Differential Responses to IFN-α Subtypes in Human T Cells and Dendritic Cells ¹²

Type I IFNs (IFN-αβ) are a family of homologous cytokines. They were originally identified for their potent antiviral activity but have also been shown to possess both antiproliferative and a wide range of immunoregulatory activities (1, 2). IFN-αβ are the most frequently used cytokines for the treatment of a wide variety of diseases, including multiple sclerosis, viral hepatitis, and melanoma (3, 4).

IFN-αβ are potently induced in response to viral infection but can also be produced upon exposure to nonviral pathogens including Gram-negative bacteria and protozoa (e.g., Escherichia coli, Leishmania major) (5, 6). The IFN-αβ form part of the innate cytokine response. They are rapidly produced upon infection and trigger early, nonspecific defense mechanisms, including the induction of genes that inhibit viral replication and the activation of NK cell cytotoxicity. IFN-αβ also provide an important link between innate and adaptive immunity by shaping many components of the Ag-specific immune response. For example, IFN-αβ promote CTL responses, enhance Ab production, and facilitate the development of naive Th cells toward the Th1 phenotype. These late immunoregulatory effects of IFN-αβ are thought to be mediated through indirect rather than direct actions, and modulation of cytokine/cytokine receptor expression and dendritic cell (DC) ⁵ function are likely to be involved (7, 8, 9).

In humans, there are multiple IFN-α genes and a single gene for IFN-β (2, 10, 11). At least 13 IFN-α genes are transcribed. The coding sequences of these genes diverge up to 8%, giving rise to 12 different functional subtype proteins (IFN-α1 and -α13 are identical). All IFN-α subtypes and IFN-β are structurally similar and share a common cell surface receptor, consisting of two known subunits, IFNAR1 and IFNAR2 (12, 13, 14). Typically, binding of IFN-αβ to the receptor complex initiates a signaling cascade through the activation of the tyrosine kinases Tyk2 and Janus kinase (Jak)1 and, subsequently, phosphorylation and dimerization of STAT1 and STAT2 (15). Activated STAT dimers dissociate from the receptor complex and translocate to the nucleus where they regulate the expression of IFN-stimulated genes (ISGs) (5). A number of transcription factor complexes are formed in response to IFN-αβ. ISG factor 3 (ISGF3) consists of STAT1, STAT2, and p48/ISGF3γ/IFN regulatory factor (IRF)9 and binds IFN-stimulated response elements (ISRE), present in many ISG promoters. STAT1-2 heterodimers and STAT1 homodimers drive the expression of a subset of ISGs through γ-activated sequence (GAS) elements. Other members of the STAT family are involved in IFN-αβ signaling. For instance, IFN-αβ have been found to activate STAT3 in most cell types, STAT4 and STAT5 in T cells and NK cells (16, 17, 18), and STAT6 in B cells (19). In addition, pathways mediated through phosphatidylinositol 3 (PI3) kinase (20, 21), p38 mitogen-activated protein (MAP) kinase (22, 23, 24), and extracellular signal-regulated kinase 1/2 (25) have been implicated in IFN-αβ signaling, but their precise roles remain to be established.

The existence of so many different IFN-α subtypes raises the question of their biological relevance. It is known that different viruses induce different patterns of IFN-α subtype expression (26, 27) and that the production of the individual subtypes is regulated by different IRFs (28, 29, 30). Distinct relative potencies have been reported for individual subtypes in vitro and in vivo (31, 32, 33, 34, 35, 36, 37, 38). However, all IFN-α subtypes likely have the ability to exert the classical IFN-αβ activities (e.g., antiviral effects), as might be expected granted that they act through a common receptor. However, there is substantial biochemical evidence that individual IFN subtypes bind to different sites of the IFN-αβR and have different binding affinities, leading to the formation of distinct signaling complexes (39, 40, 41, 42, 43, 44, 45), which, a priori, might be expected to differentially recruit downstream signaling pathways. Accordingly, it is reasonable to assume the following: 1) superimposed upon widely overlapping classical functions, such subtype-specific signaling complexes may be capable of differentially activating ancillary responses peculiar to that subtype; 2) any such differentials likely vary with cellular background and will be manifest by differentials in gene expression profiles; and 3) accepting the importance of the IFN-αβ in immune function, IFN-α subtype differentials will be most obvious and appropriately analyzed in primary cells of the immune system.

