Evaluation of Cell Proliferation in Rat Mammary Glands Is Not Predictive of the Carcinogenic Potential of Insulin In Vivo

During the nonclinical safety-assessment process of insulin analogues, regulatory guidelines recommend assessing the carcinogenic potential in vivo.^1-4 As native human insulin (HI) is mitogenic in vitro, it is of interest to determine the mitogenic potency of an insulin analogue relative to that of HI, which is therefore recommended as a reference compound in these studies.¹ Primary end points involve an evaluation of proliferative changes based upon histological examination of all major organs and tissues; additionally, the guideline suggests that assessment for cell proliferation (CP) should also be included in order to further evaluate any proliferative effects.¹

The well-established mitogenic effect of insulin in vitro in mammary carcinoma cell lines^{5- 7} is thought to also apply for mammary gland tissue in vivo. In the early 90s, the publication of a study showing a carcinogenic effect of the insulin analogue insulin X10 (hereafter referred to as X10) in rat mammary glands raised concerns regarding the possible increased proliferative effect of modified insulin analogues in vivo.⁸ Consequently, assessment of CP in mammary gland has often been included in the nonclinical studies evaluating the carcinogenic effects of new insulin analogues, and X10 is recommended as a positive control.¹ However, knowledge of HI and insulin analogues has accumulated vastly since these results were published and subsequent rodent studies challenge this in vivo proliferative effect as the incidence of mammary tumors in healthy animals dosed with HI is not increased.^9,10 Moreover, HI did not show growth-promoting effects on already established tumors in rodent mammary cancer models.^11,12 This could suggest that the proliferative effect of insulin in vivo is not as marked as previously suspected. This brings to question the value of including CP assessments when evaluating the carcinogenic effects of insulin analogues, particularly since more accurate histopathologic examinations of mammary gland tissue are included as final end points.

An additional issue is the use of X10 as a positive control. Insulin analogue X10 has a greater mitogenic effect than HI on malignant mammary epithelial cell lines.^5,7,13,14 Furthermore, 2 weeks of X10-dosing to mice promoted growth of carcinomas established by inoculation of mammary adenocarcinoma cell lines, typically used for in vitro studies.¹⁵ In contrast, in a genetic mouse model which develops mammary carcinomas over time, supra-physiologic doses of X10 only decreased latency time for tumor development but not the incidence or growth after 16 weeks.¹¹ Besides the study from 1992,⁸ no other studies have to our knowledge showed carcinogenic effects of X10 in mammary glands of healthy animals. This brings into question the relevance of using X10 as a positive control when evaluating the carcinogenic effects of insulin analogues.

Here we present results from 4 nonclinical rat studies assessing the carcinogenic potential of the long-acting insulin analogues insulin detemir and insulin degludec, as well as HI comparator groups with dosing for up to 52 weeks. These were performed according to regulatory guideline recommendations,¹ and all studies included evaluation of mammary gland CP. The studies were not performed as replacements for carcinogenicity studies, but rather as studies with focus on carcinogenic effects following long-term dosing. Data from an additional investigative study, where the aim was to evaluate CP and histopathologic changes induced by dosing with the typical recombinant HI comparators, neutral protamine hagedorn (NPH) and HI, as well as the analogue X10, is also presented.

Using the results from these studies as a starting point, and relating them to the literature, the aims of the present article were to discuss: (1) the value of CP evaluation in the nonclinical assessment of the carcinogenic potential of insulin analogues, (2) the relevance and usefulness of X10 as a sensitive positive control for CP, and (3) the reliability of the typically used CP markers.

Premature deaths for all the studies are listed in Supplemental Table 2.

Overall, dosing female SD rats with insulin detemir and degludec did not affect the incidence of mammary gland tumors. The same was true for the comparator groups dosed with recombinant HI, NPH, or HI; the positive control groups dosed with X10 also did not differ from controls or comparator groups. Consistent with this, CP was not increased following dosing with insulin degludec or X10, and the same overall picture was seen for HI/NPH. Generally, insulin detemir did not increase CP, but there were a few exceptions (see Table 7), which are discussed below. One to three different CP markers were used in all five studies; however, results from the different markers did not always align across the studies. Based on the reported results, we here discuss the value of CP evaluation in the nonclinical assessment of the carcinogenic potential of insulin analogues, the relevance and usefulness of X10 as a positive control and finally, the reliability of the typically used CP markers.

