**3. Results**

#### *3.1. Distinct Kinome Response to MET Inhibition Present in MET-Addicted MPNSTs*

To understand how RTK amplification and enhanced RTK signaling impact the MPNST kinome, we assessed the influence of both the *MET* copy number and MET kinase inhibition on the drug response and resistance. Both *MET* and its ligand, hepatocyte growth factor (HGF), are implicated in NF1-related MPNST initiation and progression [21–23]. Previously, our genomic analysis of human MPNST progression revealed that *MET* and *HGF* copy number gains are present at the earliest stage of neurofibroma transformation and increase during metastasis and resistance [6]. Moreover, studies in other cancers have demonstrated that aberrant MET signaling can drive malignant progression in a variety of RAS-deregulated human tumors and augmen<sup>t</sup> the oncogenic e ffects of RAS activation [24,25]. To understand the impact of the MET genomic status on kinome adaptations, we evaluated the response and resistance to the potent and selective MET inhibitor capmatinib in three diverse models of NF1-related MPNSTs, including an "MET-addicted" model (NF1-MET), an *Nf1*/*Trp53*-deficient model (NF1-P53), and an NF1 model (P53WT, METWT, *Hgf*-amplified). As we previously showed, NF1-MET MPNSTs were uniformly sensitive to MET inhibition, whereas a heterogeneous response to MET inhibition was observed in NF1-P53 and NF1 MPNSTs (Figure 1A–C) [6]. To characterize the kinome response to MET inhibition, we performed pathway activation mapping of 98 proteins and phosphoproteins. This was a targeted pathway activation analysis focused on actionable targets of RTK-mediated signaling, downstream PI3K-mTOR signaling, downstream RAS-ERK signaling, and motility/adhesion signaling. To assess the immediate, early, and late kinome responses to kinase inhibition, we profiled the tumor phosphoproteome after 4 h, 2 days, and 21 days of treatment. With these time points, we anticipated that both innate and acquired kinome adaptations would be observed in the various genomic backgrounds. Changes in the expression relative to the vehicle were plotted in rank order for each timepoint. For the 21-day RPPA analysis of each MPNST model, we analyzed tumors that had diverse treatment responses, while avoiding tumors that exhibited grossly anomalous growth patterns compared to the mean growth curve (see individual tumor annotations in Figure 1A–C). By including diverse tumors, we anticipated that we would detect the heterogeneity of mechanisms underlying drug resistance. For example, because the NF1-MET tumors are "*Met*-addicted", substantial growth inhibition was present at 21 days and minimal heterogeneity in the drug response was observed (Figure 1A) This homogeneous response was not observed in the other MPNST tumor-graft lines (Figure 1B,C). Correspondingly, we observed a more homogeneous kinome response in NF1-MET tumors in comparison to the responses observed in NF1-P53 and NF1 tumors (Figure 1D–F; Figure S1).

**Figure 1.** MET inhibition reveals differential innate and adaptive kinome reprogramming. Individual tumor growth curves for (**A**) Neurofibromatosis Type 1 (NF1)-MET, (**B**) NF1-P53, and (**C**) NF1 tumorgrafts plotted by treatment (colored lines) compared to the vehicle (black lines). The analysis of tumor growth data was previously reported [6]. The annotated tumors were analyzed by a reverse phase phosphoproteome array (RPPA) (**D**–**F**). The fold change relative to the mean protein expression of control tumors (i.e., #1–3) was calculated for each tumor #4–6, with the first column of Panel A at 21 days corresponding to tumor #4, the second to tumor #5, and the third to tumor #6. Ranked balloon plots of the proteins with the highest and lowest fold change in expression after 4-h, 2-day, and 21-day treatment of the NF1-MET model with capmatinib. Each column represents a single animal. Balloon color indicates the fold change in expression relative to the vehicle mean (*n* = 3) for that time point. Balloon size indicates the absolute protein expression normalized to the total protein input and background.

