Synthesis, Characterisation and Mechanism of Action of Anticancer 3-Fluoroazetidin-2-ones

The stilbene combretastatin A-4 (CA-4) is a potent microtubule-disrupting agent interacting at the colchicine-binding site of tubulin. In the present work, the synthesis, characterisation and mechanism of action of a series of 3-fluoro and 3,3-difluoro substituted β-lactams as analogues of the tubulin-targeting agent CA-4 are described. The synthesis was achieved by a convenient microwave-assisted Reformatsky reaction and is the first report of 3-fluoro and 3,3-difluoro β-lactams as CA-4 analogues. The β-lactam compounds 3-fluoro-4-(3-hydroxy-4-methoxyphenyl)-1-(3,4,5-trimethoxy phenyl)azetidin-2-one 32 and 3-fluoro-4-(3-fluoro-4-methoxyphenyl)-1-(3,4,5-trimethoxyphenyl)azetidin-2-one) 33 exhibited potent activity in MCF-7 human breast cancer cells with IC50 values of 0.075 µM and 0.095 µM, respectively, and demonstrated low toxicity in non-cancerous cells. Compound 32 also demonstrated significant antiproliferative activity at nanomolar concentrations in the triple-negative breast cancer cell line Hs578T (IC50 0.033 μM), together with potency in the invasive isogenic subclone Hs578Ts(i)8 (IC50 = 0.065 μM), while 33 was also effective in MDA-MB-231 cells (IC50 0.620 μM). Mechanistic studies demonstrated that 33 inhibited tubulin polymerisation, induced apoptosis in MCF-7 cells, and induced a downregulation in the expression of anti-apoptotic Bcl2 and survivin with corresponding upregulation in the expression of pro-apoptotic Bax. In silico studies indicated the interaction of the compounds with the colchicine-binding site, demonstrating the potential for further developing novel cancer therapeutics as microtubule-targeting agents.


Introduction
Breast cancer is the most prevalent cancer diagnosed in women, with 2.3 million women diagnosed with breast cancer and 685,000 deaths globally in 2020 [1,2], and is recognised as the leading cause of cancer-related deaths in women. Many successful anticancer agents with different biological targets are used clinically to treat breast cancer [3,4]. Approximately 70-80% of breast cancers are hormone-dependent; the majority are identified as estrogen receptor-positive (ER+) cancers which also express the progesterone receptor (ER+/PR+). The estrogen receptor is directly targeted by selective estrogen receptor modulators (e.g., tamoxifen) and selective estrogen receptor degraders (e.g., fulvestrant). The aromatase inhibitors anastrozole, letrozole and exemestane block a key step in estrogen biosynthesis and are effective clinical adjuvant therapies. Triple-negative breast cancer (TNBC) is an aggressive breast cancer subtype which lacks the estrogen receptor (ER)

Compound
Structure X-ray Representation 18 resonates much further downfield compared to C4 due to the electron-withdrawing effect of the 3-fluoro atom (δ 95.2 ppm and 58.8 ppm, respectively) as an apparent double doublet with carbon-fluorine coupling constant values of 75 and 26 Hz, respectively. The carbonyl carbon (C2) of the β-lactam ring for 32 resonates as a doublet also coupling to the 3-fluoro substituent with a J value of 31 Hz. The characteristic 19 F NMR spectrum confirmed the presence of the 3-fluoro substituent in compound 31 at δ −203.7 and 32 at δ −203.6. Table 1. X-ray crystal structures of imines 18 and 23. 18

23
