Demographics and response to treatment
Seventeen white males were randomized to receive treatment (13 to imiquimod and 4 to placebo). The mean age was 75 years (range, 62 to 89 years). All 17 subjects in the study completed the treatment and post treatment portions of the study. The median number of AK lesions at baseline was 10 per subject (range, 6 to 13 lesions). Because of the short duration of the follow-up period (4 weeks for the post treatment period), efficacy was not measured in this study. However, clinical clearance was observed in 25% of the imiquimod-treated subjects 4 weeks after the end of treatment. Imiquimod-treated subjects 1, 2, 4, and 6 and placebo-treated subject 7 were assessed as having clinical clearance of lesions 4 weeks post treatment as determined by return of the lesional site to normal skin. The complete clearance rate in a study where AK subjects were treated for 16 weeks, with a post treatment period of 8 weeks, was 57% .
Analysis of global gene expression using Affymetix GeneChips: Gene Ontology classification
A 2-way analysis of variance (ANOVA) was performed comparing the treatment response gene expression fold change (the maximum response value from week 1, week 2 and week 4 treatments) of samples from the imiquimod-treated subjects to the gene expression fold change values of their respective pretreatment AK samples. This comparison resulted in 530 unique genes that had p-values 2 for induced genes. Data for the 530 genes is documented in [Additional file 1]. Two-way ANOVA comparing pretreatment AK samples and samples taken 4 weeks after the last imiquimod treatment resulted in 111 unique genes that differentiated the 2 groups with a p-value 1]. The expression of the rest of the genes returned to basal levels 4 weeks post treatment. Of the total number of differentially regulated genes during imiquimod treatment or 4 weeks post treatment, 87% were up-regulated and 13% were down-regulated.
Table 1 summarizes the gene ontology classification of imiquimod-modulated genes from the Affymetrix analysis. P-values for representation of ontology categories are given for categories with p-values 0.05 for representation in the gene ontology category included metabolism (148 genes), development (37 genes) and regulation of transcription (24 genes). The ontology analysis shows that imiquimod treatment of AK results in global gene expression changes impacting various cellular processes, with immune response and signal transduction being the 2 major processes represented.
Validation of selected genes observed in the Affymetrix experiment using real time reverse transcriptase polymerase chain reaction
Micro array analysis may not be sensitive enough to capture all changes in gene expression. In comparison, real time reverse transcriptase polymerase chain reaction (RT-PCR) has been shown to be more sensitive for the detection of low abundance transcripts . Therefore, we analyzed AK samples and imiquimod treated AK samples for a selected set of genes using real time RT-PCR. The data is summarized in [Additional File 2] which compares the median fold change values (13 imiquimod-treated subjects) measured using both Affymetrix GeneChip analysis and RT-PCR for a subset of the genes. In general, good agreement was obtained between the 2 methods, with the RT-PCR data showing similar or higher expression than the Affymetrix data. Figure 2 shows a comparison of the 2 methods for the expression of interferon regulatory factor 7 (IRF7). The individual fold change values for the 13 imiquimod-treated subjects during imiquimod treatment were used for the regression analysis (R Square = 0.83). In general, the direction of change in expression for each gene was the same in both assays whereas the magnitude of the fold change values was higher for the RT-PCR analysis than the Affymetrix analysis. Of the 46 genes evaluated on both gene expression platforms, only 3 genes (CD80, MyD88 and TLR6) were identified as having changed in expression in the RT-PCR experiments only (p value ANOVA analysis
Imiquimod increases expression of pattern-recognition receptors of the innate immune system
Since imiquimod is a TLR7 agonist, we sought to determine if TLR7 and other TLRs were expressed in AK lesions and if treatment with imiquimod altered their expression. We analyzed expression levels of various TLRs in biopsy samples of pretreatment AK and after treatment with imiquimod using RT-PCR. We also measured the expression levels of MyD88, an adaptor molecule for various TLRs including TLR7; and IRF7, a transcription factor recently shown to be important in the regulation of interferon gene expression through the TLR7 pathway . Figure 3 shows the median expression levels of the various TLRs, MyD88 and IRF7 for subjects treated with imiquimod (n = 13) during treatment and 4 weeks post treatment. The expression of TLR1, TLR3, TLR6, TLR7, TLR8, TLR9, MyD88 and IRF7 were all increased at statistically significant levels (p-value 2]. The increase in TLR4 expression did not reach a statistically significant level (p-value = 0.066) in the imiquimod-treated samples, whereas the Affymetrix data shows a statistically significant increase (p-value of 0.028). The most statistically significant changes observed upon treatment with imiquimod were for TLR3, TLR7, TLR8, and IRF7, with good agreement between the Affymetrix and RT-PCR analysis both for the magnitude of change and for the p-values. There was no change in the expression levels of TLR5 and TLR10 upon treatment with imiquimod. The data are consistent with previous reports of induction of TLRs by various TLR agonists and IFNα [35-37][38,39]. The increased expression of the TLRs may be a result of increased expression of the genes in cells resident in the skin (e.g., DCs, macrophages) or due to the influx of cells with high expression of these genes (e.g., DCs, macrophages, plasmacytoid DCs). The increased expression of TLR3, TLR7, and TLR8 is consistent with increased expression observed in human peripheral blood mononuclear cells upon treatment with imiquimod. The data are also consistent with increased expression of TLR7 observed in imiquimod-treated AK . In summary, treatment of AK lesions with imiquimod results in increased expression of several TLRs and TLR pathway components, thus potentially priming for further amplification of the innate immune system. The increase in the expression of IRF7 is also predicted to amplify the innate immune response by increased expression of type 1 interferons and interferon-inducible genes.