After 4-h capmatinib treatment, we observed a striking repression of ERK, AKT, and RTK phosphorylation that corresponded to growth reduction in the NF1-MET tumors (Figure 1D). Overall, minimal kinome activation was observed at the 4-h time point in growing NF1-MET and NF1-P53 tumors (Figure 1D,E; Figure S1B,C); however, two of three NF1 tumors had phosphorylation changes in several pathways at the 4-h time point (i.e., PRK, AKT, and p38MAPK) (Figure 1F). After 2-day capmatinib treatment, we observed increased activating phosphorylation at several sites in the NF1-P53 and NF1 tumors, including AXL (Y702), cofilin (S3), and 4EBP1 (T37/T46) (Figure 1E,F; Figure S5), which is a finding that correlated with the relatively increased capmatinib resistance at 21 days (Figure 1B,C). In the NF1-MET tumors, NFκB demonstrated the strongest increase in phosphorylation at the 2-day time point. This probe corresponds to S536 in the transactivation domain (TAD) of NFκB/p65, which leads to transactivation. Interestingly, at the 2-day time point, NFκB/p65 was in the top three most increased phoshposites in all of the tumor models after 2-day MET inhibition. Since NFκB is a master regulator of the inflammatory response, survival, and tumor proliferation [26], and a known mediator of pathway indi fference [27], NFκB activation at the 48-h time point may represent a common kinome adaptation that is agnostic to the MPNST genomic context.

After 21 days of capmatinib treatment, significant tumor death was observed in the NF1-MET tumors (Figure 1A), with only a small layer of viable cells present at the edge of the tumors [6]. This is in contrast to the NF1-P53 and NF1 tumors that maintained a significant decrease in growth compared to the vehicle control. An upward growth trend was observed in the majority of tumors at the 21-day time point, despite ongoing treatment (Figure 1A–C). In the surviving, capmatinib-resistant cells present at 21 days in NF1-MET tumors, distinct changes in the kinome response were observed, comprising consistent AXL (Y702), EGFR (Y1068), cofilin (S3), and AKT (S473) activation (Figure 1D; Figure S5). These results sugges<sup>t</sup> that MET-addicted MPNSTs survive MET inhibition through pathway reactivation via other RTKs (i.e., AXL and EGFR) and potentially a pathway bypass through AKT signaling. In the NF1 tumors whose ascending growth patterns indicated the beginning of drug resistance, increased phosphosite activation was observed in ERK, ribosomal protein S6 kinase (S6), 4EBP1, and AKT, ye<sup>t</sup> markers of parallel RTK activation were also present. In the F1-P53 tumors, which were continuing to grow at the 21-day time point, phosphosite expression returned to levels resembling the vehicle, suggesting that broader kinome adaptation was no longer required for growth. Collectively, these data indicate that distinct mechanisms of innate and adaptive kinome reprogramming occur in genetically diverse MPNSTs.

#### *3.2. Kinome Response to MEK Inhibition Results in Bypass Activation*

The recent clinical success of MEK inhibition with selumetinib in NF1 plexiform neurofibromas and recent preclinical MPNST treatment studies highlight the therapeutic potential of targeting MEK in NF1-related MPNSTs [28–30]. To evaluate the kinome response to MEK inhibition in NF1-deficient MPNSTs with distinctive genomic backgrounds, we used the MEK inhibitor trametinib (Novartis, Cambridge, MA, USA). Trametinib is a reversible, highly selective, allosteric inhibitor of MEK1 and MEK2, which is FDA approved for melanoma, lung cancer, and anaplastic thyroid cancers with BRAF mutations. MEK inhibition significantly decreased tumor growth in all of the MPNST lines, ye<sup>t</sup> substantial response heterogeneity was observed in the NF1-MET and NF1-P53 tumors (Figure 2A–C) [6]. The most uniform tumor inhibition was observed in the NF1 MPNST tumors, whereas some NF1-P53 tumors still displayed aggressive growth after 21 days. As with capmatinib, RAS and AKT pathway inactivation (i.e., ERK1/2, mTOR, S6, and p90RSK) was observed after 4 h of trametinib in the NF1-MET and NF1-P53 tumors (Figure 2D,E). Interestingly, broader kinome activation was not observed in these same genomic contexts, suggesting that NF1-MET and NF1-P53 tumors maintain a limited MEK dependency due to innate resistance (Figure 2A,B,D,E; Figure S2). Interestingly, by 2 days, trametinib treatment induced a similar response to capmatinib in the NF1-MET tumors, strongly activating EGFR (Y1068), AXL (Y702), PKCz (L410/T403), and NFκB (S536) (Figure 2D; Figure S5). In contrast, trametinib treatment resulted in the di fferential regulation of EGFR (Y1068) and AXL (702) in NF1-P53 tumors, as AXL (702) was upregulated, while EGFR (Y1068) was the most repressed site after 2 days (Figure 2D,E). AXL (Y07) was also highly induced after 2 days of trametinib treatment in the NF1 tumors (Figure 2F), suggesting that AXL activation may be a universal early response to MEK inhibition, regardless of the genetic context of the MPNST.