ORTEP representation of the X-ray crystal structures of imines 18 and 23 with 50% thermal ellipsoids with heteroatoms numbered. The 3,3-difluoro-β-lactams 36-41 and 43-45 were similarly prepared by reaction of ethyl bromodifluoroacetate with the imines 16-20 and 22-25, respectively, in 13-65% yield (Scheme 1). Deprotection of the TBDMS ether 41 with TBAF afforded the phenol 42 in 21% yield. In the 1 H NMR spectrum of 3,3-difluoro compound 37, the H4 signal was observed as a double doublet δ 5.29 ppm (J = 12.20 Hz and 1.64 Hz). In the 13 C NMR spectrum of resonates much further downfield compared to C4 due to the electron-withdrawing effect of the 3-fluoro atom (δ 95.2 ppm and 58.8 ppm, respectively) as an apparent double doublet with carbon-fluorine coupling constant values of 75 and 26 Hz, respectively. The carbonyl carbon (C2) of the β-lactam ring for 32 resonates as a doublet also coupling to the 3-fluoro substituent with a J value of 31 Hz. The characteristic 19 F NMR spectrum confirmed the presence of the 3-fluoro substituent in compound 31 at δ −203.7 and 32 at δ −203.6. Table 1. X-ray crystal structures of imines 18 and 23. 18

23
ORTEP representation of the X-ray crystal structures of imines 18 and 23 with 50% thermal ellipsoids with heteroatoms numbered. The 3,3-difluoro-β-lactams 36-41 and 43-45 were similarly prepared by reaction of ethyl bromodifluoroacetate with the imines 16-20 and 22-25, respectively, in 13-65% yield (Scheme 1). Deprotection of the TBDMS ether 41 with TBAF afforded the phenol 42 in 21% yield. In the 1 H NMR spectrum of 3,3-difluoro compound 37, the H4 signal was observed as a double doublet δ 5.29 ppm (J = 12.20 Hz and 1.64 Hz). In the 13 C NMR spectrum of 23 spectrum for both 31 and 32, at C2, C3 and C4 of the β-lactam ring. For compound 32, C3 resonates much further downfield compared to C4 due to the electron-withdrawing effect of the 3-fluoro atom (δ 95.2 ppm and 58.8 ppm, respectively) as an apparent double doublet with carbon-fluorine coupling constant values of 75 and 26 Hz, respectively. The carbonyl carbon (C2) of the β-lactam ring for 32 resonates as a doublet also coupling to the 3-fluoro substituent with a J value of 31 Hz. The characteristic 19 F NMR spectrum confirmed the presence of the 3-fluoro substituent in compound 31 at δ −203.7 and 32 at δ −203.6. Table 1. X-ray crystal structures of imines 18 and 23. 18

23
ORTEP representation of the X-ray crystal structures of imines 18 and 23 with 50% thermal ellipsoids with heteroatoms numbered. The 3,3-difluoro-β-lactams 36-41 and 43-45 were similarly prepared by reaction of ethyl bromodifluoroacetate with the imines 16-20 and 22-25, respectively, in 13-65% yield (Scheme 1). Deprotection of the TBDMS ether 41 with TBAF afforded the phenol 42 in 21% yield. In the 1 H NMR spectrum of 3,3-difluoro compound 37, the H4 signal was observed as a double doublet δ 5.29 ppm (J = 12.20 Hz and 1.64 Hz). In the 13 C NMR spectrum of spectrum for both 31 and 32, at C2, C3 and C4 of the β-lactam ring. For compound 32, C3 resonates much further downfield compared to C4 due to the electron-withdrawing effect of the 3-fluoro atom (δ 95.2 ppm and 58.8 ppm, respectively) as an apparent double doublet with carbon-fluorine coupling constant values of 75 and 26 Hz, respectively. The carbonyl carbon (C2) of the β-lactam ring for 32 resonates as a doublet also coupling to the 3-fluoro substituent with a J value of 31 Hz. The characteristic 19 F NMR spectrum confirmed the presence of the 3-fluoro substituent in compound 31 at δ −203.7 and 32 at δ −203.6.

