Type I Interferon Drives Dendritic Cell Apoptosis via Multiple BH3-Only Proteins following Activation by PolyIC In Vivo

Background DC are activated by pathogen-associated molecular patterns (PAMPs), and this is pivotal for the induction of adaptive immune responses. Thereafter, the clearance of activated DC is crucial to prevent immune pathology. While PAMPs are of major interest for vaccine science due to their adjuvant potential, it is unclear whether and how PAMPs may affect DC viability. We aimed to elucidate the possible apoptotic mechanisms that control activated DC lifespan in response to PAMPs, particularly in vivo. Methodology/Principal Findings We report that polyinosinic:polycytidylic acid (PolyIC, synthetic analogue of dsRNA) induces dramatic apoptosis of mouse splenic conventional DC (cDC) in vivo, predominantly affecting the CD8α subset, as shown by flow cytometry-based analysis of splenic DC subsets. Importantly, while Bim deficiency conferred only minor protection, cDC depletion was prevented in mice lacking Bim plus one of three other BH3-only proteins, either Puma, Noxa or Bid. Furthermore, we show that Type I Interferon (IFN) is necessary and sufficient for DC death both in vitro and in vivo, and that TLR3 and MAVS co-operate in IFNß production in vivo to induce DC death in response to PolyIC. Conclusions/Significance These results demonstrate for the first time in vivo that apoptosis restricts DC lifespan following activation by PolyIC, particularly affecting the CD8α cDC subset. Such DC apoptosis is mediated by the overlapping action of pro-apoptotic BH3-only proteins, including but not solely involving Bim, and is driven by Type I IFN. While Type I IFNs are important anti-viral factors, CD8α cDC are major cross-presenting cells and critical inducers of CTL. We discuss such paradoxical finding on DC death with PolyIC/Type I IFN. These results could contribute to understand immunosuppression associated with chronic infection, and to the optimization of DC-based therapies and the clinical use of PAMPs and Type I IFNs.


Introduction
Dendritic cells (DC) are the most potent professional antigenpresenting cells and play a key role in the induction of tolerance and initiation of adaptive immune responses [1]. Such decisive role of DC has attracted major attention for the improvement of vaccine efficacy, including the targeting of antigens to DC and the use of adjuvants to optimally activate DC and consequently boost immunity [2].
The DC system is composed of phenotypically and functionally heterogeneous subsets [1,3]. Migratory DC reside in non-lymphoid organs and are able to capture antigens at the sites of infection or injury and transport them to draining lymph nodes. Lymphoid tissue-resident, conventional DC (cDC) are non-migratory and serve as sentinels in secondary lymphoid organs and tissues for in situ antigen capture and presentation. In the mouse, several DC subtypes can be found in the spleen: two subsets of cDC (CD8a + CD11b 2 and CD8 2 CD11b + , hereafter termed CD8a and CD11b subsets), inflammatory monocyte-derived DC, and plasmacytoid DC (pDC).
Pattern-Recognition Receptors (PRR), including the Toll like receptor (TLR) family, detect Pathogen-Associated Molecular Patterns (PAMPs) and are important for the stimulation of antigen-presenting activity, featuring up-regulation of co-stimulatory molecules and the secretion of a variety of chemokines and cytokines [4]. Importantly, following detection of PAMPs, DC switch from tolerance induction to T cell priming. Notably, the various DC subsets differ in their profile of PRR expression as well as in their cytokine responses and antigen-presenting capacities, with a consequent division of labour amongst DC subsets in pathogen recognition and immune response initiation [4,5]. Amongst the different PAMPs under research and used in clinical studies, the synthetic dsRNA analog polyinosinic:polycytidylic acid (PolyIC) has proven superiority as a vaccine adjuvant for Th1 immunity, including in DC targeted vaccines [2], with the induction of Type I Interferon (IFN) by PolyIC being key to the enhancement of DC function [6].
The control of DC lifespan is yet another critical parameter that can profoundly influence the outcome of an immune response, potentially affecting vaccine efficacy and its side-effects [7,8]. While an increase in the lifespan of DC can elicit autoimmunity [9], reduced DC numbers correlate with immunosuppressed states in several settings both in humans and in mice, including chronic viral infection, sepsis and cancer [7]. The DC compartment must be strictly regulated, in order to sustain homeostasis, to return to steady-state levels following the induction of an immune response, to prevent autoimmunity, and to avoid immunosuppresion.
Apoptotic cell death is thought to play a critical role in the regulation of the numbers and function of DC [7,8]. Mammals have two distinct but ultimately converging apoptosis signaling pathways. The ''death receptor'' (or ''extrinsic'') pathway is triggered by members of the TNF-R family, containing intracellular ''death domains'' (e.g. Fas, TNF-R1), and requires activation of the initiator caspase, caspase-8. The ''Bcl-2regulated'' (also called ''intrinsic'' or ''mitochondrial'') apoptotic pathway can be triggered by developmental cues, cytokine withdrawal or cytotoxic stress signals (e.g. DNA damage); it involves mitochondrial outer membrane permeabilization (MOMP) and consequent release of apoptogenic factors, such as cytochrome c, which promotes activation of the initiator caspase-9, unleashing the caspase cascade. This pathway is controlled by the Bcl-2 family and these proteins can be sub-divided according to structure and function [10]. The anti-apoptotic members, including Bcl-2, Bcl-x L , Mcl-1 and A1, share up to 4 regions of homology (BH or Bcl-2 Homology domains) and are essential for cell survival. Bax and Bak have three BH regions and are critical for MOMP and activation of the effector phase of apoptosis. The BH3-only subgroup, including Bim, Bid, Puma and Noxa, is essential for initiation of apoptosis and thought to activate Bax/ Bak proteins either directly or indirectly by sequestering their prosurvival relatives [11]. Both the ''death receptor'' and the ''Bcl-2regulated'' apoptotic pathways converge upon the activation of the effector caspases, caspase-3, -6 and -7, which cleave a multitude of substrates and thereby cause cellular demolition. In addition, cross-talk between ''intrinsic'' and ''extrinsic'' apoptosis pathways may occur. For instance, caspase-8 may proteolytically activate the pro-apoptotic BH3-only member Bid that will then activate the ''intrinsic'', mitochondrial death pathway.
The apoptotic mechanisms that regulate DC function are poorly understood. T-cell dependent killing of antigen-presenting DC has been described for both CD4 and CD8 T cells, via FasL and perforin secretion, respectively, as a negative feedback mechanism to attenuate immune responses [12,13]. However, there is incomplete or contradictory evidence on the consequences of microbial stimulation for DC viability. Treatment of DC in vitro with TLR-Ls results in up-regulation of the pro-apoptotic protein Bim, but TLR-Ls rather promote survival and appear to have an inhibitory effect on the spontaneous apoptosis of DC in culture [14]. Furthermore, Bim-deficient DC are partially protected from spontaneous apoptosis in vitro, regardless of the presence or absence of TLR-Ls [14]. In contrast, treatments in vivo using TLR-Ls, such as CpG or LPS, as well as infection with E.coli or LCMV-Armstrong, induce loss of splenic cDC [15,16]. To date, there is a lack of information on the possible apoptotic pathways that control DC populations during the early phases of activation in vivo, particularly in response to PAMPs and independently of other immune cells.
Here we show that PolyIC induces apoptosis of splenic cDC following their activation in vivo, with the CD8a cDC being more susceptible than the CD11b cDC subset. Proving the apoptotic mechanism involved in vivo, absence of Bim together with a second BH3-only family member (Bid, Noxa or Puma) leads to protection from DC death. We also show that Type I IFN production is required and sufficient for PolyIC-induced DC death. This study sheds light into the mechanisms of activated DC death in vivo. The results could contribute to an understanding of immunosuppression during chronic viral infection and inflammation, and may provide important clues for the optimization of vaccines and treatments featuring PAMPs and Type I IFNs.

