances it has not been established whether the protein levels of individual D-type cyclins correlate with mRNA levels measured by gene expression profiling. This is an important consideration given an earlier study of discordant protein and mRNA levels for Bcl-2 in GC and mantle zone B cells (45).

Perhaps more intriguing is the matter of why cyclin D2 is repressed in such a highly proliferative cellular compartment. One possibility is that cyclin D2 may have additional functions, independent of its role in G₁-to-S progression. Recent studies have established cyclin D as an important regulator of cell growth. For example, homozygous inactivation of the cyclin D1 gene produces a small mouse phenotype and manipulating the cyclin D pathway in Drosophila results in altered organism growth (19, 21, 46, 47). Important in this regard, Shaffer et al. (42) have noted that the gene expression signature of GC B cells favors cell proliferation at the expense of cellular growth. This observation raises the possibility that cyclin D2 might be important in the intracellular pathway leading to B cell growth. If so, GC cells may opt to repress cyclin D2 expression and, thus, to forego cellular growth in favor of cell proliferation.

Several studies have implicated D-type cyclins in modulating differentiation programs (47). Cyclins D1 and D2 were shown to inhibit myogenic differentiation through inactivation of the muscle transcription factor MyoD (47, 48, 49). In addition, cyclins D2 and D3, but not D1 inhibit the ability of 32D myeloid cells to undergo neutrophil differentiation in response to G-CSF (50). By extension, a role for cyclin D2 may be envisaged in modulating B cell fate within the GC. Naive splenic B cells in the GC environment respond to Ag either by becoming activated and differentiating into plasmacytic cells in the periarteriolar lymphoid sheath or, alternatively, by differentiating into GC B cells in the follicular region. BCL-6, a transcriptional repressor expressed in GC B cells, appears to function by skewing B cells away from plasmacytic differentiation and toward GC differentiation (42, 44). Microarray analysis of B cell lines engineered to express BCL-6 reveals a subset of genes that are specifically repressed by BCL-6, including cyclin D2 (42, 44). Should cyclin D2 function to inhibit GC B cell differentiation program, its repression by BCL-6 might then be a requisite step in GC B cell differentiation.

The first evidence supporting the concept that the signalosome is involved in regulating cyclin D2 expression was gleaned from the analysis of splenic B cells from Bruton’s tyrosine kinase (Btk)-deficient and xid mice (26, 51, 52). The latter immunodeficiency results from a naturally occurring point mutation in the pleckstrin homology domain of Btk (reviewed in Ref.52). In particular, xid splenic B cells fail to proliferate in response to thymus-independent type-II Ags due to impaired de novo cyclin D2 mRNA induction. Subsequently, it was shown that adaptor protein B cell linker (BLNK)- and Vav-deficient splenic B cells do not proliferate in response to BCR cross-linking due to defective cyclin D2 induction (53, 54). The reported loss of cyclin D2 induction in Vav-deficient B cells results, at least in part, from a failure of these cells to sustain normal calcium flux following BCR ligation (Fig. 3).

More recently, DeFranco and coworkers (55) demonstrated that BCR-dependent proliferation of mature B cells requires MEK1/2-ERK activation. These studies also ruled out a requirement for MEK1/2-ERK signaling in BCR-induced apoptosis of WEHI-231 cells and immature splenic B cells derived from autoreconstituted mice. However, there does exist disparity regarding the role of ERK in BCR-induced apoptosis of WEHI-231 cells. Lee and Koretzky (56) reported that overexpression of MAPK phosphatase-1, a phosphatase that can dephosphorylate and inactivate ERK, blocked apoptosis of WEHI-231 cells. However, it is possible that MAPK phosphatase-1, in addition to inhibiting ERK, may inactivate JNK and/or p38^MAPK in WEHI-231 cells; both JNK and p38^MAPK have been shown to play a role in BCR-induced apoptosis in human B cell lines (57, 58).

Insights into the mechanism by which MEK1/2-ERK promotes BCR-mediated proliferation in mature B cells has been provided recently by Piatelli et al. (59) with the observation that inhibition of MEK1/2-ERK blocks de novo cyclin D2 gene expression and phosphorylation of endogenous pRb. The upstream signals that regulate ERK activation include a contribution from both phospholipase C (PLC)γ2 and protein kinase C (PKC), with the latter activating Raf-1 (60, 61). Recent studies in splenic B cells and DT40 B cells point to a role for the B lymphocyte adaptor molecule of 32 kDa (Bam32) in regulating ERK activation in response to BCR cross-linking (62, 63). Presumably Bam32, Sos-Ras, and PLCγ2-PKC function as upstream regulators of ERK activity, however, the degree of cross-talk between these pathways has not been established. Bam32-deficient B cells are characterized by signaling defects in the ERK upstream regulators, hemopoietic progenitor kinase (HPK1) and MEK kinase 1 (MEKK1) (62, 63, 64). Bam32-deficient B cells also exhibit defective BCR-induced proliferation (62, 65); however, it has not been reported whether induction of cyclin D2 is impaired in these B cells. These findings suggest a model wherein the BCR signals cyclin D2 induction via a Bam32-HPK1-MEKK1-MEK1/2-ERK signaling conduit (Fig. 3).