23
ORTEP representation of the X-ray crystal structures of imines 18 and 23 with 50% thermal ellipsoids with heteroatoms numbered. The 3,3-difluoro-β-lactams 36-41 and 43-45 were similarly prepared by reaction of ethyl bromodifluoroacetate with the imines 16-20 and 22-25, respectively, in 13-65% yield (Scheme 1). Deprotection of the TBDMS ether 41 with TBAF afforded the phenol 42 in 21% yield. In the 1 H NMR spectrum of 3,3-difluoro compound 37, the H4 signal was observed as a double doublet δ 5.29 ppm (J = 12.20 Hz and 1.64 Hz). In the 13 C NMR spectrum of ORTEP representation of the X-ray crystal structures of imines 18 and 23 with 50% thermal ellipsoids with heteroatoms numbered. The 3-fluoro-β-lactams 26-31 and 33-35 were obtained by microwave-assisted Reformatsky reaction of the imines 16-21 and 23-25, respectively, with ethyl bromofluoroacetate in a short reaction time (30 min) in moderate yield (6-58%) after purification via flash column chromatography (Scheme 1). Deprotection of the TBDMS ether 31 with tetrabutylammonium fluoride (TBAF) afforded the phenol 32 in 18% yield. For compounds 26-35, the β-lactam carbonyl group was confirmed from the IR spectrum (ν 1736-1762 cm −1 ). In the 1 H NMR spectrum of compounds 31 and 32, H-3 was observed at 5.14 ppm and 5.11 ppm, respectively; geminal coupling with the 3-fluoro substituent of the β-lactam ring to H3 was demonstrated with 3 J fluorine H-F coupling constants of 47 Hz and 48 Hz for 31 and 32, respectively. Due to 2 J fluorine coupling, the typical β-lactam doublet for H3 resonates as an apparent double doublet with a second 3 J coupling of 1.7 Hz and 2.6 Hz for H 3 of 31 and 32, respectively, to the adjacent H 4 of the β-lactam ring. The characteristic coupling constant indicates that the trans isomer is isolated exclusively in this reaction for the series of compounds. Interestingly, we previously obtained a 2:1 cis/trans mixture of 3-methylazetidin-2-ones by reaction of ethyl-2-bromopropionate with the imine under similar conditions [57]. For both compounds 31 and 32, H 4 resonates also as an apparent double doublet with 3 J coupling to both the 3-fluoro substituent and adjacent trans H 3 . 3 J fluorine coupling constant values for H 4 are narrower than values for geminal fluorine coupling observed for H 3 with values of 25 and 26 Hz for 31 and 32, respectively. Additionally, interesting carbon-fluorine coupling is observed on the 13 C spectrum for both 31 and 32, at C 2 , C 3 and C 4 of the β-lactam ring. For compound 32, C 3 resonates much further downfield compared to C 4 due to the electron-withdrawing effect of the 3-fluoro atom (δ 95.2 ppm and 58.8 ppm, respectively) as an apparent double doublet with carbon-fluorine coupling constant values of 75 and 26 Hz, respectively. The carbonyl carbon (C 2 ) of the β-lactam ring for 32 resonates as a doublet also coupling to the 3-fluoro substituent with a J value of 31 Hz. The characteristic 19 F NMR spectrum confirmed the presence of the 3-fluoro substituent in compound 31 at δ −203.7 and 32 at δ −203.6. The 3,3-difluoro-β-lactams 36-41 and 43-45 were similarly prepared by reaction of ethyl bromodifluoroacetate with the imines 16-20 and 22-25, respectively, in 13-65% yield (Scheme 1). Deprotection of the TBDMS ether 41 with TBAF afforded the phenol 42 in 21% yield. In the 1 H NMR spectrum of 3,3-difluoro compound 37, the H4 signal was observed as a double doublet δ 5.29 ppm (J = 12.20 Hz and 1.64 Hz). In the 13 C NMR spectrum of compound 37, the difluoro-substituted C3 was observed at δ 121.47 ppm (further downfield compared with C-3 of monofluoro-substituted β-lactam, e.g., for compound 31 where C-3 is found at δ 95.2 ppm), and while the resonance of C4 occurred at δ 63.59 ppm. All products were obtained as racemic mixtures, with one enantiomer illustrated (Scheme 1).