Splenic cDC are reduced in numbers following PolyIC treatment
Splenic cDC were analyzed at several time-points following injection of PolyIC in WT mice. In spleen, cDC are characterized as CD11c MHC-II B220 2 and are commonly divided into the CD8a + CD24 + CD11b 2 SIRPa 2 and CD8 2 CD24 2 CD11b + SIRPa + subsets with a steady-state ratio of 1:3 [3]. The other major DC subset, the pDC, expresses intermediate levels of CD11c and is B220 + . Moreover, infiltrating monocytes can differentiate into DC expressing CD11c and Gr-1 (moDC). We therefore identified splenic cDC as CD11c + B220 2 Gr-1 2 and sub-divided them into CD8a or CD11b subsets. As expected, PolyIC induced cDC activation, characterized by up-regulation of the co-stimulatory molecules CD40, CD80 and CD86, with a peak between 10-20 h (Fig. 1A). Activation was more pronounced in the CD8a subset and correlated with a marked increase in their absolute numbers at 10 h (Fig. 1B). Between 40 and 100 h, however, the CD8a subset was dramatically reduced in numbers, with as little as 10% remaining when compared to starting numbers ( Fig. 1B and C). In comparison, the CD11b subset showed a less pronounced activation as well as less expansion and subsequent reduction. Within 130 h of PolyIC treatment, both cDC subsets returned to steady-state levels. These results demonstrate that, in response to PolyIC injection, splenic cDC undergo transient activation followed by a marked reduction in the CD8a cDC subset in vivo.

Splenic cDC depletion upon PolyIC stimulation is independent of both T cells and NK cells
Previous reports have described the killing of DC by both CD4 and CD8 T cells, via Fas ligand and perforin, respectively [12,13]. Splenic cDC depletion was therefore investigated in the Fas LPR spontaneous mutant (Fas-deficient) mice and perforin (Pfn)deficient mice at 40 h of PolyIC treatment. In neither case was the loss of splenic cDC prevented ( Fig. 2A), showing that the previously described CD4 and CD8 T cell controls of DC populations by these factors are not involved in PolyIC-induced splenic cDC depletion. To further evaluate the contribution of NK cells and/or T cells in this loss of splenic cDC, we injected PolyIC into mice lacking both cell types due to deficiency in Rag-2 and the cytokine receptor common c-chain (c c ) (Rag-2 2/2 c c 2/2 ) ( Fig. 2B and Fig. S1A). Treatment of Rag-2 2/2 c c 2/2 mice with PolyIC induced a marked reduction in the numbers of splenic cDC (Fig. S1B). The fold-change in total cells with PolyIC showed that CD8a cDC were more extensively depleted, while CD11b cDC were less reduced, in Rag2 2/2 c c 2/2 mice compared to WT mice (Fig. 2B). Interestingly, Rag-2 2/2 c c 2/2 mice showed a steady-state CD8a-to-CD11b cDC ratio close to 0.8, in contrast to WT mice which display a ratio of ,0.3. Such altered ratio is seen in all mice lacking B cells (K.S. and D.V., unpublished observations) and may have influenced the pattern of depletion upon PolyIC treatment in Rag-2 2/2 c c 2/2 . Regardless, these results demonstrate that a T cell-and NK cellindependent mechanism must reduce spleen cDC numbers and thereby limit DC responses to PolyIC stimulation.

