Introduction
The processes that characterize and trigger progression of preinvasive ductal carcinoma
in situ (DCIS) to invasive breast cancer remain elusive. DCIS is a heterogeneous disease in which neoplastic cells are confined by the basement membrane of lobuloductal or ductal lumen. Progression to invasive ductal carcinoma (IDC) is thought to occur by first degradation of the basement membrane, microinvasion of cancer cells into the surrounding stroma and growth of a solid tumor. The use of screening mammography has increased rates of detection of DCIS [
1], which has in turn expanded knowledge about the biology of these earliest stage breast cancers. However, clinical imaging provides only a snapshot of tumor biology. Basic characteristics of DCIS development over time (i.e., growth rates and changes in morphology) and progression to IDC are still largely unknown [
2].
Fundamental questions about the natural history of DCIS have remained unanswered largely because they are difficult to study in women. Due to obligate surgical excision of newly diagnosed cancers, subsequent lesion progression cannot be followed. A few studies have examined the outcome in a small number of women whose DCIS was initially misdiagnosed as benign disease, that is, treated by biopsy alone [
3,
4]. In one such study, 6 of 13 cases of DCIS progressed to invasive breast cancer in an average of nine years. In another, 11 of 28 women with misdiagnosed low-grade DCIS developed invasive carcinoma in the same quadrant, the majority within 15 years. These studies and others [
5] have prompted some to suggest that DCIS may be over-diagnosed and over-treated [
6‐
8] because not all will progress to invasive cancer. If this is the case, it is clinically important to identify predictive markers that can distinguish those DCIS that will remain indolent from those that will progress to life-threatening disease. Some studies suggest that a higher nuclear grade is related to an aggressive phenotype, because these lesions are more likely to recur as invasive tumors [
9]. Although human studies provide important insights into the natural history of DCIS, they usually suffer from having small patient numbers, having a biased lesion population (i.e., only those DCIS that were initially misdiagnosed), performing interventions that could alter disease state and progression (i.e., biopsy or lumpectomy), and focusing on outcome rather than detailed measurements of lesion morphology or biology. It is difficult to fully understand DCIS development or the key steps involved in progression of
in situ disease without detailed empirical data directly following DCIS as it develops and progresses over time.
Transgenic mouse models of human breast cancer provide an experimental framework with which to begin to understand the natural history of DCIS. Because of the small size of
in situ mammary neoplasias in mouse models, high-resolution imaging techniques are required to effectively observe how lesions develop, grow and progress over time. Recently, our laboratory reported high-resolution
in vivo magnetic resonance (MR) images of pre-invasive mammary intraepithelial neoplasias (MIN) in the simian virus 40 large T antigen (SV40 Tag) mouse model of human breast cancer [
10]. This work demonstrated that both MIN and early invasive cancers could be identified and accurately classified based on their appearance on MR imaging, using histological analysis of the same lesions for verification. In the current study, we used these new MR techniques to follow
in situ mammary neoplasias in SV40 Tag mice over time in individual animals. Specifically, the timescales and characteristics of the development and progression of
in situ to invasive carcinoma were evaluated, and predictive markers of invasive progression were explored.
Discussion
We have used serial MR imaging to study the natural history of
in situ neoplasia in a transgenic model of human breast cancer. The timescales of neoplastic initiation and progression to invasive cancer in C3(1) SV40 Tag mice that can only be derived from repeated non-invasive imaging were measured. Significantly, we found that even in these mice that are genetically predisposed to develop invasive carcinoma, a substantial proportion of
in situ cancers did not progress to invasive tumors within 21 weeks of monitoring. To our knowledge, these results provide the first detailed, high-resolution measurements of early mammary cancer natural history in mice. Our work complements work by Abbey and colleagues that investigated malignant transformation of
in situ mammary cancer in a transplantable tissue model, using positron emission tomography (PET) to provide metabolic characterization at lower spatial resolution (5 mm
3 voxel size compared with 0.0068 mm
3 in our study) [
15,
16].