Adaptive kinome reprogramming in response to trametinib was distinct for each model. In the NF1-MET tumors, AXL (Y702) remained activated after 21 days of treatment. 4EBP1 (T37/T46), CHK1 (S345), and AKT (e.g., SGK and AKT) phosphorylation was observed in response to long-term MEK inhibition (Figure 2D; Figure S5). These results implicate a bypass mechanism of resistance to MEK inhibition, particularly in disparate signaling nodes within the AKT and mTOR pathways (i.e., SGK, CHK1, AKT, 4EBP1, and HSP27). Notably, after 21 days of treatment, 90% of NF1-P53 tumors had increasing growth trends and a negligible kinome response to MEK inhibition (Figure 2B,E). ERK was consistently inhibited by trametinib in these resistant tumors, confirming that ERK pathway reactivation was not required to maintain growth. In the NF1 tumors, AKT/mTOR and protein translation pathway effectors were the strongest targets activated by MEK inhibition. Collectively, the NF1-MET kinome response to both MET and MEK inhibition suggests that RTK-dependent MPNSTs may survive kinase inhibition both by the engagemen<sup>t</sup> of alternative RTKs (i.e., AXL and EGFR) and increasing AKT/mTOR signaling pathways.

**Figure 2.** MEK inhibition reveals differential innate and adaptive kinome reprogramming. Individual tumor growth curves for (**A**) NF1-MET, (**B**) NF1-P53, and (**C**) NF1 tumorgrafts plotted by treatment (colored lines) compared to the vehicle (black lines). The analysis of tumor growth data was previously reported [6]. The annotated tumors were analyzed by RPPA (**D**–**F**). The fold change relative to the mean protein expression of control tumors (i.e., #1–3) was calculated for each tumor #4–6, with the first column of Panel A at 21 days corresponding to tumor #4, the second to tumor #5, and the third to tumor #6. Ranked balloon plots of the proteins with the highest and lowest fold change in expression after 4-h, 2-day, and 21-day treatment of the NF1-MET model with trametinib. Each column represents a single animal. Balloon color indicates fold change in expression relative to the vehicle mean (*n* = 3) for that time point. Balloon size indicates the absolute protein expression normalized to the total protein input and background.

#### *3.3. Kinome Response to Combined MET and MEK Inhibition in NF1-Related MPNSTs*

Since we observed both pathway reactivation and bypass resistance mechanisms with single-agent MET or MEK inhibition, we sought to determine whether targeting multiple signaling pathways may abrogate these kinome adaptations and achieve a more durable clinical response. Previously, we compared combined MET and MEK inhibition with monotherapy and demonstrated significant improvement in tumor inhibition and response variability with combination therapy compared to a single agen<sup>t</sup> alone (Figure 3A–C) [6]. Even in NF1-P53 tumors which had the most heterogeneous responses to monotherapy with capmatinib or trametinib, we observed stable disease in all but one tumor (Figure 3B). The kinome response to combined MET-MEK inhibition exhibited striking di fferences in comparison to single kinase inhibition. At 4 h and 2 days of treatment, ERK1/2, S6 (S240/S244 and S235/S236), and p90RSK demonstrated the strongest decrease in phosphorylation in all of the MPNST models, suggesting that the RAS/ERK and AKT/mTOR pathways are robustly inactivated with combined MET-MEK inhibition (Figure 3D–F). As with single kinase inhibition, we observed NFκB/p65 (S536) activation at the 2-day time point. We also measured an increase in PKCζ/λ (T410/T4033) and cofilin (S3) phosphorylation at the 2-day time point with both single and combined kinase inhibition (Figure S5). Cofilin is an actin depolymerizing factor known to regulate actin dynamics and cell invasion; however, recent studies have established the role of cofilin in NFκB nuclear translocation [31,32]. The atypical protein kinase C member PKCζ is involved in several survival pathways that are deregulated in cancer and is also involved in the activation of NFκB [33,34]. Together, these findings indicate that NFκB activation is an acute response that occurs in response to monotherapy or combined kinase inhibition in MPNSTs.