This lack of in vivo stimulation of mammary gland hyperplasia by HI after 1 and up to 52 weeks of dosing, as well as the analogues insulin detemir and degludec after 4 to 26 and 52 weeks, respectively (see Table 7), do not reflect the well-established mitogenic effect of HI in vitro.^5-7 Similarly, insulin detemir and degludec have also been shown to increase mitogenic activity in vitro, though less potently than HI.^6,21,22 This lack of correlation between in vitro and in vivo mitogenic effects could potentially be due to the different conditions, in particular, the presence of IGF-1 in vivo which overshadow the proliferative effects of insulin.²³ Inadequate dose levels may also contribute to the lack of observed in vivo effect of insulin on mammary gland hyperplasia. However, in the setting of the present studies, NPH doses of 144 and 65 nmol/kg/d (St2 and St4) proved to exceed maximum tolerable dose (causing clinical signs of hypoglycemia) after approximately 3 and 11 weeks, respectively. Consequently, it was not possible to test higher doses; the same was seen for insulin degludec doses of 100 and 65 nmol/kg/d (St4). Interestingly, HI and insulin detemir doses as high as 600 nmol/kg/d were used for 26 weeks (St1). However, 10/30 HI-dosed animals died prematurely during the study, primarily from 11 weeks and afterwards; a similar response was seen in the insulin detemir-dosed group (12 of 30) from week 7. Therefore, higher doses of HI and insulin detemir would not have been tolerated, at least not for dosing beyond 10 and 6 weeks, respectively. Unfortunately, histopathologic evaluation was not performed after 26 weeks in this study; however, 4 weeks of HI-dosing did not induce histopathologic changes, nor did 400 nmol/kg/d for 3 weeks (St5). Looking at the literature, administering supraphysiologic doses of HI (1,200 nmol/kg/d) to female SD rats for 52 weeks does not induce mammary gland tumours either.⁸ Likewise, a later study using the same design (performed by Novo Nordisk A/S) showed a comparable number of animals with adenocarcinomas in the HI- versus vehicle-dosed group (3 vs 1 tumor-bearing animals, P = 0.12, 17-19 animals/group, unpublished data). Therefore, the lack of a cell proliferative effect in the present studies was likely not attributed to the dose levels.

A second thing to consider is that this lack of a mitogenic effect could be attributed to an insufficient duration of the studies. Though, St3 and St5 included NPH-dosing for as long as 26 and 52 weeks, respectively. Moreover, as mentioned above, two studies dosing HI for 52 weeks did not find a carcinogenic effect in mammary glands (unpublished and study by Jorgensen et al⁸); others have shown that as long as 2 years of dosing healthy female SD rats and mice with NPH (30 and 75 nmol/kg/d) did not affect incidence of mammary adenocarcinomas either.⁹ Thus, the studies are considered to be of sufficient duration to assess the mitogenic effect.

An additional factor to keep in mind is, that in nonclinical safety assessment studies insulin is dosed to healthy rats; the lack of a mitogenic effect of HI in mammary gland tissue in vivo could potentially be attributed to this, since in vitro the mitogenic effect is evaluated in immortal cell lines. However, in high-fat fed female SD rats with already established mammary tumors (carcinogen-induced), 6 weeks of NPH-dosing (90 nmol/kg/d) did not affect tumor incidence (60%-80% in all groups) or growth¹²; the same was seen with insulin detemir, also included in that study.¹² Also, life-long (67 weeks) NPH-dosing (≤125 nmol/kg every other day) to a genetic mouse model of mammary cancer showed the same lack of effect on tumor incidence and growth rate.¹¹ So, the use of healthy rats is likely not the reason for the lack of mitogenic effect.

Overall, findings from the five studies presented here, as well as published studies, suggest that HI, as well as the insulin analogues insulin detemir and degludec, have a weak mitogenic effect in the mammary gland in vivo, despite the known mitogenic effects in vitro. Here it should be mentioned that when evaluating insulin analogues in vitro, a key factor is the binding affinity to the IGF-1 receptor (associated with mitogenic effects), and most importantly, the relative binding compared to the affinity to insulin receptors. Human insulin has a very low affinity for the IGF-1 receptor, approximately 1,000 fold lower than IGF-1, and as such, the binding of insulin to the IGF-1 receptor is not significant at physiological insulin concentrations.²⁴ Thus, for insulin analogues, it is essential to maintain this ratio between insulin and IGF-1 receptor binding affinities to achieve the desired metabolic effects of insulin and to not unfavorably alter the balance between the metabolic and mitogenic (cell growth) effects compared to that of endogenous HI. For both insulin detemir and degludec, relative to HI, the IGF-1 receptor binding affinity is slightly reduced and the mitogenic/metabolic potency ratio is ≤1.^21,22