The observation that BLNK-deficient and xid B cells exhibit defective NF-κB signaling in addition to defective expression of cyclin D2 suggests that NF-κB activity may be required for cyclin D2 induction following BCR ligation (53, 66, 67). Definitive support for the regulation of cyclin D2 by NF-κB was obtained from studies probing the molecular mechanisms underlying the dependency of BCR-mediated proliferation on PI3K activity (68, 69). The PI3Ks are divided into four classes (I_A, I_B, II, and III) (70). The class I_A enzymes are heterodimers composed of a regulatory subunit (comprised of five different subunits: p85α, p55α, p50α, p85β, and p55γ) and a catalytic subunit (comprised of three different subunits: p110α, -β, and -δ). The p85α, p55α, and p50α subunits are expressed by alternative splicing of the same gene; deletion of the first exon results in a loss of p85α expression, but not of p55α and p50α (71). These mice exhibit a partial block in B cell development at the pro- to pre-B cell transition, reduced numbers of mature splenic B cells, and impaired T cell-independent Ab production. Reconstitution of Rag2^−/− embryos with p85α^−/−p55α^−/−p50α^−/− embryonic stem cells results in mice that share many of the phenotypes found in p85α^−/− mice (72). Our laboratory and other investigators have recently demonstrated that p85α-deficient splenic B cells do not proliferate in response to BCR ligation due to a failure to phosphorylate pRb (73, 74). A key determinant contributing to defective pRb phosphorylation is the complete absence of cyclin D2 induction (73, 74). Anti-Ig-stimulated p85α-deficient splenic B cells display normal ERK, yet exhibit impaired activation of the IκB kinase (IKK)α,β,γ-IκBα-NF-κB pathway (73). Inhibition of PI3K by wortmannin or LY294002 in normal splenic B cells recapitulates the NF-κB and cyclin D2-cdk4/pRb signaling defects observed in p85α-deficient B cells. Moreover, selective inhibition of individual components of the NF-κB pathway in normal B cells abolishes anti-Ig-stimulated cyclin D2-cdk4/pRb pathway activation. These data together define a PI3K/NF-κB-dependent pathway that is essential for BCR-mediated cyclin D2 induction (Fig. 3).

Although it is well established that PI3K regulates NF-κB activation, the nature of the proximal signals that link PI3K to NF-κB activity is not fully understood. An important consideration is the downstream effector of PI3K, Akt, which has been reported to interact directly with and phosphorylate IKKα (75). However, the PKC inhibitor Gö6983 blocks BCR-induced IκBα degradation without affecting Akt phosphorylation, suggesting that PKC may represent a key upstream regulator of IKKα,β,γ in splenic B cells (69). PI3K might then initiate NF-κB activation via the downstream effector serine/threonine kinase, 3′-phosphoinositide-dependent kinase 1, which has been shown to phosphorylate and activate conventional PKCs and PKC-ζ and -δ (76, 77). The idea that a conventional or novel PKC may be involved is supported by the finding that Gö6850, an inhibitor of conventional/novel PKCs, blocks cyclin D2 induction following BCR ligation (74). Whether the Gö6850-sensitive PKC functions in the pathway leading to Raf-1/MEK1/2-ERK activation, IKKα,β,γ-IκBα-NF-κB activation, or both has not been established (Fig. 3). PKCδ is probably not involved given recent reports that 1) BCR-induced NF-κB activation is normal in PKCδ-deficient splenic B cells; and 2) treatment of normal B cells with the PKCδ inhibitor, rottlerin, does not inhibit BCR-induced NF-κB activation (78, 79). An attractive candidate is PKCβ based on the finding that PKCβ-deficient mice do not activate NF-κB in response to BCR cross-linking (80). Recent studies indicate that the caspase recruitment domain protein Bcl10 may function to couple PKCβ to NF-κB activation, apparently via a mechanism that involves phosphorylation of Bcl10 by the adaptor protein caspase recruitment domain-MAGUK1 (Carma-1) (reviewed in Ref.64). However, BCR-mediated cyclin D2 induction is normal in PKCβ-deficient splenic B cells despite loss of NF-κB activity (81). These findings suggest a requirement for an NF-κB-independent pathway in cyclin D2 induction. It would seem likely that components of the ERK pathway may be sufficient to signal cyclin D2 induction in the absence of NF-κB activation and do so via activity of a PKC(s) other than PKCβ. Yet, signaling through MEK1/2-ERK is intact in BLNK- and p85α-deficient splenic B cells, which exhibit neither NF-κB activation nor cyclin D2 induction (54, 73). These results raise the possibility that a compensatory MEK1/2-ERK- and NF-κB-independent pathway may be up-regulated in PKCβ-deficient B cells to facilitate cyclin D2 induction.