Depletion of splenic cDC also occurs upon treatment with different TLR-Ls and pathogenic infection
It has been shown that both CpG (TLR9-L) and LPS (TLR4-L) induce loss of splenic DC, albeit with limited distinction and quantitation of DC subsets [15]. We therefore examined whether preferential depletion of CD8a cDC was unique to PolyIC, or whether it could also be observed with other PAMPs. At 12 h, both CpG and LPS induced activation of the two cDC subsets (Fig. S2A), as well as a reduction in their cell numbers at 40 h (Fig. S2B). Interestingly, as in the case of PolyIC, LPS and CpG induced a greater activation and cDC loss in CD8a cDCs compared to CD11b cDCs. While CpG induced more modest effects, LPS treatment caused an extent of activation and subsequent depletion of CD8a cDCs comparable to PolyIC.
PolyIC is a synthetic dsRNA analog that mimics genome or replication intermediates often found in viruses. In order to examine whether cDC depletion also occurs during infection with live pathogens in vivo, we analyzed the splenic cDC composition in mice at early time-points following challenge with LCMV or Leishmania parasites. Infection with 209000 pfu of LCMV-Armstrong has previously been shown to reduce the numbers of both splenic cDC subsets by ,50% at day 3 [16]. In our study, we inoculated mice i.v. with 109000 pfu of LCMV-WE, a more aggressive and persistent strain of LCMV [17]. Early activation (CD86) on day 2 was accompanied by partial loss of cDC, and even more profound loss was observed on day 3 (Fig. 3A). Again, the CD8a cDC subset was more severely affected than the CD11b subset. Compared to the previously reported data on LCMV-Armstrong infection [16], the more aggressive LCMV-WE strain induced a bigger loss of cDC.
In addition, we challenged mice with the two Leishmania species, L. major and L. guyanensis, causing cutaneous and mucocutaneous leishmaniasis, respectively. Leishmania parasites stimulate surface as well as endosomal TLRs. Interestingly, L. guyanensis harbors a dsRNA virus that can provide additional TLR-3 stimulation [18], similar to PolyIC. Systemic infection of mice with L. major or L. guyanensis resulted in activation (CD86) of both cDC subsets on days 2 and 3 ( Fig. 3B). Significant loss of splenic cDC was only observed with L. guyanensis but not L. major, presumably due to the stimulation by leishmaniavirus carried in L. guyanensis (additio-nalTLR3-L component).
Altogether, these data show that loss of splenic cDC was not restricted to stimulation with synthetic PolyIC but was also found in more patho-physiologically relevant infectious settings.
PolyIC modulates expression of pro-and anti-apoptotic Bcl-2 family genes in splenic cDCs in vivo Given the dramatic loss of splenic cDC upon PolyIC treatment, independently of other immune cells, we wanted to distinguish between emigration and cell death. For this purpose, we first analyzed pro-and anti-apoptotic Bcl-2 family members in cDC subsets purified 14 h after treatment in vivo with PolyIC, a time shortly before their depletion (Fig. S3A). Due to the low numbers of DCs that can be obtained and the clearance of apoptotic cells in vivo, we examined modulation of Bcl-2 family members at the mRNA level by quantitative real time (qRT)-PCR, a potential early apoptotic event (Fig. S3B). Amongst the pro-apoptotic BH3only members analyzed, a strong induction of Bim and, albeit to a lesser extent, Puma could be consistently detected in both cDC subsets. Of note, CD8a cDCs expressed a higher steady-state level of Bim (three-fold) and Puma (two-fold) compared to the CD11b subset (PBS controls). While Noxa displayed inter-experimental variability in CD8a cDCs, being on average maintained, it was also up-regulated in CD11b cDCs. Interestingly, Bid expression was clearly down-regulated with PolyIC in CD8a cDCs, while CD11b cDCs expressed lower levels in steady-state (PBS) as compared to CD8a cDCs and maintained these levels after treatment. In the anti-apoptotic Bcl-2 family members analyzed, Bcl-xL was clearly increased in both subsets, being in CD8a cDCs more extensively up-regulated. Bcl-2 was maintained in CD11b cDCs and slightly down-regulated in CD8a cDCs, while PolyIC up-regulated Mcl-1, and A1 slightly, in both subsets.
These results show a change in the transcriptional program of pro-and anti-apoptotic Bcl-2 family members upon PolyIC treatment, with the strongest induction seen for Bim in both subsets. , and their splenic DC composition evaluated as described in Figure S1. The fold-change in total cells with PolyIC relative to PBS treatment is shown. Data are presented as mean +/2 SD. doi:10.1371/journal.pone.0020189.g002 Multiple BH3-only proteins mediate PolyIC-induced apoptosis of splenic cDC, including but not solely involving Bim Further to changes in gene expression, post-transcriptional and post-translational regulation of certain Bcl-2 family members is a common event, for instance proteolytic Bid activation. We wanted to examine whether the mitochondrial apoptotic pathway is responsible for splenic cDC depletion, and to identify the proapoptotic members possibly involved. To this end, we injected mice lacking Bim, Puma, Noxa, Bid or combinations of two of these BH3-only proteins (Puma 2/2 Noxa 2/2 , Bim 2/2 Puma 2/2 , Bim 2/2 Noxa 2/2 and Bim 2/2 Bid 2/2 ) with PolyIC and determined their cDC numbers after 40 h (Fig. 4). These different knock-out mice displayed normal DC populations (absolute numbers and 'CD8a + : CD8a 2 cDC' ratio) in steady-state as compared to WT mice (Fig.  S4). Following PolyIC treatment, mice singly deficient for Bid, Puma or Noxa displayed a reduction in both cDC subsets that was comparable to that seen in WT mice (Fig. 4C). Remarkably, while mice lacking Bim showed only a modest protection in cDC, similar to mice lacking both Puma and Noxa, a striking persistence of cDC numbers was seen in PolyIC-treated mice lacking Bim and a second BH3-only protein, either Bid, Puma or Noxa (Fig. 4C). This protection was already visually striking in the flow cytometric analysis of cDC subsets ( Fig. 4A and B), which showed that the CD8a cDC subset in particular did not disappear after PolyIC injection, in contrast to their profound depletion observed in WT animals.
These results support that apoptosis is responsible for the depletion of splenic cDC subsets in response to PolyIC, and that the mechanism requires multiple BH3-only proteins, including Bim plus either Puma, Noxa or Bid.

Type I IFN is necessary for the induction of splenic cDC depletion in response to PolyIC
We wanted to identify the signaling events downstream PolyIC detection that induced DC death. Double-stranded RNA can be detected by TLR3 in the endosome or by Rig-like Helicases (MDA-5 and RIG-I) that survey the cytosol compartment and signal via the adaptor MAVS. PolyIC in particular is a mimic of long dsRNA (.3 kbp), that may be detected by TLR3 and MDA-5, but not RIG-I [19]. CD8a cDC are known to express higher levels of TLR3 than the CD11b subset [4,20] (Fig. S5B). The expression of MAVS was found to be similar amongst splenic cDC subsets in WT mice (Fig. S5A). Recent proteomic analysis of DC subsets has furthermore confirmed such selectively higher levels of TLR3 in CD8a cDC at the protein level [21]. This might explain the greater extent of activation, expansion and subsequent depletion seen in CD8a cDC compared to CD11b cDC. PolyIC-induced cDC loss was therefore investigated in mice lacking either Ttlr3 or the adaptor Mavs (downstream of MDA-5) (Fig. 5). Surprisingly, loss of either Tlr3 or Mavs only resulted in partially increased persistence of splenic cDC following PolyIC treatment ( Fig. 5A and B). This suggested that apoptosis may be induced by converging factors downstream dsRNA detection, such as the induction of Type I IFNs. Accordingly, splenic cDC depletion following PolyIC administration was analyzed in mice lacking the alpha chain of the Type I IFN Receptor (Ifnar1 2/2 , hereafter termed 1IFNR 2/2 mice) 1 . Remarkably, loss of 1IFNR clearly protected both splenic cDC subsets from depletion following PolyIC treatment ( Fig. 5A and B).

