Early detection and screening strategies for SCLC
Effective early detection of any cancer (i.e. detection that allows complete removal of lesions before they can metastasize) has the potential to greatly reduce cancer-related mortality. Ideally, this would be achieved by non-invasive molecular imaging combined with biomarker-based tools [
145]. An effective screening method must have high specificity and sensitivity, reduce overall mortality, be acceptable to patients, and be cost effective in order to be widely adopted [
146].
To date, imaging has shown limited promise for the early detection of SCLC. [
18F] fluorodeoxyglucose positron emission tomography (FDG-PET) scanning, a method that has been used to detect SCLC tumors in paraneoplastic patients positive for anti-Hu antibodies, has not been shown to decrease mortality [
147]. This suggests that FDG-PET is unable to detect SCLC lesions in time for curative resection. Low Dose Spiral Computed Tomography (LDSCT) is under evaluation as an early lung cancer detection modality. Studies from the Early Lung Cancer Action Project (ELCAP) suggest that annual spiral CT screening can detect lung cancer at a curable stage and increase lung cancer survival [
148‐
151]. However, any early detection method causes lead time bias; the favorable survival noted in the ELCAP study is therefore not a valid measure of the efficacy of screening [
152]. In addition, LDSCT was shown to incorrectly classify many non-calcified nodules as cancers [
153,
154], potentially subjecting patients to unnecessary follow-up procedures (scans, biopsies, or resections), which are costly, invasive, and can result in patient morbidity and mortality [
149,
155,
156]. Recently, it has been hinted that the National Lung Cancer Screening Trial, a randomized control trial of high risk long term smokers subjected to either LDSCT or X-ray based screening [
157], might show a reduction in mortality in the LDSCT-screened group. The published results of the study are therefore anxiously awaited so that the potential benefits and drawbacks of screening by LDSCT can be weighed and the guidelines for the use of this tool can be established. Whether this type of screening could benefit SCLC patients is unclear. Early detection of SCLC through imaging is especially challenging because SCLC can metastasize when the primary lesion is still very small. Small tumors are more challenging to detect because they generate less signal. In addition, if SCLCs are small at metastasis, imaging-based screening may allow too little time between detection and metastasis for successful intervention. Thus, it is unlikely that imaging alone will be of use for early detection of SCLC. Some kind of molecular marker will likely be required. However, if a molecular assay depended on nucleic acids or protein molecules derived from the tumor, the same problem of tumor size would apply, since these molecules may be secreted into the blood in undetectable amounts when the tumor is small. This caveat may not pertain to an antibody response; it is unknown what size a SCLC tumor must be to trigger immune reactivity.
Autoantibodies against tumor-associated antigens are more stable and specific than other serum-derived proteins [
158], making them good candidate biomarkers for cancer detection. Because the natural history of SCLC is unknown, it is possible that small tumors or precancerous lesions may be present in patients for a long time before the tumors develop metastatic abilities. Many patients have smoked for decades before they develop SCLC. Depending on the proteins they express, preneoplastic lesions might trigger an immune response. In SCLC patients with paraneoplastic disease, the immune response is frequently noted before the tumor is detected [
75]. It is, however, unclear if the timing of events is similar in SCLC patients
without PNS. In our studies of the immune response in SCLC-prone mice, we observed one mouse with high titer antibodies and
in situ neuroendocrine lesions, but no invasive cancer [
48]. This, and other observations from the SCLC mouse model, support the notion that an immune response can precede clinical detection (Figure
2). However, the window of detection in the mouse model is small (several months at best). It remains to be seen how the timing of events in this genetically engineered mouse model system translates to the human situation in which (epi)genetic alterations, potentially occurring over the course of many years, result in the development of SCLC. Despite this drawback, at this time, the SCLC mouse model system offers the best tool to gain insight into the temporal relationship between SCLC development and the immune response. Studies of a larger number of mice, including analysis of their lungs as elevated titers of antibodies develop, will hopefully shed light on the type of lesions that are present when the autoimmune response is triggered.