X-ray Structural Study for 3-Fluoro and 3,3-Difluoro-β-lactams 33 and 43
X-ray crystallographic analysis of the 3-fluoro β-lactam 33 and 3,3-difluoro β-lactam 43 confirmed the stereochemical assignments (Tables 2-4). The β-lactam ring in both compounds 33 and 43 is observed with a conformationally restricted scaffold for the planar aryl rings A and B usually required for characteristic interaction with the colchicine-binding site of tubulin. As shown in Table 2, the X-ray structures demonstrated a rigid configuration for both β-lactams with rings A and B not coplanar. The torsional angles (ring A/B) observed for compounds 33 and 43 are 62.3 • and −76.6 • respectively (Table 4) indicating that compound 33 is more comparable with the related ring A/B torsional angles observed for colchicine [19] and CA-4 [66,67] reported as 53 • and 55 • respectively. The torsional angle value for ring B and the 3-fluoro group of compound 33 is -119.43 • which is similar to our previously reported 3-hydroxy β-lactams (117.0 • ) and 3-chloro β-lactams (114.4 • ) [33] ( Table 4). In contrast, the torsional angle value increased to 130.71 • for 3,3-difluoro β-lactam 43 suggesting that this may modulate the tubulin activity observed for this compound. The β-lactam C=O bond lengths of 1.2134 (15) Å and 1.2012 (19) Å are observed for compounds 33 and 43 respectively, in agreement with carbonyl bond lengths reported for monocyclic β-lactams [68]. The strained β-lactam four-membered ring differs from a normal amide resulting in decreased amide resonance. Monocyclic β-lactams are found to contain a longer C-2/N amide bond length (1.35-1.38 Å) when compared with normal amide bond length of 1.33 Å; also a shorter C=O bond length (1.21-1.23 Å) is observed when compared with a standard amide bond length of 1.24 Å [68]. The data obtained for compounds 33 and 43 show the β-lactam C=O bonds were within the expected range for a monocyclic β-lactam, as were the C-2/N(amide) bond lengths [(1.3706 (15) Å and 1.3724 (19) Å respectively]. The C-2/C-3 bond lengths of 1.5339 (17) Å and 1.534 (2) Å were in the expected range of 1.52-1.55 Å for a β-lactam, while the N/C-4 bond lengths (1.4851 (15) Å and 1.4851 (18) Å respectively were also within the expected range of 1.49-1.51 Å [68,69]. (Numbers in parentheses refer to the second crystallographically independent molecule in the asymmetric unit).

Stability Study for β-Lactams 33 and 39
The β-lactam ring is known to undergo hydrolysis/degradation depending on the substituent type on the ring. For example, degradation of monocyclic β-lactam antibiotic aztreonam increased with elevated humidity, temperature and acidic pH buffer [70]. Ezetimibe, a monocyclic β-lactam drug used to reduce blood cholesterol, undergoes degradation and hydrolysis more rapidly in neutral and basic pH conditions than in acidic pH [71]. The stability of representative compounds 33 and 39 was also determined at pH 4, pH 7.4 (plasma) and pH 9 (intestine) by HPLC. Interestingly, compound 33 bearing a fluoro substituent at C-3 of β-lactam together with a 3-fluoro on ring B is stable at buffered systems pH 4 and pH 7.4 with 76% and 87%, respectively, remaining at 24 h, while 26% remained at pH 9. The electron-withdrawing effect of the 3-fluoro substituent may result in decreasing stability of the β-lactam ring in response to acid hydrolysis. For the 3,3-difluoro compound 39, 50% remained after 24 h at pH 7.4, 39% at pH 4, while the compound was less stable at pH 9 with 19% remaining at 24 h. Compounds 33 and 39 were further examined under degradation conditions; 33 was found to be stable with 90%, 88%, 60%, 65% and 65% of the compound remaining after 4 h treatment in heat (60 • C), UV light, acidic (0.1 M HCl), alkaline (0.1M NaOH) and oxidative (3% H 2 O 2 ) conditions, respectively. The 3,3-difluoro compound 39 was slightly less stable in these conditions, with 82%, 61%, 54%, 60% and 50% of the compound remaining, respectively (see Supplementary Materials, Table S7).
; e Ring C = β-lactam ring; * = 2 independent molecules in the asymmetric unit. Each angle given but only the first atom numbering scheme is outlined above.