In this study, signaling through the Jak/STAT pathways and gene expression profiles were examined for human T cells and DC treated with different IFN-α subtypes. In addition to quantitative differences between IFN-α1, -α2, and -α21 in their ability to activate STAT1–5 and ISGs, evidence for IFN subtype and cell type specificity and differential requirements for additional pathways were found. A striking example was provided by the induction of IFN-γ-inducible protein-10 (IP-10), a chemokine which plays an essential role in the recruitment of Th1 cells and NK cells to inflammatory sites. It was highly induced by IFN-α2 and -α21 but to a much lower extent by IFN-α1 in DC. In contrast, it was poorly induced by all of these IFN-α subtypes in T cells. Optimum induction of IP-10, unlike that of other classical ISGs analyzed, was dependent on the activation of p38 MAP kinase(s). Inducible NO synthase (iNOS) and IL-12Rβ2 were also differentially induced in DC and T cells. Substantial differences in the kinetics of accumulation of mRNAs for different subsets of ISGs add to the complexity of the IFN-α response.

PBMC were isolated from buffy coats by density centrifugation on Lymphoprep (Nycomed, Oslo, Norway). To obtain T cells, PBMC were activated with PHA (Murex, Kent, U.K.) and maintained in RPMI 1640 supplemented with 10% inactivated FCS and human rIL-2 (20 ng/ml) for 1 wk. Before IFN-α treatment, T cell blasts were washed and subsequently cultured for 48 h in the absence of rIL-2. To generate DC, monocytes were isolated from PBMC by MACS sorting using anti-CD14-conjugated magnetic microbeads (Miltenyi Biotec, Bergisch Gladbach, Germany) and cultured for 6 days in RPMI 1640 supplemented with 10% inactivated FCS, 50 ng/ml GM-CSF, and 50 ng/ml IL-4 (both from R&D Systems, Minneapolis, MN). The following bacterial rIFN-α subtypes were used: IFN-α1 (IFN-αD; specific activity, 5.0 × 10⁷ U/mg), IFN-α21 (IFN-αF; specific activity, 6.3 × 10⁸ U/mg), both from PBL Biomedical Laboratories (New Brunswick, NJ), and IFN-α2 (IFN-α2a; Roferon; specific activity, 5.4 × 10⁸ U/mg) from Roche (Basel, Switzerland). All cytokines were free of endotoxin, as determined by the Limulus amebocyte lysate assay (i.e., <0.1 endotoxin U/ml). The SB203580 and LY294002 compounds were from Sigma-Aldrich (St. Louis, MO).

Cells were treated for 20 min with IFN-α subtypes and subsequently lysed in 0.5% Nonidet P-40, 50 mM Tris (pH 8.0), 10% glycerol, 150 mM NaCl, 1 mM DTT, 1 mM EDTA, 1 mM sodium orthovanadate, 1 mM PMSF, 100 U/ml aprotinin, and 1 μg/ml leupeptin. Lysates were cleared by centrifugation and were either used directly for SDS-PAGE or for immunoprecipitation with anti-STAT2 Abs (Santa Cruz Biotechnology, Santa Cruz, CA) and protein A-Sepharose. Affinity purification of DNA binding proteins was performed as described previously (46), using a biotinylated DNA oligo containing a consensus GAS element (5′-GTGGCTTTCCGGGAATCCTTG-3′). Proteins eluted in loading buffer were electrophoresed on 7.5% polyacrylamide-SDS gels and transferred to polyvinylidene difluoride membranes (Immobilon-P; Millipore, Bedford, MA). Membranes were probed with the following Abs: anti-phosphotyrosine Abs PY20 (Affinity, Nottingham, U.K.) and 4G10, anti-STAT1 phosphoserine 727 (both from Upstate Biotechnology, Lake Placid, NY), anti-STAT1 phosphotyrosine 701, anti-phospho-p38 and anti-p38 (all from New England Biolabs, Beverly, MA), anti-STAT1 p91 (NovoCastra, Newcastle, U.K.) and anti-STAT2, anti-STAT3, anti-STAT4, and anti-STAT5B (all from Santa Cruz Biotechnology). Following ECL and autoradiography, membranes were stripped in 62.5 mM Tris (pH 6.8), 2% SDS, and 100 mM 2-ME, and then reprobed as required.