Identifying biomarkers highly
specific for a malignant condition is one great challenge in developing a serological detection tool. In the case of autoantibodies, multiple environmental factors, pathogen invasion, and autoimmune disease can result in the production of a high-level of IgG and IgM autoantibodies that recognize various antigens and thus reduce the specificity of the antibody [
159‐
161]. The prevalence of smoking in healthy individuals may also affect their immune response (reviewed in [
162]). In addition, the detection of background reactivity may be influenced by technology; the more sensitive the technique, the greater the probability that background reactivity will be detected. A recent study analyzed a panel of tumor-associated autoantibodies in 205 healthy individuals and found a subgroup of individuals that showed elevated levels of immunoreactivity against most of the antigens tested [
163]. Unfortunately, no SCLC-associated antigens were examined in this study. In another analysis, the presence of anti-Hu reactivity was examined in plasma from 120 subjects, including 79 healthy controls (smokers and non-smokers) and 41 SCLC cases [
52]. The latter was a population-based study matched on age, race, sex, and smoking status in which anti-Hu reactivity was examined in relation to lung cancer risk. Although anti-Hu reactivity was found to be significantly higher in cases than controls, low level anti-Hu reactivity was detected in healthy non-smokers (~39%) as well as smokers (~44%). Dalmau and colleagues also observed anti-Hu reactivity in non-cancer individuals; they considered reactivity in normal subjects as background and used it as cut-off to score anti-Hu reactivity in SCLC patients with and without PEM/SN [
26]. In other studies, reactivity of anti-VGCC autoantibodies in normal controls was used as a cut-off for positive reactivity [
66,
164]. The presence of background reactivity in non-cancer patients will likely limit the specificity of any anti-Hu antibody-based SCLC test, and will require the stringent determination of cut-off values for reactivity considered positive.
Background reactivity does not appear to occur for all antigens. In a study examining anti-recoverin response, no healthy individuals showed positive reactivity [
87]. However, SCLC is not the only cancer associated with anti-recoverin antibodies, which can also be found in several other cancers (Figure
1). No anti-SOX1 antibodies were found in healthy individuals in another study testing the sera of 27 healthy blood donors [
165]. These examples illustrate that careful choices must be made when considering which SCLC antigens might be useful for early detection tests.
The second big challenge in developing a serological detection tool is to attain sufficient
sensitivity. It is clear that a detection test utilizing any single SCLC antigen will show insufficient sensitivity because any single antibody is present in only a fraction of SCLC patients (Table
1). However, examination of autoantibody reactivity in SCLC patients shows that different patients can exhibit distinct immune responses and in some occasions, an immune response against multiple antigens [
29,
31,
166,
167]. Thus, a
panel of SCLC-related autoantigens may provide the necessary increased sensitivity, as was seen, for example, by Titulaer and colleagues when examining four SOX family members and the three neuronal Hu antigens in SCLC patients with and without LEMS [
31]. The authors from this study find 67% sensitivity and 95% specificity in discriminating between SCLC patients with LEMS and LEMS patients without tumors when testing against a panel of the SOX family members (Sox 1, 2, 3, and/or 21) [
31]. This latter study underlines the need to continue searching for new SCLC-associated autoantigens.