Predicted Physicochemical and ADME Properties
Consideration of physicochemical and pharmacokinetic properties of active molecules is useful in the early stages of drug discovery to determine the potential of compounds for further development. The physicochemical properties and metabolic stability predicted for the panel of 3-fluoro and 3,3-difluoro-β-lactams 26-45 were compiled to assess the relevant drug properties of the series (see Supplementary Materials for Tier 1 profiling screen, Tables S1 and S2). Pipeline Pilot Professional [72] was used to calculate the relevant physicochemical and pharmacokinetic properties. The synthesised compounds followed Lipinski and Veber rules with a molecular weight less than 500 Da (within the range of 361-458 Da), ≤10 hydrogen bond acceptors, ≤5 hydrogen bond donors and ≤10 rotatable bonds. The logP for all synthesised compounds was determined to be less than 5 (Supplementary Materials, Table S2) and was in the range 2.68-4.00. The topological polar surface area (TPSA), indicating the ability of the compound to form hydrogen bonds and to permeate cells (indicating good intestinal absorption), was calculated to be in the range 57.23-77.46 Å 2 and within the acceptable limit of ≤140 Å 2 . All compounds were predicted to have good passive gastrointestinal absorption properties, high blood-brain barrier (BBB) absorption levels and good plasma protein binding properties (>90%) and were not predicted to inhibit CYP2D6. The synthesised compounds are predicted to be un-ionised at physiological pH. The theoretical pKaH values of phenolic compound 32 calculated with Marvin are 9.82 (phenolic OH) and 11.47 (CH-F), while the corresponding values for compounds 33 and 42 are 11.46 (CH-F) and 9.82 (CH-F), respectively. The compounds are predicted to have low aqueous solubility, with the exception of phenolic compound 32 which is predicted to have good aqueous solubility (logSw = −3.8020 mol/L) (see Supplementary Materials, Table S1). These compounds were soluble in EtOH and DMSO for biological evaluation; improved water solubility may be achieved using phosphate esters, as for CA-4 [73]. In addition, the compounds did not signal an alert for pan-assay interference compounds (PAINS) [74]. As the compounds are predicted to demonstrate good drug-like parameters and bioavailability [75], we proceeded with further biochemical studies to examine their mechanism of action.
The antiproliferative effect of 3,3-difluoro β-lactams 36-39 containing different para substituents on ring B is displayed in Figure 3B. All compounds in this series exhibited much weaker activity with 59-67% viable cells at 10 µM and >90% at 1 µM compared to the corresponding 3-fluoro β-lactam derivatives (26)(27)(28)(29). In a similar trend, 3,3-difluoro compounds 40 and 42-45 containing various meta substituents on ring B showed poorer antiproliferative activity with >60% cell viability at 10 µM and >80% cell viability at 1 µM compared to their corresponding 3-fluoro β-lactam derivatives (30, 32-35) ( Figure 3B). The exception in this series was compound 42 containing meta-hydroxy substituent on ring B, which elicited potent anticancer effects at both concentrations (1 and 10 µM) with 17% and 37% viable MCF-7 cells remaining, respectively. The results suggested that the introduction of an additional fluorine substituent at C-3 of the β-lactam ring significantly reduced the cell growth inhibition activity for most of the compounds and affects the interaction of the β-lactam with the target tubulin-binding site.   Table S4). 