Generally, insulin detemir did not increase CP (see Table 7); however, there were a few exceptions. Cell proliferation was increased in a few insulin detemir-dosed groups in two of the studies (St1, St2). Four weeks of insulin detemir-dosing at dose levels of 150 and 300 nmol/kg/d (low- and mid-dose groups) and 13 weeks with 300 nmol/kg/d (high-dose group) increased CP significantly compared to vehicle-dosed groups. Unfortunately, CP was not evaluated in the NPH-dosed group in the 13 week study (St2); therefore, it cannot be assessed if CP in the insulin detemir-dosed group was comparable to that seen with HI, illustrating the importance of such a comparator group. However, in the 4-week study (St1), CP in the insulin detemir-dosed groups was not different from the NPH- or HI-dosed comparator groups. Also, increased CP was only seen in the low-and mid-dose groups in the 4 week study (for 1 of 2 CP markers), and not in the high-dose group, making it unlikely that this is an insulin detemir-specific effect. Moreover, none of these groups had malignant hyperplastic changes in mammary glands. These results show that CP in mammary gland does not seem to be affected by short- or long-term dosing with insulin detemir and degludec. Looking at mammary gland CP in HI/NPH-dosed animals, the present studies showed no differences compared to controls following dosing periods spanning from 1 to 26 weeks (St1, St5). One exception was 4 weeks of dosing with relatively high doses of HI (St1, 600 nmol/kg/d), which resulted in increased CP LI. However, this was only for CP evaluated using PCNA, but not with Ki-67, and CP was not different from the NPH-dosed group. Consequently, it is considered an incidental finding. Cell proliferation was determined in the NPH-dosed group after 52 weeks (St4), but since no changes were seen in the insulin degludec-dosed groups, the result was not adjusted for estrous stage for statistical comparison to the vehicle group. However, looking at the unadjusted CP LI in the NPH-dosed group, this was not different from the vehicle group. This is in line with a 6 month study in female SD rats, where mammary gland CP was unchanged in HI-dosed animals (240 nmol/kg/d).¹⁰ Thus, dosing healthy rats with HI for as long as 52 weeks did not induce any increases to CP in the mammary glands.

Overall, HI, insulin detemir, and degludec did not promote tumor formation/CP, even when dosed for extended time periods and/or at high doses in the present, as well as published studies. This indicates that insulin may not be as mitogenic in mammary glands in vivo as previously suspected. This corresponds to what is seen in diabetic women treated with HI or insulin analogues, where the risk of breast cancer is not increased.^25,26 Also, as stated by others,²⁷ it is important to keep in mind, that increased mitogenicity, such as that seen in vitro, does not necessarily lead to increased carcinogenicity, that is, a malignant transformation of normal cells. If insulin itself only has a weak mitogenic effect in vivo, CP may not be a strong predictor of the carcinogenic potential of new insulin analogues. This brings into question the value of CP assessment in mammary gland tissue during nonclinical assessment of the potential carcinogenic effects of new insulin analogues.

In contrast to CP LI, the standard histopathological evaluation of mammary gland tissue included in nonclinical studies allows for the distinction between benign and malignant tissue changes as well as grading of the severity. This distinction may be important when assessing potential carcinogenicity. For instance, in the 52 week repeat dose study (St4), 11 of 49 animals in the NPH-dosed group had hyperplastic changes in mammary gland tissue; however, 8 of these were benign and only 3 malignant, as recognized by histologic evaluation. So instead of a result showing hyperplastic changes in 22% of the animals, it was recognized that only 6% of the animals had malignant hyperplastic changes, similar to the control group. With CP LI this distinction cannot be made. Therefore, in line with the discussion above, including CP evaluation in studies assessing the carcinogenic potential of insulin analogues in vivo may not contribute with any additional information of value.

Furthermore, hyperplasia and tumor development may be attributed to decreased apoptosis rather than increased CP. Human insulin as well as several insulin analogues decrease cell apoptosis in vitro, primarily in cancerous cell lines.^28-33 In rats with chemically induced mammary carcinomas, the hyperplastic and preneoplastic lesions showed unchanged CP, but decreased numbers of apoptotic cells.³⁴ This suggests that neoplastic transformation of mammary epithelial cells may be associated with a longer life span of cells rather than increased CP. Other mechanisms involved may include changes to early differentiation of mammary epithelial cells (discussed elsewhere³⁵). Also important is that the CP LI could fail to reveal increased CP in already hyperplastic tissues as the proportion of proliferating compared to the total cell number may be unchanged in this scenario. Thus, preneoplastic hyperplasia may be present despite unchanged CP LI and results from CP assessment should be interpreted with this in mind.