Left unanswered by these studies is the identity of the trans-acting factors that couple NF-κB or MEK1/2-ERK activities to cyclin D2 gene promoter activation in B cells. In regard to the former, previous studies have indicated that Rel is necessary for c-myc transcription in response to BCR cross-linking (82). The mouse c-myc gene promoter is a direct target of Rel/NF-κB proteins and these trans-acting factors bind two sequence elements (URE and IRE) positioned upstream of the P1 promoter and in exon 1, respectively (82). In keeping with our studies in p85α-deficient B cells, Grumont et al. (83) have shown that PI3K inhibition blocks IκBα phosphorylation and degradation, nuclear translocation of NF-κB1/c-Rel, and c-myc expression.

Several lines of evidence now support the notion that cyclin D2 is a direct target gene of Myc (84, 85). For example, in primary human fibroblasts, conditional Myc expression results in induction of cyclin D2 transcription (84). The cyclin D2 promoter contains two canonical E-boxes, of which only the distal E-box appears to be a target of Myc (86). Bouchard et al. (86) have shown that the cyclin D2 gene promoter is positively regulated by the recruitment of Myc/Max heterodimers to the distal E-box. In contrast, repression of cyclin D2 gene transcription is mediated by Mad/Max complexes binding to the distal E-box. An analogous role for Myc in cyclin D2 transcriptional regulation via the BCR has not been established; however, a recent study in T cells suggests that Myc is not involved in IL-2R-mediated cyclin D2 induction, in part, because the E-boxes appear to preferentially bind upstream stimulatory factor-1 and upstream stimulatory factor-2 (87). Moreover, both E-boxes are dispensable for cyclin D2 gene promoter activation (87). Cyclin D2 transcription appears to be mediated by Stat5, which binds to an enhancer element located at −1204 in the cyclin D2 gene promoter (87, 88). Interestingly, IL-2R-induced cyclin D2 transcription is dependent on PI3K activity in T cells, similar to the situation in B cells stimulated through the BCR.

Understanding the signal transduction pathways leading to B cell proliferation has important implications for designing molecularly targeted therapies for the treatment of B cell diseases. DLBCL represents the most common type of non-Hodgkins lymphoma, comprising 30–40% of adult non-Hodgkin lymphomas (89). From a clinical perspective, only 40% of these patients are cured by chemotherapeutic regimens. Gene expression profiling of DLBCL tumors has made clear the existence of two DLBCL subgroups that are characterized by distinct gene expression signatures (41). Notably, GC B-like (GCB) DLBCLs are characterized by a gene expression signature that resembles normal tonsillar GC B cells, whereas activated B cell-like (ABC) DLBCLs express a gene signature mirroring that of human peripheral blood B cells following BCR cross-linking (41). GCB DLBCLs correlate with a more favorable prognosis compared with ABC DLBCLs (41). DNA microarrays reveal that ABC DLBCLs express genes known to be NF-κB targets, whereas GCB DLBCLs generally have low expression of these genes (41, 42). The NF-κB target genes include among others, IκBα and cyclin D2. ABC DLBCL cell lines exhibit constitutive IKK, IκBα degradation, and nuclear NF-κB activity, whereas GCB DLBCL cell lines are devoid of constitutive NF-κB activity (90). Ectopic expression of a superrepressor form of IκBα or a dominant-interfering IKKβ results in cell death to ABC DLBCL cell lines. The activity of NF-κB is also required for cell cycle progression insofar as inhibition of NF-κB signaling results in G₁-phase arrest in ABC DLBCLs. Interestingly, the human cyclin D2 gene promoter contains two NF-κB binding sites (91), suggesting that G₁-phase arrest in response to inhibition of NF-κB in ABC DLBCLs is due to a loss of sustained cyclin D2 gene expression. Because cell cycle arrest causes apoptosis in many tumor cell types, the loss of NF-κB-targeted cyclin D2 induction may contribute to the cytotoxic effects in ABC DLBCLs. These findings suggest that molecularly targeted therapies that inhibit components of the NF-κB/cyclin D2 pathway may prove useful in treating clinically intractable DLBCLs.

Recent work has suggested some new ways to think about the function of cyclin D2 in B cells. That cyclin D2 may have additional roles in B cells, beyond regulating cdk4/6 and pRb, represents an exciting area of investigation that may provide novel insights into the regulation of B lymphocyte growth and differentiation. We are also beginning to define the signal transduction networks that serve to link the BCR to cyclin D2 induction in mature B cells. On this point, MEK1/2-ERK and NF-κB pathways have emerged as major signaling conduits. Finally, understanding the regulation of cyclin D2 has important implications for the treatment of lymphoid malignancies, such as DLBCL.

I apologize to those investigators whose work was not cited due to space limitations. I thank Dr. Jennifer Mataraza (Department of Biology, Boston College, Chestnut Hill, MA) and Craig Kasprzak (Boston College) for a critical reading of the manuscript. I also thank Dr. Edward A. Clark (Departments of Microbiology and Immunology, University of Washington, Seattle, WA) for providing helpful comments.