3-Fluoro β-lactam 33 exhibited notable antiproliferative activity against the ER-positive MCF-7 breast cancer cells with an IC 50 value of 0.095 µM ( Figure 4A). The positive control CA-4 gave an IC 50 value of 0.0035 µM in the MCF-7 cell line, which is in good agreement with the reported values for CA-4 in the MCF-7 human breast cancer cell lines [76]. Removal of the ring B 3-fluoro substituent as in compound 26 resulted in a reduction in potency with IC 50 = 0.312 µM. The 3-fluoro-β-lactam (ring B 3-hydroxy) 32 demonstrated significant potency (IC 50 = 0.075 µM) compared to its corresponding 3,3-difluoro compound 42 which demonstrated a 4-fold reduction in potency (IC 50 = 0.321 µM), suggesting that the hydroxyl group contributes to antiproliferative activity but 3,3-difluoro substitution negatively impacts the activity. This effect was also demonstrated in the reduction in activity observed for compound 43 (IC 50 = 1.65 µM). It is interesting to see that the phenolic compound 32 with lower logP value (2.68) was more potent (IC 50 value = 0.075 µM) than the corresponding ring B 3-fluoro compound 33 (logP 3.10, IC 50 value = 0.095 µM); a similar trend was observed for the 3,3-difluoro compounds 42 (logP 2.85, IC 50 value = 0.321 µM) and 43 (logP 3.29, IC 50 value =1.65 µM), suggesting that the interaction of the phenolic group at the colchicinebinding site is more favourable than the fluorine for optimal tubulin effect. The introduction of the fluorine substituent at C-3 of the β-lactam ring resulted in antiproliferative activities for compounds 32 and 33 at nanomolar concentrations in MCF-7 breast cancer cells similar in potency to our previously reported 3-chloro, 3-vinyl and 3-methylazetidin-2-ones in MCF-7 breast cancer cells [32,33,57], thus indicating the importance of these substituents for the antiproliferative and antimitotic activity of these compounds.  [77,78]. Compound 32 was also evaluated in the triple-negative breast cancer cell line Hs578T and its isogenic subclone Hs578T(i)8 ( Figure 4C). Hs578Ts(i)8 cells have 2.5-fold more migratory capacity and 3-fold more invasive capacity through the extracellular matrix than the parental cell line (Hs578T) and have 30% more CD44+/CD24-/low cells and in vivo proliferation [79]. Compound 32 demonstrated potent antiproliferative activity in Hs578T cells (IC 50 0.033 ± 0.005 µM) and Hs578Ts(i)8 cells (IC 50 = 0.065 ± 0.003 µM) and compares well with CA-4 (IC 50 = 0.008 µM in Hs578T and 0.020 µM in Hs578Ts(i)8 cells). These results demonstrate the potential application of this compound as an anticancer agent in the inhibition of tumour invasion and angiogenesis, which are recognised features of tumour growth and metastasis in aggressive breast cancers. Additional evaluation of the compounds 26-45 in the colon cancer cell line SW-480 indicated moderate activity, with compounds 33, 42, 26 and 29 being the most potent with 62%, 56%, 58% and 60% cells remaining at 10 µM concentration (see Supplementary Materials, Figure S11). The cytotoxicity of the potent β-lactam 33 was investigated in the non-tumourigenic HEK-293T cell line (normal human embryonic kidney) at 72 h. As shown in Figure 4D, the cell viability of HEK-293T cells was notably higher than that observed for MCF-7 cells at 50, 10 and 1 µM concentrations of compound 33. For example, the cell viability of HEK cells at 1 µM was 88% in contrast to the cell viability of 36% for MCF-7 cells at 1 µM. The IC 50 value for compound 33 was greater than 50 µM in HEK-293T cells (0.095 µM in MCF-7 cells), demonstrating its selectivity towards cancer cells and the lack of toxicity in non-cancerous cells.
2.5.2. NCI 60 Cell Line Screening for β-Lactam Compounds 33, 37 and 43 The antiproliferative effects of selected compounds 33, 37 and 43 were initially evaluated in the NCI 60 cell line screen at 10 µM concentration [80] (see Supplementary Materials Table S3). Compounds 33, 37 and 43 showed mean growth inhibition of 73.34%, 58.95% and 56.43% at 10 µM over the 60 cell lines tested. Compound 33 demonstrated broad-spectrum activity against the nine panels of cell lines tested (leukaemia, CNS, melanoma, ovarian, renal, non-small-cell lung, colon, breast and prostate cancers). The most potent growth inhibition for all compounds 33, 37 and 43 was observed in the leukaemia panel with a mean growth inhibition of 90.6%, 91.4% and 86.3%, respectively. The activity in MCF-7 confirmed our results, with 87.2%, 82.8% and 84.9% growth inhibition for compounds 33, 37 and 43, respectively.