In the present study, we measured changes in image-based features that are highly predictive of disease stage. As this was a serial imaging study, we could not perform histologic confirmation of the image-based findings. This underscores a significant challenge with noninvasive imaging of disease progression: it is very difficult to determine lesion histology without sacrificing the animal, thereby losing longitudinal information. With larger and more detailed sensitivity/specificity studies that correlate image-based features with a wide variety of histologic presentations, this limitation can be mitigated.
If image-based measurements are to be robust predictors of disease stage, their reproducibility must be established. Our preliminary work suggests that assessment of in situ vs. invasive disease (via determination of lesion type) is reproducible, as are measurements related to size and morphology of early invasive tumors. However, in situ lesion morphology (i.e., distribution or pattern) may not be adequately reproducible. Further studies are needed to quantify reproducibility of MR measurements of early murine mammary cancer.
The C3(1) SV40 Tag mouse model is being used for a wide variety of studies, ranging from evaluating effects of interventional and preventive therapies [
17‐
30], to understanding molecular and genetic alterations occurring at various stages of disease progression [
31‐
36]. Our results contribute new observations regarding this mouse model of breast cancer. MIN lesions grow slowly on average, and can both progress to invasive tumors or remain indolent, as has been suggested to be true for DCIS in women [
2‐
4]. This is consistent with additional stochastic mutations taking place as secondary transformation events beyond expression of Tag as required for cancer progression in this model [
33]. We have also found evidence for
in situ cancer regression, which if validated in larger numbers with detailed pathology correlation, would be direct demonstration of spontaneous breast cancer regression [
37‐
39]. The heterogeneity of progression paths demonstrates that the C3(1) SV40 mouse model may be a good candidate for assessing the effect of therapies that delay the progression of DCIS. Early invasive tumors show less variability in morphology as they grow compared with MIN. Although there was a wide variability of growth rates of early invasive cancers, overall they grew much faster than
in situ cancers, and none decreased in size or regressed. There was a trend for increased MIN growth rate to be a predictor of both the eventual development of invasive carcinoma, and a higher invasive tumor growth rate, in the same region. Unfortunately, in this pilot study the number of cases was too small to draw conclusions with statistical significance.
It is important to note that our results pertain to this specific mouse model; it will be important to establish imaging techniques and assess similar characteristics in other mouse models of human breast cancer. If some features can be found that persist across mouse models, they may ultimately demonstrate applicability to human disease.
The natural history of breast cancer is still an open question, and there are many theories of the mechanisms governing the growth and progression of early breast cancers in women. The 'angiogenic switch' is thought to be a crucial step during breast tumorigenesis, and has been hypothesized to occur at or before the
in situ stage [
40]. Franks and colleagues have used non-linear mathematical models to predict that invasion will occur at the middle of ducts distended by DCIS due to increased mechanical pressure [
41‐
43]. Tabar and colleagues suggest that true
in situ lesions in fact originate in lobules, and that a separate more aggressive disease representing a duct-forming invasive carcinoma is being incorrectly included with other
in situ cancers [
44,
45]. Due to an absence of empirical data of the detailed morphologic changes and other changes that occur during progression of
in situ cancers, such theories may be difficult to evaluate. Our work and extensions thereof, for example the use of dynamic contrast-enhanced MR imaging to probe changes in vasculature, can provide detailed and direct measurements of tumorigenesis on which these mathematical and physiologic models of disease initiation and progression can be evaluated.
There are several limitations to this study. First, lesion morphology was assessed using 2D axial slices rather than a 3D rendering of the inguinal glands. This could compromise the assessment of lesion morphology, particularly of MIN located in the lower gland area.