Total RNA was isolated from cells using Trireagent (Helena Biosciences, Sunderland, U.K.) according to the manufacturer’s instructions. RNA quantity and quality was assessed by determination of the OD₂₆₀ and OD₂₈₀ using spectrophotometry and additional visualization by agarose gel electrophoresis.

Radiolabelled cDNA was generated from 5–10 μg of total RNA by reverse transcription with Superscript II (Life Technologies, Rockville, MD) in the presence of [³³P]dCTP. Residual RNA was hydrolyzed by alkaline treatment at 70°C for 20 min, and the cDNA was purified using G-50 columns (Amersham Pharmacia, Bucks, U.K.). Before hybridization to the macroarrays, the labeled cDNA was mixed with 50 μg of COT-DNA (Life Technologies) and 10 μg of poly(A) DNA (Sigma-Aldrich), denatured at 95°C for 5 min and hybridized for 1 h to minimize nonspecific binding. Preparation of the macroarrays (representing 150 known ISGs), hybridization of the radioactive cDNAs, and scanning and analysis of the macroarrays were conducted as described previously (47).

RPAs were conducted as before (48). The IP-10, IL-12Rβ2, iNOS, suppressor of cytokine signaling (SOCS)1, SOCS3, and GAPDH probes were purchased from BD PharMingen (San Diego, CA). Briefly, probes were synthesized from SP6/T7 transcription vectors and labeled with [³²P]UTP to 2–5 × 10⁸ cpm per microgram of input DNA. Aliquots equivalent to 1–3 × 10⁵ cpm of each probe and 5–10 μg of total RNA were used per assay. The intensities of radioactive bands were quantified using a PhosphorImager (Storm; Molecular Dynamics, Sunnyvale, CA). Bands of interest were quantitated, corrected for background, and normalized to the control band (i.e., GAPDH or γ-actin). Results are expressed as PhosphorImager intensities.

DC were cultured in 96-well plates (1 × 10⁵ DC per 200 μl) and treated with IFN-α subtypes for 18 h. IP-10 protein levels in supernatants were determined by specific ELISA (HyCult Biotechnology, Norwood, MA).

The ability of IFN subtypes IFN-α1, -α2, and -α21 to activate the Jak/STAT pathway and induce a substantial set of ISGs was examined in T cells and DC. These subtypes were chosen because 1) IFN-α1 and -α2 are generally the major IFN species produced in response to many viruses and poly(I:C) (49, 50, 51), 2) rIFN-α2 is widely used therapeutically (4), and 3) IFN-α21 exhibits enhanced growth inhibitory activity compared with IFN-α2 (45). In addition, both recombinant and natural IFN-α1 have a ∼10-fold lower specific antiviral activity than the majority of subtypes including IFN-α2 and -α21 (52, 53). Consistent with this, the specific (antiviral) activities of the rIFNs used were in the range of 5 × 10⁷ and 5 × 10⁸ IU/mg protein for IFN-α1 vs -α2 and -α21, respectively. In accordance with current practice, absolute concentrations of the subtypes are given throughout. For easy reference, absolute concentrations of 2 ng/ml for IFN-α2 and -α21 and 20 ng/ml for IFN-α1 each correspond to ∼1000 IU/ml based on antiviral activity.

Human T cells and DC were treated for 20 min with a range of concentrations of the different IFN-α subtypes. Whole-cell extracts (Fig. 1, A, upper panels, and C) or immunoprecipitated STAT2 proteins (A, lower panels) were fractionated by SDS-PAGE, and phosphorylation of STATs was examined by Western blot analysis with Abs specific for phosphorylated Tyr and/or Ser residues. Alternatively, activated STAT dimers were affinity purified by their ability to bind to a biotinylated oligonucleotide containing a consensus GAS element, and the bound, activated STATs were similarly analyzed (Fig. 1 B).