Compound 33 was next selected for evaluation in the NCI 60 cell line five-dose screen. The GI 50 (50% growth inhibition), TGI (total growth inhibition) and LC 50 (50% lethal concentration) were determined in the NCI panel of 60 cell lines, using the sulphorhodamine B (SRB) protein assay ( Table 5). The GI 50 value provides the growth inhibition of the selected compound, while the cytotoxic effect is evaluated in the LC 50 value. Compound 33 showed good potency in all leukaemia, breast cancer, ovarian, colon and prostate cell lines and excellent antiproliferative activity at submicromolar concentrations in all cell lines except for melanoma cell lines UACC-257 and CNS SNB-75 (Table 5 and Supplementary Materials Table S3). The . The compound also displayed good antiproliferative activity in the glioblastoma cell lines U251 (GI 50 = 0.0236 µM) and SF-539 (GI 50 = 0.0280 µM). Glioblastoma is an aggressive and fast-growing brain tumour with a median survival time of 9-16 months from diagnosis, and these tumours quickly evolve resistance to temozolomide chemotherapy, which is the only FDA-approved treatment [81]. Additionally, the cytotoxicity of compound 33 (determined as the LC 50 value) was greater than 100 µM in all cell lines (except COLO 205 (LC 50 Table 6) indicated a significant therapeutic window between the concentration required for inhibition of growth of the cancer cells and the concentration of the compound that is lethal to the cells, demonstrating the potential for further development. Further mechanistic and cellular studies were carried out and are described below.
To gain insight into the potential cellular target and molecular mechanism correlating with the cytotoxic effect of 33, NCI COMPARE analysis was carried out [82]. The NCI five-dose cell growth data of 33 were compared with the NCI databases (Standard and Synthetic). The COMPARE algorithm ranks the compared compounds from the NCI databases (Standard and Synthetic, >55,000 compounds) to the seed compound 33 (NCI ref 792959), using Pearson correlation coefficients for the correlation. A correlation coefficient of greater than 0.8 indicates a strong correlation and suggests that the seed compound may have a mechanism of action similar to that of the highly correlated compounds [83,84]. The comparison with the NCI standard database of clinically used drugs resulted in the top-ranked compounds of macbecin II (0.53, Hsp90 inhibitor) and vincristine (0.5, tubulin inhibitor), with values below the acceptable value for useful prediction of the mechanism of action (see Supplementary Materials Table S5, Figure S12). However, the COMPARE analysis using the NCI Synthetic Database of compounds which have known and unknown mechanisms of action resulted in the identification of 19 compounds with correlations in the range of 0.85-0.75 based on GI 50 values (see Supplementary Materials Table S6 for details, together with the structures of the compounds identified). Of these, 12 were identified as acting as microtubule-targeting agents with tubulin binding activity, including the highest-ranking compounds which target the colchicine-binding site of tubulin. Compound 33 was designed to target microtubules, so further investigations were carried out to confirm this mechanism of action.   Based on the cell viability results for compound 33, combined with induction of apoptosis, further effects of 33 on the expression of members of the Bcl-2 apoptosis regulatory protein family were next investigated. The Bcl-2 family of apoptosis regulatory proteins includes anti-apoptotic and pro-apoptotic members and is the best-characterised family of proteins involved in the regulation of apoptotic cell death [85]. The pro-apoptotic protein Bax is an important regulator of the intrinsic or mitochondrial apoptosis pathway and triggers the release of caspases, while the anti-apoptotic protein Bcl-2 prevents apoptosis by sequestering caspases or by inhibiting the release of the mitochondrial apoptogenic factors cytochrome c and apoptosis-inducing factor (AIF) into the cytoplasm [86,87]. The selective Bcl-2 inhibitor venetoclax was developed as a BH3-mimetic that binds to the pro-survival protein Bcl-2 and inhibits its ability to bind Bax or Bak. Venetoclax has been approved for clinical use in the treatment of chronic lymphocytic leukaemia (CLL) [86]. MCF-7 cells were treated with compound 33 (0.05, 0.1 and 0.5 µM) for 48 and 72 h as shown in Figure 6. Compound 33 induced downregulation of Bcl-2 expression with corresponding upregulation of Bax at 48 h in a dose-dependent manner. This effect was further increased at 72 h.