Will a combination of antigens ever attain the required sensitivity? To try to answer this question, we can examine two extreme hypothetical situations. In one, all patients show non-overlapping immune responses against different autoantigens. In this case, a panel may greatly increase sensitivity; the sum of the percentages of all SCLC patients immuno-positive for currently known antigens would allow detection of SCLC in 100% of patients (see Table
1). In the opposite scenario, immunoreactivity is only ever present in the same small subset of SCLC patients that could exhibit a wide variety of antibodies. In this case, the sensitivity one could achieve using a panel of antigens might be as small as 25%. The latter scenario is unlikely. For example, no correlation between the presence of anti-Hu and anti-VGCC autoantibodies was found in 200 newly diagnosed lung cancer patients. Although four patients were positive for both autoantibodies, individually, up to 25.5% and 5% were immuno-reactive against Hu proteins and VGCC, respectively [
50]. Thus, the combination of carefully selected antigens may ultimately provide a marker panel that could be of use for early SCLC detection. Indeed, recent studies show that sensitivity and specificity is greatly enhanced when a combination of antibodies is used to detect lung cancer [
31,
168,
169]. Aside from careful choices of markers, defining an at-risk group that could be screened for early disease would be crucial to increase the feasibility of any test. In the case of SCLC, longtime heavy smokers would be prime candidates for screening. When examining anti-Hu reactivity in smoking and non-smoking healthy controls [
52], one of the smoking controls with detectable anti-Hu reactivity and healthy at the time of the study died of SCLC several years after the conclusion of the study. Based on the expected frequency of lung cancer in this group and the relatively minor contribution of SCLC, the SCLC case is remarkable. However, such incidental findings are impossible to interpret. In addition, three highly anti-Hu positive cases did not develop SCLC within the time frame of the study. Interestingly, one of the anti-Hu positive cases died of prostate cancer, which can occasionally exhibit anti-Hu antibodies (Figure
1A) and can show a neuroendocrine "small cell" type [
89].
It is clear that questions of sensitivity and specificity of a combined panel of SCLC-associated autoantigens for detection of SCLC will require further study. A retrospective study using plasma/serum collected from a lung cancer cohort, such as from the Carotene and Retinol Efficacy Trial (CARET) study [
170] and the Prostate, Lung, Colorectal, Ovarian (PLCO) Cancer Screening Trial [
171], could provide insight into the potential of such antibodies for early cancer detection. However, a large prospective study would ultimately be required. Before any such study would be possible, stronger data justifying the feasibility of autoantibody markers for early SCLC detection would be needed. To this end, further analysis of the SCLC-prone mouse model system [
48,
131] could provide important data on the potential sensitivity and specificity of antigen panels.
While early detection of SCLC would be the holy grail, a much simpler application of SCLC-associated autoantibodies is the establishment of a diagnosis. If the timing of autoantibody generation does not precede cancer development, or does not precede it by sufficient time, these antibodies would still be useful tools to establish or confirm a SCLC diagnosis. In PNS patients, detected autoantibodies often trigger a tumor search. One could envision that an antigen panel to confirm suspected SCLC could be established. Retrospective studies of archival human serum and plasma aimed at exploring SCLC-associated antibodies for early detection will provide data that would also be very useful to determine utility for diagnosis. Ultimately, however, a prospective study in which serum and plasma from suspected SCLC are tested and correlated with the confirmed clinical diagnosis, would be required.
Survival and prognostic benefits of a SCLC-associated immune response
How could SCLC-associated autoantibodies affect survival? For one, in patients exhibiting neurological symptoms, a diagnosis of SCLC-related autoimmunity will prompt a tumor search, which would allow the cancer to be detected at an earlier time point when intervention might be more effective. The fact that autoantibodies can precede the diagnosis of the tumor [
172,
173] suggests the cancer can still be very small, and it opens a window for improving the overall outcome in these patients, perhaps by beginning antitumor treatment before metastasis occurs.
Secondly, the immune response may itself be protective. A number of studies suggest that patients with SCLC-related autoimmune disease have a better prognosis than patients with histologically identical tumors without a paraneoplastic disorder [
139,
174‐
176]. The fact that some PEM/SN patients have small or virtually undetectable SCLC lesions that are only identifiable at autopsy supports this observation [
42]. Furthermore, it has been reported that in SCLC patients
without paraneoplastic disease, low titers of anti-Hu autoantibodies are correlated with comparatively indolent tumor growth relative to antibody-negative patients [
42,
88]. The presence of other SCLC-associated autoantibodies (VGCCs, CV2/CRMP5) in patients has been correlated to slower tumor growth, complete response to therapy, and longer survival in some studies [
32,
41,
42,
52,
88,
137‐
142]. Other studies, however, have found no survival benefit in correlation with SCLC-associated autoantibodies [
28,
31,
50,
88,
143,
144]. This discrepancy may be due to sample size, treatment protocols, secondary endpoints used to measure survival, the nature of the antibodies, the presence of clinical features, and whether or not the clinical status of the patient is known at the close of the study.