In addition to the various TLR(s) which recognize viral and bacterial components, an intracellular antiviral pathway which detects viral RNA and results in the induction of type1 interferons has recently been described [40-42]. The central components of this pathway, which are also inducible by type1 interferons [43,44] are DDX58 (RIG-I, retinoid acid inducible gene) and IFIH1 (MDA5, melanoma differentiation antigen 5). Both genes contain helicase domains responsible for detection of double-stranded RNA. They also contain caspase recruitment domains (CARDdomains), which are responsible for signaling through TBK1 (resulting in activation of NFKB and IRF3), as well as induction of type-1 interferons. The activation of this pathway leads to growth inhibition as well as antiviral activity . Figure 4a and 4b show changes in expression of DDX58 and IFIH1 upon treatment with imiquimod as determined from Affymetrix GeneChip analysis. The expression of both genes was increased to statistically significant levels during treatment. The expression of DDX58 remained higher than its pretreatment level (p-value = 0.004), four weeks, post treatment, whereas that of IFIH1 returned to basal level. The induction of several members of the cytoplasmic helicase innate immune pathway, as well as several TLRs, indicates that in addition to activation of the TLR7 pathway, treatment with imiquimod also results in priming of other innate pathways. These pathways may augment other aspects of the innate immune response and may be important for eliminating pre-neoplastic cells.
Imiquimod induces a large number of type I interferon-inducible genes with growth inhibitory and immune-stimulatory activity
Imidazoquinoline TLR7 agonists such as imiquimod and resiquimod are know to induce various cytokines, including interferon-α, IL6, MCP-1 and IL12 as well as the co-stimulatory markers CD80 and CD86 [15,46]. Type I interferons are known to be powerful regulators of the innate and adaptive immune system through the induction of various genes with antiviral, anti-tumor and immune regulatory functions [43,44,47-51]. In this study, we did not detect increased expression of interferon upon treatment with imiquimod, but observed the increased expression of a large number of IFN-inducible genes (114 genes), [see Additional file 3]. The lack of detection of type 1 interferons after imiquimod treatment in this study may be due to the early induction and degradation of their respective mRNA. Biopsies were taken approximately 8 to 16 hr after application of the drug. We have observed that mRNA for type 1 interferons in human blood mononuclear cells treated with imiquimod peaks in 1 to 2 hours and declines to basal levels 6 to 8 hours post treatment (unpublished internal data).