In T cells, we observed that all three IFN-α subtypes induced the tyrosine phosphorylation of STAT1 to -5 (Fig. 1, A and B). However, phosphorylation/activation of these STATs by IFN-α1 was comparatively poor: 10- to 100-fold higher absolute concentrations of this subtype were required for detectable tyrosine phosphorylation for each of the STATs (Fig. 1, A and B). Similarly, 10- to 100-fold higher concentrations of the IFN-α1 subtype were required for a positive signal to be detected in the less-sensitive Ser⁷²⁷ phosphorylation assay for STAT1 (Fig. 1,A). No such phosphorylation was detected by the whole-cell extract assay at any of the IFN-α1 concentrations tested (up to 10 ng/ml; Fig. 1,A). However, at higher concentrations (20 and 200 ng/ml), Ser⁷²⁷ phosphorylated forms of STAT1 could be observed in the GAS oligo pull-down assay. In this assay (Fig. 1 B), the two STAT1 forms with distinct electrophoretic mobility represent STAT1 phosphorylated on tyrosine (lower band) and STAT1 phosphorylated on both tyrosine and serine (higher band) (46). For IFN-α2 and IFN-α21, similar dose-response curves for both tyrosine and serine phosphorylation were observed, although IFN-α2 appeared slightly more potent at lower concentrations.

In DC, comparison of the responses to IFN-α1 and IFN-α2 similarly revealed large quantitative differences between these subtypes in their ability to induce tyrosine and serine phosphorylation of STAT1 (Fig. 1 C) and STAT3 (data not shown) and to induce formation of an ISGF3 complex as determined by EMSA (data not shown).

The kinetics of STAT activation in T cells (Fig. 2 A) and DC (B) was also examined. Phosphorylation of STAT1 on Y701 was sustained for 1–2 h and declined thereafter. Allowing for the quantitatively lower activity of IFN-α1, no apparent differences between the IFN-α subtypes in the kinetics of STAT1 Y701 phosphorylation were observed. Similar analyses of the kinetics of activation of STAT2, -3, and -4 in T cells and STAT3 in DC revealed no differences between IFN-α subtypes (data not shown).

Taken together, the results establish that there are substantial quantitative, but not qualitative, differences between the IFN-α subtypes with respect to STAT activation in both T cells and DC.

Customized cDNA-based macroarrays, each representative of 150 known ISGs in triplicate (Materials and Methods and Ref.47), were used in an initial screen for differential gene expression in response to the three IFN-α subtypes. In prior experiments, for IFN-α2, in HT1080-based cell lines, optimum induction of virtually all of the ISGs represented was typically observed at 2 ng/ml for 6–8 h (47). Accordingly, in this study, the responses to all three subtypes were analyzed at 2 ng/ml, corresponding to physiological concentrations of ∼1000 IU/ml for IFN-α2 and IFN-α21 and 100 IU/ml for IFN-α1 (Materials and Methods). To compensate for the lower specific antiviral activity, the IFN-α1 subtype was also assayed in DC at the higher absolute concentration of 20 ng/ml (1000 IU/ml). Total RNA from untreated and subtype-treated cells was analyzed. The results are shown as fold induction compared with untreated cells. Data for those genes induced 2-fold or more (27 in T cells and 36 in DC), including the classical ISGs MxA, 2′-5′-oligoadenylate synthetase, and PKR known to be involved in antiviral responses, are presented (Table I). As expected for ISGs, no significant repression (>2-fold) of any of the genes represented was observed. As for STAT activation, in both T cells and DC, in general, ISGs were more potently induced by IFN-α2 and -α21 than by IFN-α1. Once again, IFN-α21 was slightly less effective than IFN-α2 (Table I). IFN-α1 at 2 ng/ml is sufficient to give STAT1 and -2 activation (Figs. 1 and 2). Consistent with this, although less effective than IFN-α2 or -α21, IFN-α1 at 2 ng/ml clearly induced, particularly in DC, the more highly inducible of the spectrum of ISGs observed with the more potent IFN-α2 and -α21 subtypes. That any differential is largely quantitative was essentially confirmed by the closer match observed on increasing the IFN-α1 concentration from 2 to 20 ng/ml (Table I, DC, first column). Interestingly, however, IFNs α2 and α21 induced the expression of the chemokine IP-10 and of iNOS mRNA in DC but not T cells. This provides a further example of cell type specificity in the IFN-α response (47). More importantly in this study, with respect to IFN-α subtype specificity, no significant induction of the IP-10 gene was observed in DC in response to IFN-α1, even at the increased concentration of 20 ng/ml, whereas it was highly induced in these cells by 2 ng/ml of IFN-α2 (14.4-fold) and IFN-α21 (6.6-fold) (Table I). Potentially similar IFN-α subtype-specific differentials were observed for a number of other ISGs (e.g., RING4, GBP-1, IFI16, and Caspase 1). However, these differentials were both less profound and less consistent among the three donor samples analyzed and remain to be established. Of those analyzed, IP-10, therefore, emerged as the most likely candidate for an IFN-α subtype-specific response gene. Accordingly, the induction of IP-10 mRNA and protein was analyzed further.