Abnormal or ectopic expression of antigens by a non-neuronal cell type could activate immuno-competent cells that promote an inflammatory response. The inflammatory response can trigger apoptosis, which could then initiate an antitumor immune response [
22,
108]. Apoptosis of tumor cells can expose intracellular antigens, mediating an immune response to self-antigens that are expressed both in the tumor and the nervous system [
177‐
179]. Furthermore, tumor cells can be phagocytosed by dendritic cells that further activate antigen-specific CD4+ and CD8+ T cells and B cells through normal immunologic processes. These antibody-producing B cells could act in concert along with cytotoxic T cells specific for the tumor antigen and induce cytotoxicity, potentially retarding tumor growth. The antibodies could trigger complement-mediated cytotoxicity [
180] or antibody-dependent cell-mediated cytotoxicity [
181]. Furthermore, the antibodies may bind to their target and interfere with the function of the antigen. This might disrupt tumor progression if the antigen is necessary for tumor growth. The latter may provide an explanation for how tumors in some antibody-positive SCLC patients are smaller or slower growing. Definitive evidence of this theory is lacking, however. The nature of the immuno-response in SCLC requires further characterization to determine the protective nature of antibodies and T cells. Again, the SCLC-prone mouse model discussed in Section 5 would be a useful resource to study this phenomenon.
SCLC autoantigens as targets for anti-cancer therapy and imaging
The development of SCLC vaccines presents a formidable challenge because an immune response against antigens expressed both in the tumor and nervous system may cause neurotoxicity. The main objective in treating a patient with SCLC-related PNS is to cure the underlying cancer and to improve or stabilize the neurological dysfunction [
140]. Interestingly, in mice that were implanted with neuroblastoma, immunization with DNA encoding HuD has been shown to retard tumor growth [
110,
111], and the mice did not develop neurologic abnormalities in this study nor in a study where mice were immunized with recombinant purified HuD antigen [
111]. The SCLC-prone mouse model [
131] can be utilized in a similar manner to determine if immunization with various SCLC-associated autoantigens would protect against the development of the SCLC tumor; however, neurotoxicity is a potential side effect and would need to be closely monitored.
A very different situation arises if the antigens triggering the immune responses are mutated or post-translationally modified in the tumor compared to the native nervous system. In that case, a strategy to specifically target the modification would be an option. However, when using immunizations aimed at inducing a cancer-specific response to a modified protein, a unique tumor-specific epitope may not preclude a destructive autoimmune response; reactivity might expand through epitope spreading. For any immunotherapy strategy, finding the right balance between attacking the tumor and avoiding neurotoxicity is imperative. A more promising avenue may be to develop molecules tailored to recognize the antigen using
in vitro evolution in small scaffolds, like knottin peptides [
182]. Such "magic bullets" could be targeted toward any cancer-specific modification to deliver molecules for imaging or treatment. Visualization of small tumors is critical to establish the precise location to perform successful resection. Most current imaging modalities are not specific for cancer, although metabolism-based markers show higher signals in cancer cells based on their increased metabolic activity. Imaging agents specifically targeted to cancer cells can be powerful aids in cancer detection and could increase our ability to detect small SCLC lesions.
The potential to use the immune system or targeted molecules to attack the tumor underlines the importance of obtaining a detailed mechanistic understanding of SCLC-associated autoimmunity. Until the nature of the antigen and the immune response are clarified, it will be very difficult to devise fruitful applications. The biggest challenge may be that different SCLC-associated immune responses may occur through distinct mechanisms.