Analysis of gene ontology classification of the interferon-inducible genes increased by treatment with imiquimod identified 46 genes with immune response classification. Figure 5 shows a 2-way hierarchical clustering of the log2 transformed fold changes of all of the 530 imiquimod-induced genes [see Additional file 1]. Figure 6 exhibits clustering of the 46 interferon-inducible genes which classify as immune response genes Two main clusters are apparent in Figure 5. One cluster consists of 8 imiquimod-treated samples. The second and larger cluster consists of all of the pretreatment AK samples, the vehicle-treated samples (designated as placebo in the figure) and five of the imiquimod-treated samples for subjects 02, 05, 09, 12 and 17 (IMIQ-02, IMIQ-05, IMIQ-09, IMIQ-12 and IMIQ-17). The fact that the vehicle-treated samples cluster with pretreatment AK lesions indicates the lack of a significant vehicle-effect in the gene expression profile. In Figure 6, 10 of the imiquimod-treated samples appear in 1 cluster, with IMIQ-02 and IMIQ-05 now as part of the IMIQ-treated cluster. This cluster is characterized by high expression of the interferon-inducible genes such as MX1, IFIT1 and IFIT3. The cluster also contains the samples placebo-03 and AK-11, indicating that these AK lesions were already manifesting spontaneous immune response. The imiquimod-treated samples 09, 12, and 17 are clustered with the pretreatment AK samples in both figures, indicating that Subjects 09, 12, and 17 had little gene expression response to imiquimod. Thus, the 46 interferon-inducible immune response genes segregate the imiquimod-treated samples from pretreatment AK and placebo-treated samples better than the whole set of 530 genes. This group of genes may therefore be good predictors of gene expression response to imiquimod. Collectively, these genes have been reported as 'interferon-signature' genes induced by several IFNα subtypes and IFNβ in monocytes .
Heterogeneity of pretreatment AK lesions was not documented by clinical assessment in this study. However, the gene expression profiles shown in Figure 5 and Figure 6 indicate heterogeneity in lesions. In addition to AK 11 and Placebo 03 in Figure 6 which cluster with the imiquimod treatment group indicating some level of immune response in these samples, other AK lesions also show low levels of expression of interferon inducible genes. For example, AK 06, AK 03, and AK 07 show low levels of expression of several interferon-inducible genes whereas AK 16, AK 17, AK 12 and AK 15 show normal to slightly depressed levels (Figure 6).
Interferon-inducible genes regulate diverse cellular processes, such as cell growth and differentiation, cell death, and T-cell co stimulation, activation, and migration. These genes have been reported to possess antiviral [52-54], pro-apoptotic [55-57], and anti-proliferative activities [58,59]. The interferon-inducible genes which increased following imiquimod treatment include those known to be induced by viruses as well as those with known anti-viral activity. These include the 2'5'-oligoadenylate synthetases OAS1, OAS2, OAS3 and OASL, the genes encoding the interferon-inducible proteins with tetratricopepetide repeats, IFIT1, IFIT2, and IFIT3, IFTM1, and other interferon inducible genes such as IFI35, IFI16, MX1, MX2, EIF2AK2 (PRKR), G1P2 (ISG15), G1P3, ISG20, RSAD2 (Cig5), CCL8 (MCP2), CXCL10 (IP10) and CXCL11 (ITAC) [43,44,47,49-52,54,60,61]. Several of the interferon-inducible genes were also increased at statistically significant levels 4 weeks post treatment. These included: IRF7, IFI44, IFIT2, IFIT3, IFITM1, IFI35, RSAD2 (Cig5), G1P2, MX1, OAS1, and OAS2 [see Additional file 1].
In addition to interferon-inducible genes with antiviral activity, several genes known or predicted to possess growth-inhibition and/or cell-differentiating activities [49,50] were induced by imiquimod, including IFI16, AIM2 [62,63], IFIH1 (melanoma differentiation antigen, MDA5) , CXCL10 (IP10)  and EIF2AK2 (PRKR) . Some of the interferon-inducible genes also possess pro-apoptotic activity, including MX1, TNFSF10 (TRAIL), OAS1, and PRF1 [55,57,65]. Figure 7a and Figure 7b show changes in the expression of the pro-apoptotic genes TNFSF10 (TRAIL) and MX1 with imiquimod treatment as determined from the Affymetrix analysis. The expression of MX1 remained elevated at statistically significant level 4 weeks post treatment, whereas the expression of TNFSF10 returned to basal levels. The data are consistent with the observation of increased expression of IFNα-inducible genes in imiquimod-treated BCC and cutaneous T-cell lymphoma (CTCL) [66,67]. Thus, interferon-inducible genes with pro-apoptotic activity (e.g., MX1, TNFSF10) and growth inhibitory activity (e.g., IFIH1, AIM2, IFI16) may result in growth inhibition of neoplastic cells whereas those with immune-stimulatory activity such as CXCL10 (IP10), CXCL11 (ITAC), and CCL8 (MCP2) may facilitate cell-mediated lesion destruction by recruiting immune cells into the lesions. Indeed, further evidence for the recruitment of immune cells into AK lesions with imiquimod treatment is presented in the following sections.