The kinetics of induction of IP-10 mRNA was compared by RPA with those for a selection of IFN-α-inducible mRNAs in both DC and T cells treated with IFN-α1 (2 and 20 ng/ml) or IFN-α2 (0.2 and 2 ng/ml) for 1, 4, and 8 h (Fig. 3). Differentials 1) in the kinetics of induction of different ISGs, 2) between cell types, and 3) between the IFN-α subtypes, were observed.

The induction of SOCS3 mRNA was exceptionally early and transient, that of ISG-56K was early but prolonged, whereas that of even the primarily induced 6-16, 9-27, and STAT1 mRNAs appeared slightly delayed. Clearly, the point of detection depends on the sensitivity of the assay, but the differentials in the shapes (transient vs prolonged) of the curves remain. A transcriptional response for all of the major mRNAs identified (Fig. 3 and Table I) is well established. As we are looking at mRNA accumulation, additional modulation through differentials in mRNA stability cannot, of course, be excluded.

In response, for example, to 2 ng/ml IFN-α2, IP-10 and iNOS were highly induced in DC (IP-10, 242-fold at 4 h; iNOS, 30-fold at 1 h; Fig. 3,A, lanes 8 and 4) but only marginally in T cells (IP-10, 4-fold at 4 h; iNOS 2-fold at 1 h; A, lanes 21 and 17), thus confirming the cell type differentials observed in the macroarrays (Table I). An additional example of cell type specificity was provided by the IL-12Rβ2 gene, which was up-regulated in T cells (Fig. 3 A, lanes 19–22), but not in DC (A, lanes 1–13).

In DC, at all times analyzed, IP-10 mRNA was highly induced by IFN-α2 (0.2 and 2 ng/ml), but much less efficiently by 10-fold higher concentrations (2 and 20 ng/ml) of IFN-α1 (Fig. 3). This is especially evident at the 8 h time point in Fig. 3,A (compare lanes 10 and 11 with 12 and 13) and is in accord with the virtual absence of an IFN-α1-induced IP-10 response at 8 h in the macroarray (Table I). Consistent with this, elevated IP-10 mRNA levels were still detectable at 24 h in response to IFN-α2 (2 ng/ml) but not IFN-α1 (20 ng/ml) (data not shown). Furthermore, importantly, significantly less IP-10 protein was produced in response to IFN-α1 than IFN-α2 (Fig. 4). Indeed, at least 100-fold higher concentrations of IFN-α1 (20 ng/ml) than IFN-α2 (0.2 ng/ml) were required to induce comparable amounts of IP-10 protein (Fig. 4), a result entirely consistent with the similar differential in the efficiency of induction of the IP-10 mRNA (Fig. 3, A, compare lanes 2 and 5, 6 and 9, 10 and 13, and B). This contrasts with the data for the vast majority of ISGs, for which the induction in response to IFN-α2 (0.2 and 2 ng) and 10-fold higher concentrations of IFN-α1 (2 and 20 ng/ml) were highly comparable (Fig. 3 and Table I). Thus, although with absolute concentrations adjusted 10-fold to compensate for the 10-fold difference in specific antiviral activity and receptor affinity (see above), IFN-α1 and -α2 can exert similar antiviral activities and induce most ISGs to a similar extent, they differ much more in their ability to induce IP-10. The data suggest that the optimum induction of IP-10, but not the majority of ISGs scoring as induced in Table I, may require, at least in DC, an additional signal(s) activated by IFN-α2, but not IFN-α1.