TLR3 and MAVS cooperate in IFNß production in response to PolyIC treatment in vivo
The results showed that production of Type I IFN was the critical event that caused DC loss in vivo in response to PolyIC. Deficiency of either TLR3 or MAVS did not protect DC from PolyIC, suggesting that both TLR3 and MAVS cooperate in Type . Loss of two BH3-only proteins, including Bim, protects cDC subsets from PolyIC-induced depletion. Mice of the genotypes indicated were injected with PolyIC (n = 2 or 3 per strain) or PBS (control; n = 2 per strain, except n = 1 for Bid 2/2 ) and splenic cDC numbers determined after 40 h. A. Strategy for the flow cytometry analysis of splenic cDC subsets. Total splenocytes were gated for CD11c CD45RA 2 (R1), then CD11c high Gr-1 2 (R2). In order to exclude splenic DC precursors, CD11b cDC were gated as CD8a 2 SIRPa high cells within the cDC (R1xR2), since fully developed CD11b cDC express high levels of Sirpa and intra-splenic cDC precursors express only intermediate levels of SIRPa and CD11c [64]. One representative example of a WT and a Bim 2/2 Puma 2/2 mouse injected with either PolyIC or PBS (control) is shown. B. Splenic cDC subset segregation as outlined in (A). One representative sample is shown per strain and treatment. C. The fold-change in total cells elicited by injection with PolyIC relative to PBS treatment is shown. Data are represented as mean +/2 SD. P-values indicate significance per mouse strain as compared to WT controls. doi:10.1371/journal.pone.0020189.g004 I IFN production. We therefore assessed the relative contributions of TLR3 and MAVS in IFNß production in response to PolyIC. First, systemic production was analyzed in sera of WT, TLR3 2/2 , MAVS 2/2 or 1IFNR 2/2 mice after 4.5 h of PolyIC injection. In agreement with previous reports [6,22], MAVS was required for systemic production of this cytokine (Fig. 5C). Secondly, we analyzed IFNß production by DC. Splenic DC represent a minor fraction of cells in vivo (e.g. CD8a cDC are 0.25% of total cells in spleen) and their relative contribution to IFNß production might not be detectable in serum or cultures of whole splenocytes. We therefore analyzed DC-rich spleen fractions (containing cDC subsets, pDC and B cells) as well as purified CD8a or CD11b subsets. These were obtained from WT, TLR3 2/2 , MAVS 2/2 or 1IFNR 2/2 mice and cultured with PolyIC or medium alone for Figure 5. Type I IFN mediates splenic cDC loss and is cooperatively induced by TLR3 and MAVS in vivo. A and B. WT, TLR3 2/2 , MAVS 2/2 and 1IFNR 2/2 mice were treated with PBS (control) or PolyIC. Splenic cDC were analyzed at 40 h as described before. The segregation of cDC into CD8a and CD11b subsets is shown (A) for one representative experiment per strain. The fold-change in total cells with PolyIC relative to PBS treatment is shown (B) for experiments with TLR3 2/2 (n = 4) and MAVS 2/2 (n = 3) mice with WT mice as controls (n = 4), and for 1IFNR 2/2 (n = 3) mice with WT mice as controls (n = 3). P-values indicate significance per mouse strain as compared to WT controls. C. WT, TLR3 2/2 , MAVS 2/2 , and 1IFNR 2/2 mice were injected with PolyIC and the levels of serum IFNb examined 24 h before and 4.5 h after treatment (n = 3, representative of two independent experiments). P-values indicate significance comparing PolyIC treatment to PBS controls, per strain. D. Splenic DC fractions were isolated from WT, TLR3 2/2 , MAVS 2/2 and 1IFNR 2/2 mice as depicted in Figure S6 and cultured in medium (control) or with PolyIC for 4 h. IFNb production was analyzed in supernatants (n = 6 for DC-rich fractions, n = 5 for cDC subset fractions, from two independent experiments). P-values indicate significance comparing PolyIC treatment to medium alone, per strain. Data are presented as mean +/2 SD. doi:10.1371/journal.pone.0020189.g005 4 h (Fig. 5D and Fig. S6). In WT samples, IFNß was induced in both the DC-rich fraction and the purified CD8a subset, whereas the CD11b subset showed no secretion (Fig. 5D). Importantly, loss of TLR3 abrogated production of IFNß by CD8a cDC, while MAVS 2/2 CD8a DC showed impaired although still detectable induction of IFNß (Fig. 5D). These results support that TLR3 and MAVS cooperate to induce IFNß in response to PolyIC, in a celltype specific manner. Isolated splenic CD8a cDC required TLR3, while other cells relied on MAVS and produced IFNß systemically. Presumably, the relatively small contribution of splenic CD8a cDC to systemic IFNß production is not detectable in the serum of MAVS 2/2 mice.