The insulin analogue X10 is often included as a reference when evaluating carcinogenic potential of new insulin analogues in vivo, as recommended by the guideline.¹ Insulin analogue X10 has been deemed a positive control, based on the findings from a single published study showing its carcinogenic effect in mammary glands in healthy rats in response to suprapharmacological doses.⁸ In vitro, X10 appears more mitogenic than HI, primarily in cancerous mammary epithelial cell lines,^5-8,13,14 and both the relative IGF-1 and insulin receptor binding affinity as well as the mitogenic/metabolic potency ratio are >1^6,21,22 Consistent with this, dosing mice with X10 for 2 weeks (150 nmol/kg/d) promoted growth of mammary carcinomas induced by injecting such malignant mammary cell lines (HI was not included).¹⁵ Here, X10 did not increase tumor incidence or CP in mammary glands in rats after up to 3 weeks of dosing with 400 nmol/kg/d (St5) nor after up to 26 weeks of dosing with 600 nmol/kg/d (St1). This could be attributed to the lower doses and shorter duration of these studies compared to the previous study showing a carcinogenic effect of X10 in rats following highly supraphysiologic doses (1,200 nmol/kg/d) for 52 weeks.⁸ During the 26 weeks of dosing in St1, 13 of 30 animals were found dead or prematurely sacrificed for welfare reasons (clinical signs of hypoglycemia), thus a higher dose level would not be feasible. However, a study in a genetic mouse model of mammary cancer used X10 doses of up to 1,800 nmol/kg every 2 days for 67 weeks, with no effect on incidence or growth rate of carcinomas.³⁶ Furthermore, a 52 week study with X10- and HI-dosing was subsequently performed in healthy female SD rats, using the same doses as Jorgensen et al.,⁸ no differences were seen in incidence of mammary gland adenocarcinomas between these groups (3 tumor-bearing animals in each of the X10- and HI-dosed groups, Novo Nordisk A/S, unpublished data). The fact that no other studies, including one using the same study design, have replicated the results reported by Jorgensen et al. in 1992, and the seemingly low in vivo mitogenic effect of HI itself (discussed above), question the relevance of these results and indicate that X10 may not be suitable as a positive control when assessing the carcinogenic potential of new insulin analogues in vivo. As an alternative, detection of estrous cycle-driven fluctuations in CP can be used as a positive control.

The present studies showed a lack of consistent correlation among the CP markers PCNA, Ki-67, and BrdU (see Table 7). Particularly PCNA varied within and between studies and importantly, the PCNA LI was not increased in the estradiol + progesterone-dosed positive control group after 26 weeks (St1) and showed weak correlations with CP fluctuations during the estrous cycle (St1-3). PCNA is endogenously expressed at high levels in replicating cells as well as at very low levels in resting cells, and has a long half-life,^17,37,38 increasing risk of false positive CP signal in resting cells. Also, discriminating positive from negative PCNA-stained cells is problematic,^16,37-43 emphasized by lack of correlation between the number of PCNA positive versus mitotic cells in human breast cancer tissue.^39,44,45 Collectively, this suggests that PCNA is an unreliable marker for CP in mammary glands. The CP marker Ki-67 is also expressed endogenously in replicating cells, but has a shorter half-life, separation of positive from negative cells seems straightforward, and the mitotic count in human breast carcinomas is positively correlated with the Ki-67 signal.^42,46-50 BrdU is a synthetic nucleoside incorporated into cellular DNA during replication, meaning no disturbing endogenous signal.^49,51 Both Ki-67 and BrdU LIs were increased in the estrogen + progesterone-dosed group in St1 and showed strong correlations with hormone-induced fluctuations in CP during the estrous cycle (St1-5), and both markers showed highest CP during met-estrous and lowest during pro-estrous, in agreement with the literature.^16,52,53 This illustrates that these markers can detect increased mammary gland CP, when it is present, such as seen during the estrous cycle.^16,52,53 Overall, BrdU seems to be the most robust marker for the evaluation of mammary gland CP, followed by Ki-67, whereas PCNA may present important technical issues and inconsistent results. The above should be considered when including CP evaluation in nonclinical studies.

In summary, neither HI (HI or NPH), insulin detemir, degludec, nor X10 induced mammary tumours or increased CP in the present studies. The data presented here and results from the literature suggest that, in contrast to the general belief, the mitogenic effect of insulin in rat mammary glands is weak in vivo, even in already established malignant mammary tumours in rats and mice. Evaluation of mammary gland CP as an addition to histopathologic evaluation during nonclinical assessment of carcinogenicity may not be predictive of the carcinogenic potential of insulin in vivo or contribute with any additional information of value. Moreover, the usefulness and relevance of including an X10-dosed group as a positive control in such studies are questionable due to the lack of a consistent increase in CP or induction of tumor formation. Instead, detection of fluctuations in CP during the estrous cycle can be used as a positive control. It should also be stressed that most important is that CP of a new insulin analogue is not different from that of HI. Furthermore, the quality/reliability of the typically used CP markers differ substantially, which should be considered when including CP evaluation in nonclinical studies and when interpreting the results.

In conclusion, CP does not correlate with incidence of mammary gland tumors and may not be predictive of the carcinogenic potential of insulin in vivo, therefore, the value of including evaluation of CP in nonclinical studies is debatable. Furthermore, the use of X10 as a positive control in such studies is not recommended, and the typical CP markers vary greatly in quality with BrdU being most reliable (BrdU > Ki-67 >> PCNA).