The IP-10 promoter contains an ISRE (54). Activated STAT1 and -2 are required (together with preexisting p48/ISGF3γ/IRF9) for the formation of ISGF3, which mediates IFN-α responses through ISREs. Therefore, it can be reasonably assumed that activated STAT1 and -2 are essential (although not necessarily sufficient) for the induction of IP-10 by IFN-αs. No differential in the activation of STAT1 and -2 or in the formation of ISGF3 in response to IFN-α1 and -α2 was observed (e.g., Fig. 1). The differences in IP-10 induction by the IFN-α1 and -α2 subtypes cannot, therefore, be explained by differences in STAT activation. In addition to Jak/STAT signaling, the p38 and PI3 kinase pathways have been implicated in IFN-α responses (20, 21, 22, 23, 24). Therefore, the role of these pathways in the induction of IP-10 and other classical ISGs in response to IFN-α1 and -α2 was investigated. DC were treated with IFN-α1 and -α2 (at 20 and 2 ng/ml, respectively) in the absence or presence of well-characterized inhibitors of the PI3 kinase (LY294002) and p38 (SB203580) pathways (Fig. 5,A). Transcriptional responses were assessed at 8 h, at which time point the differential between the IFN-α subtypes at these concentrations was most evident (Fig. 3,A). The induction of IP-10 in response to both IFN-α subtypes was strongly inhibited by SB203580, but not by LY294002, indicating that the p38, but not the PI3 kinase pathway is involved in regulation of this gene. SB203580 inhibited IP-10 mRNA induction in a dose-dependent fashion, inhibition of IP-10 being observed with as little as 1 μM of this compound (data not shown). In contrast, this inhibitor did not affect the induction of the ISG-56K, 2′-5′-oligoadenylate synthetase, and 6-16 mRNAs (Fig. 5, A and B), even at substantially higher concentrations (20 μM; data not shown). Although LY294002 was without effect on the induction of IP-10 mRNA, it was active in these cells. It strongly inhibited the accumulation of IP-10 protein in the medium of treated cells (ELISA; three independent experiments; data not shown). This most likely reflects the requirement for PI3 kinase for protein secretion through exocytosis of secretory vesicles (55).

In accordance with a requirement for p38 for optimum induction of IP-10, both IFN-α1 (at 20 ng/ml) and IFN-α2 (at 2 ng/ml) induced phosphorylation of p38 (Fig. 5 C). Phosphorylation of protein kinase B, a substrate of PI3 kinase, was not observed (data not shown). Clearly, the p38 pathway is used by both IFN-α subtypes and is required for the optimum induction of IP-10, but not the other ISGs analyzed.

In this study, we report, for the first time, an analysis of signaling through Jak/STAT and additional pathways and of gene expression in human T cells and DC in response to different IFN-α subtypes. Substantial variations were observed, including differentials in IFN-α-subtype and cell-type responses and ISG expression. Importantly, the chemokine IP-10 was differentially induced in DC by IFN-α1 vs IFN-α2 and IFN-α21. At physiological concentrations (2 ng/ml), IP-10 mRNA and protein were highly induced by IFN-α2 and -α21, but only poorly by IFN-α1 (Table I and Figs. 3 and 4). In contrast to the other ISGs analyzed, increasing the concentration of IFN-α1 to 20 ng/ml did not result in the induction of IP-10 levels comparable to those induced by 2 ng/ml IFN-α2.