CD8a DC lines rely on TLR3 for IFNß production and apoptosis with PolyIC
In addition to IFNß production by DC, we wanted to validate that DC undergo apoptosis in response to PolyIC, in isolation from other cells (in vitro), and assess the requirement of TLR3. Of note, several technical difficulties render the study of DC apoptosis challenging, including the scarcity of DC in the mouse and the fact that dying cells are rapidly cleared in vivo (rendering their detection difficult). We initially assessed whether apoptosis of DC induced by PolyIC treatment in vivo could be detected by a subsequent enrichment and 4 h culture of DC ex vivo, allowing for phosphatidylserine translocation (AnnexinV staining) with minimal phagocytosis by surrounding cells (Fig. S7A). Although a slight increase in AnnexinV staining was seen in cDC from PolyIC-treated mice (Fig. S7B), the rapid spontaneous activation and apoptosis of splenic cDC ex vivo (4 h in cDC from PBS-treated, Fig. S7B and C) rendered analysis of PolyIC-induced apoptosis in splenic cDC extremely difficult.
In order to avoid spontaneous apoptosis of DC, we have taken advantage of novel CD8a DC lines derived from the culture of tumors arising in our recently reported transgenic mouse model of Langerhans cell hiostiocytosis [23], harbouring the SV-40LgT oncogene under the control of the DC-specific CD11c promoter (Fig. S8A). The DC lines derived share all tested features with freshly-isolated CD8a DC, such as PAMP-induced up-regulation of MHC class II and co-stimulation molecules, cytokine secretion upon activation, and antigen-presentation and cross-presentation (Fuertes Marraco and Grosjean et al, manuscript in preparation). Following activation with PolyIC (Fig. S8B), DC lines displayed induction of apoptosis as confirmed by the standard AnnexinV/7-AAD staining as well as mitochondrial depolarization (Fig. S8C). Pertinently, the kinetics of apoptosis in DC lines was reminiscent of the depletion of splenic cDC observed in response to PolyIC in vivo (Fig. 1).
Several genetically deficient DC lines were derived from tumors in CD11c:SV40LgT-TG mice crossed to various genetically deficient backgrounds, including 1IFNR 2/2 , TLR3 2/2 and MAVS 2/2 . Upon treatment with PolyIC for 48 h, 1IFNR 2/2 DC lines were protected from PolyIC-induced apoptosis ( Fig. 6A and  B), in agreement with the requirement for the 1IFNR for splenic cDC loss in vivo (Fig. 5A and B). Strikingly, TLR3 2/2 DC lines were also resistant to PolyIC, whereas MAVS 2/2 DC lines remained normally susceptible (Fig. 6A and B). Importantly, MAVS 2/2 DC lines produced IFNß in response to PolyIC, whereas the resistance of TLR3 2/2 DC lines correlated with a lack of IFNß production (Fig. 6C), in agreement with the requirement of TLR3 for IFNß production by freshly isolated splenic CD8a cDC (Fig. 5D).
IFNb is sufficient to induce apoptosis of DC lines in vitro and splenic cDC loss in vivo We next investigated whether Type I IFN was not only necessary but also sufficient to induce apoptosis of splenic cDC.
Notably, except for 1IFNR 2/2 DC lines, all other DC lines readily underwent apoptosis after 48 h of treatment with IFNß, including the PolyIC-resistant TLR3 2/2 DC lines ( Fig. 6A and B). In addition, effector caspase-3/-7 and initiator caspase-8 activities were clearly evident in WT DC lines after ,30 h of stimulation with either PolyIC or IFNß (Fig. 6D).
Further to the preliminary evidence in DC lines, it was critical to investigate the sufficiency of IFNß to induce loss of splenic cDC in vivo. Experiments using WT and 1IFNR 2/2 mice showed that injection of IFNß was sufficient to deplete WT splenic cDC subsets to an extent similar to that achieved by treatment with PolyIC (Fig. 7). Moreover, TLR3 2/2 MAVS 2/2 double deficient mice were generated in order to test whether TLR3 and MAVS have a cooperative action in the loss of splenic cDC in response to PolyIC. Remarkably, TLR3 2/2 MAVS 2/2 splenic cDC were found to be resistant to PolyIC injection, but remained susceptible to treatment with IFNß (Fig. 7). These data support the co-operative action between TLR3 and MAVS in IFNß production in vivo, and demonstrate that IFNß is alone sufficient to induce splenic cDC loss in vitro and in vivo.

Discussion
PolyIC kills splenic cDC in vivo, in particular CD8a cDC, via the conjunction of multiple BH3-only proteins, including but not solely involving Bim There are contradictory reports on the effects of microbial stimulation on DC viability, and to date no reports exist on apoptotic mechanisms controlling DC populations in vivo. One prediction is that recognition of pathogens would enhance DC survival to guarantee potent DC function, thus promoting efficient adaptive immune responses. Conversely, given the harmful consequences of exaggerated immune responses, strong microbial stimulation might be expected to create a need for increased DC turnover. We show here that splenic cDC undergo apoptosis in response to PolyIC, particularly affecting the CD8a subset. Furthermore, loss of splenic cDCs occurred in the absence of T cells or NK cells, pointing to an apoptotic mechanism that is intrinsic to DCs and distinct from previous reports [12,13]. In our hands, concomitant CD40 cross-linking did not rescue CD8a DC from PolyIC-induced death (data not shown) [24].
Remarkably, while loss of Bim alone afforded only minor protection of DC, no DC loss was seen in PolyIC-injected mice lacking Bim plus either Puma, Noxa or Bid. Such correlation supports that DC apoptosis takes place in vivo in response to PolyIC. Whether DC that are deficient for two BH3-only proteins (including Bim) are generally more resistant to apoptosis cannot be excluded. To test the latter hypothesis, alternative unrelated treatments that induce apoptosis in splenic cDC would need to be identified and investigated in the different BH3-only deficient mice. Of note however, the different mice deficient for BH3-only members did not display increased DC populations in steady-state (neither numbers nor CD8a + -to-CD8a 2 cDC ratio, Fig. S4). This excludes the presence of abnormally high starting numbers of splenic cDC in mice deficient for BH3-only members that would then render DC loss with PolyIC less significant. The involvement of Bim together with other BH3-only proteins in splenic cDC loss suggests that it is the overall balance between the pro-apoptotic BH3-only proteins and the anti-apoptotic Bcl-2 members, which regulates cDC lifespan after PolyIC stimulation. A similar overlapping action of several BH3-only proteins has been reported in neutrophil apoptosis and in the context of inflammatory arthritis [25,26]. To our knowledge, this is the first report on the apoptotic mechanism regulating activated DC lifespan in vivo, mediated by the overlapping action of pro-apoptotic BH3-only proteins, including but not solely involving Bim.