Imiquimod induces the expression of chemokines responsible for recruitment of immune cells to AK lesions
Gene ontology classification (Table 1) shows that 106 genes classified as immune responsive were induced by imiquimod, some of which are also shown to be interferon α/β-inducible [see Additional File 3]. Out of the 106 immune response genes, several were classified as cytokines/chemokines and chemokine receptors. In addition to CXCL10, CXCL11 and CCL8 mentioned above, other chemokines were induced by imiquimod treatment, including CCL3 (MIP1a), CCL4 (MIP1b), CCL5 (Rantes), CXCL12 (SDF1), and CXCL16, and the chemokine receptors CCR1, CCR5, and CXCR4. Figure 8a and 8b illustrate changes in expression of CXCL10 and CXCL11 observed upon treatment of AK lesions with imiquimod. The magnitude of the changes observed in the expression of chemokine genes during treatment ranged from a median fold change value of 1.8 for CXCL12 (SDF1) to 40.8 for CXCL11. The expression of both genes returned to pretreatment levels 4 weeks post treatment. These data are consistent with previous reports of increased expression of chemokines genes and their receptors upon topical treatment of BCC and CTCL with imiquimod , as well as in vitro studies of human blood mononuclear cells stimulated with other imidazoquinoline TLR7 agonists showing the induction of CXCL10 and CXCL11 proteins . The cocktail of chemokines up-regulated by imiquimod is consistent with recruitment and/or activation of macrophages, DCs, plasmacytoid DCs, gamma/delta T cells, cytotoxic T cells, and natural killer (NK) cells, and is also consistent with the cell surface markers indicating the presence of these cells upon imiquimod treatment. Indeed, topical treatment of various neoplastic skin conditions with imiquimod cream have shown inflammatory conditions at the site of treatment, indicating the infiltration of immune cells into the site [17,19,20,68,69]. In this study, the gene expression fingerprints that indicate recruitment of various immune cells are further discussed below.
Imiquimod increases the expression of genes predictive of infiltrating macrophage and dendritic cells
Of the imiquimod-induced genes classified as immune response in gene ontology, several have receptor activity and (or) are hematopoietic cell surface markers (Table 1). The increased expression of these genes indicates the recruitment of various immune cell types to the lesion sites. Macrophage and/or monocyte infiltration of AK lesions upon treatment with imiquimod was indicated by an increase in CD14, CD163, and CLECSF9 (CLECSF4, MINCLE), a C-type lectin found on activated macrophages . The increase in expression of genes of the classical complement pathway, C1QA, C1QB, C3AR1, and C5R1, also indicates an increase in and/or the activation of macrophages . The data are consistent with histologic observation of macrophage and/or monocyte infiltration after application ofimiquimod in the treatment of AK  and in lentigo maligna [73,74].
The presence of DCs is shown by increases in the co stimulatory molecules CD86 and CLEC4A (DCIR, CLECSF6), as well as by 3 leukocyte immunoglobulin receptors: LILRB3 (ILT3) and LILRB1 (ILT2) which are expressed in both myeloid and plasmacytoid DCs , while ILT7 (LILRA4, CD85g) is restricted to plasmacytoid DCs [76,77]. Figure 9a and Figure 9b illustrate the increase in expression of CD86 and ILT7 with imiquimod treatment. The expression of CD86 remained elevated 4 weeks post treatment. Increased expression is taken as indication of recruitment of these cells to the site of treatment. These observations are consistent with previous studies using topically-applied imiquimod in the treatment of human BCC  and melanoma in mice , showing recruitment of plasmacytoid DCs into the site of treatment.
It is interesting to note that CD1C is markedly decreased upon treatment with imiquimod [see Additional file 3]. CD1c is found on Langerhans cells as well as on DCs [79,80]. The decrease in CD1C expression may reflect (although not exclusively) the migration of CD1C+ Langerhans cells out of the dermis. Migration of Langerhans cells out of mouse dermis was observed after topical treatment with imiquimod [78,81]. In the case of the Palamara studies  in mouse melanoma, Langerhans cells were observed to return to normal levels in the dermis by day 20. The decrease in CD1C observed upon treatment of AK lesions with imiquimod therefore suggests the activation and migration of Langerhans cells to the lymph nodes and is consistent with previous observations.