At 21 days, significant tumor inhibition was observed in NF1-MET tumors and in the surviving cells, the kinome adaptations observed with single MET inhibition were intensified (Figure 3D; Figure S3A). Specifically, AXL (Y702), EGFR (Y1068), and AKT (S473) are strongly activated. Intriguingly, combined MET-MEK inhibition also resulted in the strong activation of AXL (Y702) in the NF1-P53 and NF1 tumors, which was not observed with single-agent treatment of either drug (Figure 3E,F). The NF1-P53 tumors maintained an inflammatory kinome response after 21-day treatment (PKCζ/λ, NFκB), which stands in contrast to the NF1-MET and NF1 tumors, where an inflammatory response was only observed at 4 h and 2 days (Figure 3D–F; Figure S5). Rather, after 21 days of combination therapy, the surviving cells of NF1 tumors robustly activated S6 (S240/S244 and S235/S236) and 4EBP1 (T37/T46), along with AXL (Y702) (Figure 3F). Given that AXL is activated in response to MET and MEK inhibition in all three of these genomically diverse MPNST models, AXL activation may be a common mechanism of therapeutic resistance to RAS pathway inhibitors.

#### *3.4. Kinome Response to Doxorubicin in NF1-Related MPNSTs*

Doxorubicin is a topoisomerase II inhibitor that prevents cellular replication by indirectly stabilizing double-stranded DNA breaks [35]. It has also been implicated in direct DNA damage through free radical production. Doxorubicin is currently being tested in combination with multiple kinase inhibitors for sarcoma (e.g., PDGF α inhibitor), or to treat anthracycline-resistant sarcomas [36–38]. Although the results of these trials are mixed, it is unclear whether doxorubicin resistance is mediated at least in part through kinome adaptation. How NF1-related MPNSTs confer doxorubicin resistance is likely multifactorial [39]; however, the patterns of compensatory kinase signaling have not been studied to date. Following the doxorubicin treatment of NF1-MET, NF1-P53, and NF1 tumorgrafts, significant resistance and response heterogeneity was observed (Figure 4A–C). NF1 tumor growth was significantly slower than controls; however, no tumors ultimately responded to treatment. RPPA analysis revealed a broad and diverse response to doxorubicin at early and late timepoints across all genomic contexts. Early responses at 4 h included RTK activation (EGFR, IGF1R, PDGFR, and MET), pro-inflammatory signaling mediators (p38, NFκB, and STAT3/5), and upstream kinases (SRC and RAF) (Figure 4D–F; Figure S4). Kinome responses at 2 days and 21 days of treatment were more diverse, with the emergence of increases in AXL (Y702), EGFR (Y1068), and cofilin (S3) as dominant signaling mediators (Figure S5). Qualitatively, doxorubicin resulted in the broadest pathway responses compared to single-agent capmatinib (Figure 1), trametinib (Figure 2), and combination therapy (Figure 3). Interestingly, the PI3K/AKT/mTOR pathway response did not appear to be significantly activated in response to doxorubicin, as evidenced by the inactivation of AKT (S473), S6RP, and CHK1 (Figure 4D–F, Figure S4). These results indicate that doxorubicin treatment causes both acute and persistent kinome changes in several pathways. The diversity and perseverance of the doxorubicin-mediated kinome response may underlie innate resistance observed in sarcomas.

**Figure 3.** Combination MEK and MET inhibition reveals differential innate and adaptive kinome reprogramming. Individual tumor growth curves for (**A**) NF1-MET, (**B**) NF1-P53, and (**C**) NF1 tumorgrafts plotted by treatment (colored lines) compared to the vehicle (black lines). The analysis of tumor growth data was previously reported [6]. The annotated tumors were analyzed by RPPA (**D**–**F**). The fold change relative to the mean protein expression of control tumors (i.e., #1–3) was calculated for each tumor #4–6, with the first column of Panel A at 21 days corresponding to tumor #4, the second to tumor #5, and the third to tumor #6. Ranked balloon plots of the proteins with the highest and lowest fold change in expression after 4-h, 2-day, and 21-day treatment of the NF1-MET model with combination therapy. Each column represents a single animal. Balloon color indicates the fold change in expression relative to the vehicle mean (*n* = 3) for that time point. Balloon size indicates the absolute protein expression normalized to the total protein input and background.