IP-10 was originally identified as an IFN-γ-induced protein (56). In vivo studies using either neutralizing IP-10 Abs or IP-10-deficient mice indicate that this chemokine plays an important nonredundant role in the recruitment/trafficking of effector Th1 cells into tissue sites of inflammation (57, 58). It has been proposed that early expression in particular of IP-10 in vivo is crucial in establishing protective cell-mediated immune responses against certain pathogens, and that components of the innate immune system, rather than IFN-γ, are responsible (57). IFN-αβ are likely candidates for such components. Indeed, IFN-αβ-induced IP-10 production by DC in response to Mycobacterium tuberculosis was recently shown to be critical for the selective recruitment of activated T cells (59). Therefore, it is not unreasonable to assume that the impaired induction of IP-10 by IFN-α1 at physiological concentrations may have important consequences for immune regulation. It should be noted that IFN-α1 is a major player in the IFN-α response, because it generally is one of the main subtypes produced (49, 50, 51).

In addition to a protective role for IP-10 in infections that require strong cell-mediated immune responses, IP-10 may also contribute to progression of certain diseases. The expression of IP-10 has been observed in many Th1-type inflammatory/autoimmune diseases, including psoriasis (60), multiple sclerosis (61, 62), rheumatoid arthritis (63), and type 1 diabetes (64). Moreover, in vivo mouse models of Th1 type autoimmunity have established a critical role for IP-10 in disease onset and progression (58, 65, 66). Clearly, the induction of IP-10 and the subsequent recruitment of Th1 cells to inflammatory sites are not desirable in these diseases. In this respect, it is of interest to note that the therapeutical application of IFN-α is contraindicated in patients with a history of Th1-related disorders. Moreover, there have been examples from the clinic where administration of IFN-α has caused the induction of autoimmune diseases (67, 68) and has led to exacerbation of psoriatic lesions (69). Accepting the essential role of IP-10 in Th1-type inflammatory diseases, the observation that induction of this chemokine is impaired in response to IFN-α1 may be clinically relevant, because this subtype may be a more suitable candidate for therapeutical applications in cases where Th1 inflammatory responses are not desirable. Although this possibility remains to be established, it is tantalizing that IFN-α1 caused less side effects than IFN-α2 in patients with malignant diseases (70).

Two previous studies have applied global gene expression profiling tools to address the related question of whether there are functional differences between IFN-α2 and IFN-β (37, 71). These studies demonstrated that, at a given concentration, more genes were induced by IFN-β than -α2 (37, 71), but that the patterns of ISG induction became identical upon increasing the IFN-α2 concentration (37)—a result similar to that for the majority of ISGs in response to IFN-α2 vs IFN-α1 in this study (Table I). Accepting that the IFN-αβ act through a common receptor, this overlap in response is arguably not surprising. That said, the current data show that, even at the relatively high physiological concentration of 2 ng/ml, induction of IP-10 by IFN-α1 was severely impaired. Although induction of the IFN-αβ may be high in response to at least some virus infections (72), it may be relatively low in response to other pathogens (6), and there is also good evidence for low constitutive expression (73). Thus, differences in IFN-αβ actions at lower concentrations may also be biologically relevant.

Substantial quantitative differences between IFN-α1, -α2, and -α21 in the ability to activate STAT1 to -5 and a spectrum of ISGs were found. At high IFN-α concentrations, all subtypes induced phosphorylation of STAT1 to -5 in T cells and STAT1 to -3 in DC (STAT4 and -5 were not analyzed) and similar patterns of ISG expression (Fig. 1 and Table I). Differences between subtypes in this respect are quantitative, not qualitative, and similar to those reported for IFN-α2 and IFN-β (37). Importantly, no significant differences in the kinetics of STAT activation by the different subtypes over a range of concentrations were observed (Fig. 2 and data not presented). Therefore, there is no apparent specificity of STAT activation for these particular subtypes in these particular cells. Recently, Cull et al. (38) reported the selective activation of STAT5 in response to different IFN-α subtypes in an immortalized murine erythroblast cell line. However, it remains possible that this again reflects a quantitative concentration-dependent difference, particularly because activation of STAT1 and -3 by various subtypes was shown to be highly dependent on the dose used.