The CD8a cDC subset is predominantly lost
Intriguingly, in response to PolyIC, LPS and CpG, the CD8a cDC subset showed greater up-regulation of co-stimulatory markers (e.g. CD86) and also greater subsequent depletion than the CD11b subset. Notably, CD8a cDCs displayed a higher steady-state level of gene expression for Bim, Puma and Bid as compared to CD11b cDCs (PBS controls), supporting the hypothesis of a higher intrinsic susceptibility of CD8a cDCs to apoptosis. Clearly in both subsets, PolyIC treatment further upregulated Bim and Puma. Interestingly, CD8a cDC also strongly up-regulated anti-apoptotic Bcl-xL in response to PolyIC, and down-regulated pro-apoptotic Bid to the levels of CD11b cDC, which were maintained upon treatment. This could reflect an attempt by CD8a cDC to overcome the initially higher proapoptotic to pro-survival Bcl-2 family member ratio than CD11b cDC, as a negative feedback mechanism in order to temporarily sustain their activation. The aforementioned results using mice genetically deficient for pro-apoptotic Bcl-2 family members nevertheless show that ultimately, in both subsets, Bim with a second BH3-only protein is sufficient to tilt the pro-to-anti-survival balance towards cell death. Possible speculations on such preferential death of the CD8a cDC subset are discussed further below. Type I IFN is necessary and sufficient for PolyIC-induced DC death, with a cell-type specific and cooperative involvement of TLR3 and MAVS in vivo A fundamental question was to dissect the signaling events that induced apoptosis of DC, and not only their activation, in response to PolyIC treatment. We show that DC depletion required the 1IFNR, and IFNb was necessary and sufficient to kill DC lines and deplete splenic cDC in vivo. In our hands, IFNß was predominantly detected over IFNa in DC culture supernatants (data not shown). However, while we addressed induction of DC apoptosis with IFNß, it cannot be excluded that other members of the Type I IFN family, notably the different IFNa proteins, may induce similar DC death.
Interestingly, neither endosomal TLR3 nor cytosolic MAVS were individually indispensable, rather double deficiency (TLR3 2/2 MAVS 2/2 ) was necessary to prevent DC loss in vivo. In response to PolyIC, TLR3/TRIF and MDA-5/MAVS signaling axes have been reported to cooperate in DC activation, antibody production, as well as CTL and NK responses [6,22,27]. In agreement with the latter reports [6,22], systemic IFNb production in response to PolyIC required MAVS. However, we have found that splenic cDC, in particular the CD8a subset and DC lines, rely on TLR3 for IFNb production. Such TLR3 requirement in CD8a cDC is in contrast to a previous report using isolated DEC205+ (CD8a+) DC [6], where rather MAVS is required. To our knowledge, we provide first evidence that CD8a cDC rely on TLR3 for IFNß production and in vitro apoptosis in response to PolyIC, in line with their characteristic expression of TLR3 [20,21]. Importantly, treatment with IFNb alone was sufficient induce apoptosis in TLR3 2/2 DC lines and loss of splenic cDC in mice doubly deficient for both TLR3 and MAVS.
Our findings support the hypothesis that both TLR3 and MAVS cooperate in vivo in a cell-type specific manner to produce IFNb upon PolyIC treatment and consequently splenic cDC loss. However, different cell types display a differential requirement for either of these signaling pathways, with systemic IFNß production depending on MAVS, and splenic CD8a cDC (including DC lines) requiring TLR3. This furthermore implies that both autocrine and paracrine Type I IFN production induces DC death.
Type I IFNs are well recognized anti-viral factors; virtually all cells may respond to Type I IFN, inducing diverse, and ubiquitous or cell-type specific effects [28]. Due to the pleiotropic nature of Type I IFN effects, many of the details of Type I IFN mechanism of action and their particular relevant scenarios remain to be elucidated [29]. Inducing an antiviral state, Type I IFNs may cause growth arrest and apoptosis in several cell types, particularly in transformed cells. As part of their anti-viral effectiveness, Type I IFN also induce immune activation, and may further induce proliferation and sustained survival of certain immune cell types such as T cells [30].
Induction of Type I IFN by PRR has been recently shown to be critical for DC function, in particular in the context of the adjuvant effect of PolyIC [6]. There is however limited evidence concerning the impact of Type I IFN in DC viability. In vitro, Type I IFN can antagonize the survival effects of TLR-Ls in fibroblasts and human MoDC [31], and IFNb induces death in BM-DC cultures only when they are concomitantly treated with a stimulatory cocktail [32]. Generally, Type I IFN alone is not sufficient to induce apoptosis in most cell types [29]. Rather, the pro-apoptotic effect of Type I IFN generally relies on sensitization to concomitant apoptotic triggers, such as treatments with TNF family members [33]. In addition, Type I IFN may itself induce TNF family members such as TRAIL, and consequently apoptosis [34], as shown in the context of multiple myeloma. TNFa is another important cytokine for DC activation, for example during viral infection [35]. Furthermore, the contribution of Bid in PolyIC-induced DC loss that we observed (Fig. 4) raises the question of whether death receptor signaling contributes to DC apoptosis. We therefore tested for a possible involvement of TNFa or TRAIL in splenic cDC depletion with PolyIC-induced Type I IFN. Reduction in both cDC subsets with PolyIC still occurred as extensively in TNFa 2/2 mice (Fig. S9A) or in WT mice treated with TRAIL-R2:Fc blockade (blocking soluble TRAIL) (Fig.  S9B). Other genetic deficiencies within the TNF or TNF-R family, including LTa 2/2 or TNFR1 2/2 TNFR2 2/2 mice (data not shown), did not affect the loss of splenic cDC with PolyIC.
Our results show for the first time that IFNb is sufficient to induce death of splenic cDC in vivo, in absence of accompanying stimuli, demonstrating a novel apoptotic effect of Type I IFN in a critical antigen-presenting cell.
What may be the physiological significance of DC death upon PAMP/Type I IFN stimulation?
Given the critical function of DC for the induction of adaptive immune responses, the apoptosis of DC in response to a microbiallike stimulus and/or Type I IFN (critical anti-viral mediators), affecting in particular the cross-presenting CD8a cDC subset, may appear paradoxical. Several speculations may be advanced concerning the physiological significance that such DC death may have.
DC apoptosis may limit the spread of infection and allow for an adequate shutdown of immune responses. Critically, DC apoptosis must maintain a healthy balance between excessive DC clearance leading to immunosuppression and the maintenance of immune competence without causing autoimmunity. Aberrant accumulation of DC results in chronic lymphocyte activation and systemic autoimmune manifestations, as shown in transgenic mice expressing the baculoviral pan-caspase inhibitor p35 under the DCspecific CD11c promoter [9]. Similarly, Bim-deficient DC display a decrease in spontaneous death in culture and their adoptive transfer induces autoantibody production in vivo [14]. Conversely, a reduction in DC numbers has been linked to states of immune suppression, such as sepsis [36,37] and chronic viral infection. Impairment of DC function during LCMV infection was reported and linked to immunosuppressive states, with the LCMV-WE being more immunosuppressive and persistent than LCMV-Armstrong [17]. Interestingly, we observed that LCMV-WE infection causes more extensive cDC depletion than previously reported for LCMV-Armstrong [16]. Several studies have detected an impairment of DC function upon TLR stimulation [38], similarly to chronic inflammatory or infectious states [39,40]. This has been thought to be a consequence of loss of antigen uptake and processing in activated DC, but may well (at least in part) have been due to the here-described DC death.
1IFNR 2/2 mice succumb to infection with many viruses, showing the importance of Type I IFN for antiviral responses, including stimulation and clonal expansion of CTL [41,42]. However, a Type I IFN response may also be detrimental under certain circumstances, such as infection with listeria monocytogenes [29,43]. While Type I IFN is required for the initiation of adequate anti-viral responses, the DC apoptosis we report could be a negative secondary effect of Type I IFN, either when ''inappropriately'' induced or upon persistent infections. In support of this idea, persistence of viral infection and immune suppression has been correlated with reduced DC numbers and sustained levels of Type I IFN, in models such as LCMV-Armstrong infection in mice [44], as well as infections with HIV-1 in humans [45] and SIV in monkeys [46]. Collectively, our findings suggest that one major cause of transient or chronic immunosuppresion during infectious conditions is a consequence of apoptotic depletion of DC linked to sustained and/or aberrantly high Type I IFN levels.
The DC apoptosis we report is particularly relevant for the therapeutic exploitation of TLR-Ls and Type I IFNs (e.g. adjuvants, multiple sclerosis, cancer, Langerhans cell histiocytosis) [2,47,48,49]. Both the adjuvant immune stimulatory and direct anti-proliferative properties of Type I IFN have a strong potential for therapeutic exploitation for cancer treatments [50,51]. Similarly, IFNß has been used for the treatment of autoimmune diseases such as multiple sclerosis [52]. It would be critical to investigate whether current or developing therapies featuring the use of Type I IFN or TLR-Ls induce DC loss. Accordingly, therapeutic regimes should be carefully designed (e.g. dosage, timing of antigen/adjuvant treatment) in order to avoid the potentially deleterious side effects of DC apoptosis leading to immunosuppression.
CD8a DC are particularly potent at phagocytosing dying cells for antigen cross-presentation and at inducing CTL responses [53], being thus one key target for vaccines against cancer and infectious diseases [54,55,56]. Strategies for vaccine enhancement include the targeting of antigens to CD8a cDC [2,57], for which PolyIC has been shown to be an optimal adjuvant [2]. Pertinently, several lines of evidence support that TLR3 expression by CD8a DC and downstream Type I IFN production is critical for the adjuvant effect of dsRNA (PolyIC), including cell-associated dsRNA for cross-presentation [58,59,60,61]. Given the unique functions of CD8a cDC, their preferential death with a virus-like stimulus (PolyIC) and Type I IFN (anti-viral mediator), as compared to the other CD11b DC subset, appears contradictory. However, on the one hand, apoptosis may be an evolutionary adaptation to limit the spread of intra-cellular pathogens within such critical DC subset. On the other hand, apoptosis may serve as an efficient mechanism to amplify the response to intracellular pathogens, by an apoptosis-mediated ''DC to DC cross-presentation'' of antigens. Via this mechanism, infected cells (DC) would be sacrificed, while pathogen would be continuously processed and diluted amongst neighboring cross-presenting DC, correlating with pathogen load. Such apoptosis-driven ''DC to DC crosspresentation'' would be particularly relevant when antigen is limiting, or considering the naturally low frequency of DC, to mount an adequate and rapid adaptive immune response.