**Figure 4.** Doxorubicin reveals differential innate and adaptive kinome reprogramming. Individual tumor growth curves for (**A**) NF1-MET, (**B**) NF1-P53, and (**C**) NF1 tumorgrafts plotted by treatment (colored lines) compared to the vehicle (black lines). The annotated tumors were analyzed by RPPA (**D**–**F**). The fold change relative to the mean protein expression of control tumors (i.e., #1–3) was calculated for each tumor #4–6, with the first column of Panel A at 21 days corresponding to tumor #4, the second to tumor #5, and the third to tumor #6. Ranked balloon plots of the proteins with the highest and lowest fold change in expression after 4-h, 2-day, and 21-day treatment of the NF1-MET model with doxorubicin. Each column represents a single animal. Balloon color indicates the fold change in expression relative to the vehicle mean (*n* = 3) for that time point. Balloon size indicates the absolute protein expression normalized to the total protein input and background.

#### *3.5. Combined MET-MEK Inhibition with Doxorubicin Decreases Response Heterogeneity*

Because combined MET and MEK inhibition resulted in an improved treatment response in all MPNST lines, we investigated the efficacy of doxorubicin in combination with MET and/or MEK kinase inhibition. As discussed earlier, the kinase inhibition of MET or MEK resulted in distinct kinome adaptations compared to doxorubicin treatment. We focused our tumor growth analysis on the NF1-MET and NF1-P53 tumors since these two MPNST models had the most aggressive growth and distinctive responses to MET and MEK inhibition. In NF1-MET tumors, doxorubicin treatment caused a significant decrease in tumor growth (Figure 5A; *p* < 0.0005); however, doxorubicin treatment was inferior to capmatinib or trametinib (Figure 5B). Even though the mean growth reduction

was significant, the heterogeneous response to doxorubicin was substantial (Figure 5C). Combined doxorubicin and kinase inhibition significantly improved tumor inhibition and reduced tumor heterogeneity. For example, trametinib alone resulted in moderate tumor inhibition in NF1-MET tumors (Figure 5A), ye<sup>t</sup> combined trametinib + doxorubicin significantly improved the response in comparison to single-agent treatment with trametinb or doxorubicin (Figure 5B). Since these MPNST tumors are MET-addicted, capmatinib resulted in impressive tumor regression, ye<sup>t</sup> combined capmatinib + trametinib was superior to capmatinib alone (Figure 5B), whereas capmatinib + doxorubicin did not significantly improve the treatment response. The treatment that resulted in the least response variability (SD = 19 mm) was the capmatinib + trametinib + doxorubicin combination, with each tumor showing consistent growth inhibition.

For the MPNST tumorgraft line, the NF1-P53 tumors had the most aggressive growth, highest response heterogeneity, and least impressive response to single-agent treatment. In NF1-P53 tumors, doxorubicin treatment did not result in tumor regression (Figure 5D–F). Combined doxorubicin and kinase inhibition reduced tumor growth in comparison to doxorubicin alone, ye<sup>t</sup> this combination was not better than capmatinib or trametinib single-agent treatment (Figure 5E). The triple combination of capmatinib + trametinib + doxorubicin was not significantly better than capmatinib + trametinib; however, this treatment combination resulted in the least heterogeneity in the response (Figure 5F; SD = 343 mm3). The heterogeneity of response and growth patterns observed correlated with the diversity and intensity of the innate and acquired kinome responses delineated in these genomically distinct MPNST tumors.

**Figure 5.** *Cont*.

**Figure 5.** Combined doxorubicin, MET, and MEK inhibitor treatment reduces the response heterogeneity. Tumor growth of (**A**) NF-MET and (**D**) tumorgrafts are plotted as means with standard errors. 95% confidence intervals for the pairwise differences between the growth rates of the select treatments in the (**B**) NF1-MET and (**E**) NF1-P53 tumors, estimated and tested using linear mixed-effects models with random slopes and intercepts, and false discovery rate-adjusted contrasts. Statistically significant differences (*p*-value < 0.05) between compared therapies are highlighted in red. Individual tumor growth curves for (**C**) NF1-MET and (**F**) NF1-P53 tumorgrafts plotted by treatment (colored lines) compared to the vehicle (black lines). The analysis of tumor growth data and differences in treatment response were previously reported for single-agent treatment of capmatinib and trametinib, and combination treatment of capmatinib + trametinib [6].