Second, there may have been some errors in lesion identification. Although the descriptors of 'nonmass' and 'mass' are highly specific to MIN and invasive tumors, respectively, the 'mass' descriptor is not perfectly correlated [
10] implying that some MIN lesions may have been misidentified as invasive tumors. More generally, in this study it may have been difficult to distinguish focal MIN from invasive cancer, or to pinpoint the exact point of transition from MIN to invasion due to the two to three-week sampling interval. On the other end of the progression spectrum, distinguishing MIN from benign conditions, such as atypical ductal hyperplasia or epithelial proliferative diseases, may be a challenging task in mice as it is in women. A much larger sensitivity/specificity study will be required to better correlate a wider variety of image-based features with histology, in order to minimize such confusion so that disease stage can be assessed with increased confidence. Similarly, MR imaging of nontransgenic normal mice should be performed to document the presentation of normal murine mammary gland anatomy, upon which findings in transgenic models can be compared.
Third, the numbers of lesions studied was rather small, limiting the statistical significance of our findings. To address this, both an increased number of mice should be imaged as well as an increased number of mammary glands in each mouse (rather than only the inguinal glands on one side). In this way, cancer development can be assessed in the whole mouse, and data can be analyzed to determine whether lesions in the same animal can be considered as independent.
Fourth, in this study we did not consistently perform end-point histologic evaluation of imaged mammary glands.
Fifth, mice were followed to 20 or 21 weeks of age rather than until natural death, so that studies of cancer progression could not be performed past this age. Although the range of cancer development in this model spans 10 to 24 weeks of age, and most female mice must be sacrificed at this time due to increased tumor burden in the mammary glands, additional time points should be acquired to definitively assess in situ cancer progression.
Sixth, changes in the parenchyma that preceded the development of MIN could not be easily observed because of the poor signal-to-noise ratio of the normal tissue. Recent improvements in imaging methods have provided greatly enhanced images of normal parenchyma, opening up the possibility of studying changes in the normal mammary glandular tissue that are precursors to cancer development.
Finally, the new framework we have presented for analyzing early carcinogenesis in mice may need improvement. For example, the number of MIN lesions that progressed to invasive tumors may have been over-estimated. Our criterion was only that an invasive tumor appeared in the same region on subsequent imaging; however, this tumor may have been independent of the original MIN detected previously. The transition from MIN to invasive tumors was rarely observed directly. In addition, cancer growth rates could only be calculated for lesions imaged at least twice, that is, 15 of 21 MIN and 8 of 14 invasive tumors. The remaining lesions were excluded from any analysis of growth rates, which may have introduced a bias. These two limitations could be addressed by conducting serial imaging at higher frequency (i.e., every few days) so that each MIN lesion can be definitively linked with its subsequent invasive phase, and so that growth rates can be measured for all. Lastly, the statistical model we used to identify indolent lesions could most likely be improved or modified.
In prior work, we introduced MR imaging methods for imaging early murine mammary neoplasias and invasive cancers, and subsequently reported on how those techniques could be used to better interpret clinical MR imaging of the breast [
46]. Here, we have established a new role for MR imaging in preclinical studies of the natural history of early breast cancer. In future work, we plan on performing more detailed studies of carcinogenesis, by imaging more frequently and at higher resolution. In addition, we will explore additional MR imaging techniques, such as dynamic contrast enhanced MR imaging, diffusion weighted imaging and high spectral-spatial resolution imaging, to probe the changes in vasculature and cellularity that occur during progression to invasive cancer. Finally, molecular imaging and gene/protein expression studies will be explored in conjunction with MRI to interrogate the molecular mechanisms involved in cancer initiation and progression.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
SJ conceived of the study and experiment design, conducted the imaging experiments, performed the data analysis and drafted the manuscript. SC participated in conception and design of the study and provided transgenic mice for imaging. XF helped to draft the manuscript and perform data analysis. EJ is the veterinary technician that participated in all imaging experiments. GM helped conceive and design the study. GK conceived of the study and experiment design, and participated in its coordination and helped to draft the manuscript. All authors read and approved the final manuscript.