Perspectives
Future experiments should shed light on the physiological impact of DC death with PolyIC and Type I IFN. It will be addressed whether DC death promotes amplification of crosspresentation, protects from autoimmune pathology, induces sideeffect immunosuppression, and/or affects vaccine efficiency. In addition to antigen targeting to DC and the use of TLR-Ls as adjuvants, efforts to optimize vaccine efficacy also include increasing DC viability after activation [62,63]. Our findings have the potential to contribute to the optimization of vaccine therapies, providing an important basis for the design of strategies to enhance DC viability and consequently function.

Ethics statement
Animal experiments were performed in strict accordance to the Swiss Federal Regulations. The protocol was approved by the ''Service de la consommation et des affaires vétérinaires du Canton de Vaud'', Switzerland (Permit Number: 1847.1). All efforts were made to minimize suffering and minimize the number of mice needed to assess statistical significance and experimental reproducibility. The generation and use of mouse DC lines for in vitro research represents a considerable advance towards the implementation of the 3R (Refine, Reduce, Replace) principle in animal experimentation.

DC isolation, cultures and treatments
The medium composition was IMDM-glutamax (GIBCO 31980) supplemented with 8-10% heat-inactivated FCS, 10 mM Hepes, 50 uM b-mercaptoethanol, and 50 U/mL of penicillin and 50 mg/mL streptomycin. Cells were cultured at 37uC with 5% CO 2 . Different DC populations were obtained from spleens as depicted in Fig. S6. Briefly, splenocyte suspensions were fractionated by density centrifugation in isohexol carbohydrate medium (Nycodenz, Axis-Shield, Norway) at 1.077 g/cm 3 [64]. The DC-enriched, light density fraction was collected as the 'DCrich fraction' (Fig.S6, fraction 1). Purified cDC subsets (.95%) were obtained from DC-rich fractions using anti-CD11c antibodycoupled magnetic micro-beads (Miltenyi Biotech) and flow cytometry cell sorting of CD11c + /B220 2 cells into CD8a DC or CD11b DC (Fig. S6, fraction 2 or 3, respectively) using a FACSAria cell sorter (Becton Dickinson). DC lines were derived as detailed in Materials and Methods S1. For apoptosis assays, DC lines were seeded at 2610 5 cells/mL the day prior and treated with medium, PolyIC at 5 mg/mL, or CTRL and IFNbcontaining supernatants as detailed below. For IFNß production assays, supernatants were obtained from DC cultures (DC lines or DC ex vivo) at 10 6 cells/mL treated with medium or PolyIC at 25 ng/mL for 4 h. The PolyIC used was Poly(I:C)-HMW from Invivogen and no transfecting reagent or liposomal vehicle was used.

Recombinant IFNß production and treatments
Recombinant murine IFNß was produced in house by lentiviral transduction of 1IFNR 2/2 DC lines, allowing to obtain a DCconditioned medium containing IFNß, in the absence of secondary effects from the IFNß on the producer cells (see Materials and Methods S1 for the details). Concentrated IFNß (1-5 ug/mL) was obtained using high cell density culture systems (CELLine AD1000, Integra-Biosciences) and diluted to 25 ng/mL of IFNß for experiments in vitro, with mock-transduced DC line supernatant used as a control. For treatments in vivo, mice were injected i.p. with 5610 6 IFNß-producing cells or mock-transduced cells (CTRL).