#### *3.6. ERK Reactivation is Observed in Cells Resistant to MET or MEK Inhibition*

RPPA revealed the consistent de-repression or reactivation of ERK (T202/Y204) in surviving cells throughout 21 days of kinase inhibitor treatment. To determine if ERK reactivation was specific to resistant subpopulations or the entire tumor, we stained the tumors for phospho-ERK (T202/Y204) after 21 days of single or combination therapy. In vehicle-treated tumors, we observed moderate to strong pERK staining; however, distinct ERK activation patterns were observed in each tumorgraft line (Figure 6A). In the NF1-MET vehicle tumors, ERK activation was intense at the invasive edge of the tumor, whereas ERK activation was moderate to strong and uniformly expressed in the NF1-P53 and NF1 tumors (Figure 6A). After 21 days of single-agent treatment with either capmatinib or trametinib, ERK activation was robust in the NF1-MET and NF1 tumors (Figure 6A–C). Interestingly, ERK activation decreased with single-agent treatment in the NF1-P53 tumors, except for minor cell populations at the invasive edge in some tumors (Figure 6B,C). This decrease was even more pronounced in the combined capmatinib-trametinib-treated NF1-P53 tumors, while the capmatinib-trametinib NF1-MET and NF1 tumors maintained high levels of ERK activation (Figure 6D). To directly compare RPPA and IHC and to understand how ERK phosphorylation changed over time in each model, we plotted the normalized absolute protein expression measured by RPPA for ERK (T202/Y204) for each treatment (Figure 6E). Overall, ERK maintained a similar level of activation or increased over time in the NF1-MET and NF1 models. Moreover, long-term capmatinib treatment induced strong ERK activation in these tumors, whereas NF1-P53 tumors consistently maintained lower levels of ERK activation compared to the NF1-MET and NF1 tumors (Figure 6E).

**Figure 6.** *Cont*.

**Figure 6.** ERK reactivation or pathway indifference drive resistance to kinase inhibition. Phospho-ERK1/2 T202/Y204 expression in each genetic model after 21 days of (**A**) vehicle, (**B**) capmatinib, (**C**) trametinib, or (**D**) combination treatment. (**E**) phospho-ERK1/<sup>2</sup> T202/Y204 expression values were measured by RPPA and calculated as the absolute protein expression normalized to the total protein input and background. Points represent the mean (*n* = 3) for each treatment-genotype-time group. Shaded bars represent +/− SEM.

#### *3.7. AXL NFkB Co-Activation Associated with Therapy Resistance in MPNSTs*

As both AXL and NFkB were highly activated in several treatment conditions, including therapy-resistant tumor growth, we sought to determine whether AXL and NFkB phosphorylation were correlated in our models. Recent studies in other cancer contexts sugges<sup>t</sup> that AXL induces NFkB activation in response to a variety of therapies, including kinase inhibitors (22410775, 23474758, and 25568334). This novel therapy mechanism has been underreported to date. We determined that pAXL expression was tightly correlated to pNFkB (Spearmen's rank correlation rho = 0.729, *p* value = 3.91 × <sup>10</sup>−21) and grouped strongly by time point (Figure 7A,B). Expression was also grouped by treatment, as the phosphorylation of both proteins was highest in the doxorubin and combination capmatinib + trametinib treatment groups (Figure 7B), regardless of genotype (Figure 7A).

**Figure 7.** *Cont*.

**Figure 7.** AXL Y702 and NFkB p65 S536 phosphorylation are highly correlated. Expression of phoshpo-AXL Y702 and phospho-NFkB p65 S536 plotted by (**A**) time and genotype group or (**B**) time and treatment group. Colors indicate treatment time and point shape indicates treatment or genotype groups. Lines indicate loess-predicted fit for each time point; shaded regions indicate 95% confidence intervals. Spearmen's rank correlation rho = 0.832, 0.835, and 0.881 with *p* value = 3.72 × <sup>10</sup>−9, 9.75 × <sup>10</sup>−13, and 6.40 × 10−<sup>15</sup> for the 4-h, 2-day, and-21 day groups, respectively.