The less efficient phosphorylation/activation of STATs by IFN-α1 than -α2 (Fig. 1) correlates with the known lower binding affinity for the IFN-αβ receptor (20-fold lower K_d) (39) and lower (5- to 20-fold) specific antiviral activity, depending on cell type (52). However, induction of most ISGs was only 2- to 4-fold lower in response to IFN-α1 than to IFN-α2 and -α21 (all at 2 ng/ml). Increasing the IFN-α1 concentration to 20 ng/ml yielded ISG induction comparable to IFN-α2 and -α21 (with the exception of IP-10), despite lower STAT activation even at this higher IFN-α1 concentration (Table I and Fig. 1). Thus, the quantitative differential between IFN-α1 and IFN-α2 or -α21 appears smaller for the induction of ISGs than for the activation of STATs. This was not simply due to differences in the sensitivities of the assays used, because the relatively small differences in ISG induction were observed over a wide range of concentrations (for IFN-α1, 20 to 0.2 ng/ml; IFN-α2, 2 to 0.02 ng/ml), and the shapes of the dose-response curves for IFN-α1 and IFN-α2 were similar (data not shown). Differentials between STAT activation and functional responses have been reported previously; for example, differences in antiviral and antiproliferative activities of a number of IFN-α2/21 hybrids did not correlate with their ability to activate STAT1/2 (45), and the antiviral activity of IFN-β in a cell line expressing a truncated IFN-αβ receptor was strongly impaired in contrast to activation of Jak/STAT signaling (43). Accordingly additional rate-limiting factors are likely to be involved.

Both the p38 and PI3 kinase pathways have been implicated in IFN-α signaling (20, 21, 22, 23, 24). In DC, activation of p38, but not PI3 kinase, was differentially required for optimal induction of IP-10, vs other ISG mRNAs (Fig. 5). Both IFN-α1 and IFN-α2 activate p38 kinase(s) (Fig. 5 C). Accordingly, differential p38 induction seems unlikely to be the basis for the observed differential in IP-10 induction in response to the two IFN-α subtypes. Requirement for p38 has been reported for both ISRE- and GAS-driven expression of reporter constructs (22, 23, 24). The requirement for p38 activation for IFN-α to optimally induce IP-10 is, to the best of our knowledge, the first demonstration of the involvement of this pathway in the regulation of an endogenous ISG by IFN-α. Among the ISGs analyzed in this study, IP-10 was, therefore, unique in being 1) differentially regulated by IFN-α subtypes, 2) dependent on activation of p38, and 3) induced in DC but not T cells.

Although not a specific focus, this study has also provided further evidence for cell type specificity for the IFN-α response (see Ref.47) and major differences in the kinetics of expression of subsets of ISGs. In addition to IP-10, examples of cell type specificity include the induction of iNOS in DC but not T cells and IL-12Rβ2 in T cells but not DC (Table I and Fig. 3). Consistent with the latter, IL-12Rβ2 is a STAT4-responsive gene, and immature DC do not express STAT4 (74, 75, 76). Among the ISG mRNAs examined, very different kinetics of accumulation were observed (Fig. 3,B and data not presented). For example, SOCS1 showed very early transient kinetics, IP-10 and ISG-56K were similarly transient but slightly delayed, whereas 6-16, 9-27, and STAT1 showed a more sustained induction (Fig. 3). Clearly, the point of detection depends on the sensitivity of the assay and the time points chosen, but the differentials in the shape of the curves, which may or may not reflect major differences in mRNA stability, remain. Interestingly, such differentials were not observed in a similar analysis for IFN-α2 in HT1080 human fibrosarcoma cells. In these, the mRNA accumulation for all 150 ISGs analyzed was prolonged and optimal at 6–8 h (Ref.47 ; J. F. Schlaak and I. M. Kerr, unpublished data). These contrasting results emphasize both the complexity of the responses and their dependence on cell type. Clearly, in this area, data obtained in one cell system cannot necessarily be extrapolated to others.

In conclusion, the current study provides novel information on IFN-α subtype responses in human T cells and DC. Substantial variations in IFN-α subtype activity, in cell type specificity, and in requirements for the p38 MAPK pathway were observed. Induction of IP-10 was the most striking example in this respect. Further examples of variations in IFN-α-subtype activity will likely continue to prove important in the developing understanding of the role(s) of the different subtypes in the immune response.

We thank Drs. Doreen Cantrell, Patrick Costello, Sandra Diebold, and Facundo Batista for critically reading the manuscript.