IFNb quantitation
IFNb quantitation was performed on cell culture supernatants or peripheral blood sera by ELISA (PBL 42400, PBL Interferonsource).
Caspase activity assays. The fluorogenic substrates used were ac-IETD-amc for Caspase-8 and ac-DEVD-amc for Caspase-3/-7 (Enzo Life Sciences). Caspase-activity assays were performed on DC line samples according to the manufacturer's protocol, measuring caspase-3/-7 activity after 60 min and caspase-8 activity after 120 min and normalizing data to protein content quantitated by the standard Bradford reaction.

Statistical analyses
P-values were obtained using two-tailed unpaired t-tests with 95% confidence intervals (ns = not significant; * = p,0.05; ** = p,0.01; *** = p,0.001). Figure S1 Flow cytometry analysis of Rag2 -/c c -/mice treated with PBS or PolyIC. Rag2 -/c c -/mice were injected i.p. with 50 ug PolyIC (n = 3) or, as a control, PBS (n = 3), and their splenic composition compared to that of similarly treated WT mice (n = 3 for PolyIC, n = 3 for PBS). Consistent with previous observations [65], flow cytometric analyses of the spleens of these rag2 -/c c -/mice was complicated by the presence of autofluorescent cells. A. Rag2 -/c c -/mice lack B cells, T cells and NK cells. The autofluorescent cell population in the staining for B cells, T cells and NK cells is visible on a fluorochrome-empty flow cytometry channel (FL4). Staining profiles for representative PBStreated mice is shown for each strain. B. Gating strategy for the analysis of splenic cDC, excluding pDC, monocyte-derived DC and autofluorescent cells by a first gate on GR1 and B220 doublenegative cells (an additional flow cytometry channel was not available in the staining combination for cDC analysis). CD11c + cells from the GR1 2 B220 2 were selected for segregation into CD8a + vs CD11b + subsets. Data from one representative mouse for each strain and treatment is shown. C. Total numbers of each cDC subset were calculated on the basis of total splenocytes harvested multiplied by the fraction of cells per gate as indicated in (b). To obtain the fold-changes in the cDC subset cellularities elicited by injection with PolyIC, the total number of DC in the spleens of mice treated with PolyIC were divided by the total number of DC in the spleens of PBS-treated mice. P-values indicate significance per strain as compared to WT controls. (TIF) Figure S2 Loss of splenic cDC after CpG and LPS injection. Mice (C57BL6) were injected i.p. with PBS (control; n = 4), 50 ug of CpG (n = 3), 10 ug of LPS (n = 3) or, as comparison to data shown in Figure 1, with 50 ug PolyIC (n = 1). Splenic cDC subsets were analyzed at 12 h post-injection for expression of the activation marker CD86 (GeoMFI = geometric mean of fluorescence intensity) and at 40 h for total numbers of cDC subsets. P-values indicate significance per treatment compared to injection with PBS. (TIF) Figure S3 Changes in gene expression of Bcl-2 family members in cDC treated with PolyIC in vivo. Mice were injected with 50 ug PolyIC (n = 9) or PBS (control; n = 9). After 14 h, the spleens were harvested and pooled for cDC subset isolation and gene expression analysis. The experiment was performed twice in two independent laboratories as described in the Methods section, with a total of four biological replicates per sample. a. Strategy for flow cytometric analysis, gating and isolation of CD8a or CD11b cDC subsets from total splenocytes. One representative experiment is shown. b. Analysis of the levels of mRNA for pro-apoptotic Bcl-2 family members Bim (n = 4), Puma (n = 4), Noxa (n = 4), Bid (n = 3) and anti-apoptotic Bcl-2 (n = 4), Mcl-1 (n = 4), A1 (n = 3) and Bcl-xL (n = 3). Open symbols (#, e) correspond to qRT-PCR data generated at the WEHI (Melbourne), normalized to b-actin as a house-keeping gene. Closed symbols (N, ¤) correspond to qRT-PCR data generated at the DB-UNIL (Lausanne), normalized to TBP as a house-keeping gene. The primers used were as detailed in Table  S2A. In order to obtain an inter-experimental and inter-laboratory comparison, a second normalization was performed relative to CD8a cDCs treated with PBS. (TIF) Figure S4 Steady-state splenic cDC populations in mice deficient for BH3-only members. Splenic cDC subsets were quantitated as described before amongst WT (n = 6), Bim -/-(n = 4), Puma -/-(n = 2), Noxa -/-(n = 4), Bid -/-(n = 3), Puma -/-Noxa -/-(n = 2), Bim -/-Puma -/-(n = 2), Bim -/-Noxa -/-(n = 4), Bim -/-Bid -/-(n = 2). A. Ratio of CD8a + -to-CD8a 2 cDC per mouse strain as indicated. B. Total numbers of each cDC subset per spleen per mouse strain as indicated. Data are represented as mean +/-SD. P-values indicate significance per strain as compared to WT controls. (TIF) Figure S5 Expression levels for TLR3, MAVS and the IFNaR1 in splenic cDC subsets and DC lines. Purified splenic CD8a as well as CD11b cDC subsets and DC line samples (n = 2 per sample) were examined for the levels of mRNA for Mavs and Ifnar1 by qRT-PCR using the primers detailed in Table S2B. Tlr3 mRNA expression was analyzed by semiquantitative PCR. (TIF) Figure S6 Strategy and flow cytometry analysis for the purification of DC-rich and cDC subset fractions. Splenocyte suspensions were fractionated by density centrifugation in isohexol carbohydrate medium (Nycodenz, Axis-Shield, Norway) at 1.077g/cm3. The light density fraction was collected as the DC-rich fraction (fraction 1, as indicated). CD11c+ cDC were isolated from DC-rich fractions using anti-CD11c antibody coupled magnetic micro-beads (Miltenyi Biotech). Purified cDC subsets were obtained from cDC by flow cytometry cell sorting separating CD11c + /B220 2 cells into CD8a or CD11b subsets, collecting fractions 2 and 3, respetively, as indicated. One representative sample is shown. The cDC represent 1-2% of total splenic leukocytes. Within the cDC population, typically ,20% were CD8a + cDC and ,60% CD11b cDC. At least 5 mice were required to isolate ,0.5610 6 CD8a + cDC. Density centrifugation allowed for a ,10-fold enrichment in DC prior to immunomagnetic bead selection. Enriched DC preparations subjected to flowcytometric cell sorting contained .75% cDC, and resulting cDC subsets were purified to at least 95%.