Imiquimod increases the expression of genes predictive of infiltrating cytotoxic T cells and natural killer cells
Natural killer cells mediate lysis of tumor cells as well as virally-infected cells. Natural killer cells preferentially express several calcium-dependent (C-type) lectins, known as the NKG2 family, which have been implicated in the regulation of NK cell function, and are believed to be important for NK cell-mediated tumor rejection and T-cell mediated immunity . Transcripts for 3 of these C-type lectins, KLRC1/C2, KLRK1 (NKG2D) and KLRF1 were increased in expression in imiquimod-treated samples. In addition, several genes important to the cytolytic function of NK cells and cytotoxic T cells, including TYROBP (DAP12) and CD69 (early T-cell activation antigen), ITGAL (CD11a), ITGB2 (CD18), ICAM1, and the ligands for the NKG2 receptors, MICA and MICB [Additional file 1] [83-88] were increased in expression with imiquimod treatment. Also, genes of the granule products of NK cells and cytotoxic T cells, such as granzymes GZMA, GZMB, GZMK, and GYNYL (granulysin, NKG5), PRF (perforin) and NKG7 (GIG1, GMP-17) were increased in expression, indicating that these cells are cytolytically active [89-93]. Figure 10a and 10b show changes in the expression of GZMA and NKG7 after imiquimod treatment. The expression of both genes returned to pretreatment levels 4 weeks post treatment. Thus, the coordinate increased expression of genes important for NK cell recognition of tumor cells, and for the activation and cytolytic response of NK cells and cytotoxic T cells, suggests that these cells are at least in part responsible for the antilesional activity of imiquimod.
Imiquimod increases the expression of genes predictive of the activation of the adaptive immune system
The presence of activated T cells in imiquimod-treated AK lesions is indicated by the expression of several members of the T-cell activation pathway [Additional File 1]). These include the T-cell receptor (TCR) subunits TRD and TRG, TCR-signaling pathway genes such as Fyn, Fyb and LCP2 [94,95], and other genes associated with T-cell activation such as HCK, CD69, PTPRC (CD45) SELL (CD62L, L-Selectin), ITGA4 (antigen CD49D, alpha 4 subunit of VLA-4 receptor), and LAG3. In addition, treatment of AK lesions with imiquimod resulted in a small but significant increase in the expression of CD8β, suggesting that there is an increase in CD8 T cells [96,97]. The presence of CD8 memory T cells in imiquimod-treated subjects is also suggested by increases in expression of SELL, NT5E (CD73), LGALS2 (galectin 2), and LAIR1 (leukocyte-associated immunoglobulin-like receptor 1), which was recently identified to be differentially expressed in mouse memory Tcells . Figure 11a and Figure 11b illustrate the increase in expression of SELL and NT5E respectively. The expression levels of these genes returned to pretreatment levels 4 weeks post treatment. Taken together, the increased expression of Tcell receptor genes, T cell activation marker genes and genes important for T cell co stimulation suggests that treatment with imiquimod results in the infiltration of T cells into AK lesions. The increased expression of genes important for the development of T-cell memory (e.g. SELL, NT5E), as well as genes associated with cytotoxic T cells (e.g. granzymes and perforin) suggests that imiquimod treatment recruits both memory and effector T cells. These results are consistent with the observation of increased CD3, CD4 and CD8 positive cells after topical application of imiquimod to AK lesions , as well as infiltration of CD8 T cells in BCC [19,66] and in cutaneous squamous cell carcinoma treated with topical imiquimod . The increased expression of genes important in the activation and co stimulation of cells of the adaptive immune system is consistent with the infiltration of cells important in the development of the adaptive immune system. These observations are also consistent with the reported adjuvant activity of imiquimod and its analog resiquimod , and the low recurrence rates observed in animal studies  and clinical studies with imiquimod . These same observations suggest that imiquimod may also prevent recurrence of AK lesions as well.
In summary, the data suggest that the therapeutic effect of imiquimod in the treatment of AK involves the stimulation of both the innate and adaptive immune responses. The role of type 1 interferons in imiquimod's mechanism of action is underscored by the induction of large numbers of IFNα/β-inducible genes with tumor growth-inhibitory and immune stimulatory activity. Data indicate that the antilesional activity of imiquimod is a result of the induction of a strong immune cell-mediated cytolytic and apoptotic gene expression program that leads to destruction of AK lesions and sun-damaged cells. The development of immune memory indicated by this study, as well as the observation of low recurrence in clinical studies, makes imiquimod a unique therapy for AK as well as for other cutaneous neoplasms.