Survivin is a member of the inhibitor of apoptosis (IAP) protein family. It is an essential anti-apoptotic protein marker that is overexpressed in most tumour cells and is associated with a poor clinical outcome. Survivin inhibits caspase activation, which is used as an indicator of apoptosis cascades [88][89][90]. Survivin selective inhibitor molecules have been identified as cancer therapeutics, including survivin-partner protein interaction inhibitors, survivin homodimerisation inhibitors, survivin gene transcription inhibitors and survivin mRNA inhibitors [91]. We examined the effect of a range of concentrations of compound 33 (0.05, 0.1 and 0.5 µM) on the level of survivin expressed. Compound 33 caused a downregulation of the expression of survivin in a dose-and time-dependent manner, as demonstrated by Western blotting (Figure 6), confirming the pro-apoptotic effect of 33 in MCF-7 cancer cells. The upregulation of the pro-apoptotic protein Bax and the downregulation of the anti-apoptotic proteins Bcl-2 and survivin support the pro-apoptotic mechanism of action suggested for compound 33, also indicated from the Annexin V/PI flow cytometric analysis.

Effect of Compound 33 on Tubulin Polymerisation
The effect of representative 3-fluoro β-lactam CA-4 analogue compound 33, which exhibited potent anticancer effects in vitro, on the polymerisation of the purified tubulin protein was examined. Tubulin polymerisation was investigated using a turbidimetric assay which determines the light scattering by microtubules (absorbance at 340 nm) that is proportional to the microtubule polymer concentration. Paclitaxel (10 µM), which stabilises the tubulin when compared to the vehicle control (DMSO), was the positive control. Tubulin polymerisation results obtained for 33 showed a 3.5-fold reduction in the V max (maximum rate of reaction) at 10 µM compared to the vehicle with an increase to 4-fold reduction at the higher concentration of 30 µM (Figure 7), whereas in our previous work we demonstrated that CA-4 induced a 6.3-fold reduction at 10 µM [33]. This result indicated that tubulin is the molecular target of the antiproliferative 3-fluoro β-lactam compound 33 as reported previously for related heterocyclic CA-4 analogues [33]. The tubulin-targeting effect of compound 33 on the microtubule network of MCF-7 cells was further evaluated by immunofluorescence studies to identify the cellular effects that are relevant to its mechanism of action by inhibiting the polymerisation of tubulin. MCF-7 cells displayed a well-organised microtubular network in the control cells (0.1% ethanol (v/v)) ( Figure 8). Clearly, significant stabilisation of microtubules in paclitaxel (1.0 µM)-treated cells was observed, while CA-4-treated cells (0.01 µM, positive control) demonstrated depolymerisation and destabilisation of the cell membrane of microtubules. Treatment of MCF-7 cells with compound 33 (0.1, 0.5 and 1.0 µM) induced cell rounding and distinct abnormalities of the spindle formation as well as a loss of the structured tubulin that impacts the microtubule network structure in a dose-dependent manner. These results confirmed our findings that compound 33 acts by destabilising microtubules, supporting the results from the in vitro tubulin polymerisation assay.

Molecular Modelling Study for Compounds 32, 33, 42 and 43
A series of molecular docking calculations using MOE 2020.09 were undertaken on both enantiomeric pairs of the 3-fluoro-β-lactam compounds 32, 33, 42 and 43, using the tubulin co-crystallised with DAMA-colchicine ligand (X-ray crystal structure PDB entry 1SA0) [19] (Figure 9). It is evident from 1 H NMR spectroscopy that only the trans isomers of the β-lactam compounds were obtained, and therefore the 3S/4S and 3R/4R enantiomeric pairs were selected for the modelling analysis. For the compounds with two stereogenic centres, in all cases the S,S enantiomers were found to be more highly ranked than the corresponding R,R enantiomeric pair and so will only be discussed in this study. All selected compounds overlaid their B-rings on the C-ring of DAMA-colchicine (resulting in the characteristic hydrogen bond acceptor interactions with Lys352 residue). The 3,4,5trimethoxyphenyl substituted A-rings occupied the colchicine-binding site, and the 3-fluoro and 3,3-difluoro substituents were located in an open accessible region of the β-tubulin binding site at the monomer interface and did not form interactions with adjacent residues.   Table S8. When the conformers were generated with OMEGA [92,93] and docking was run with FRED [94], a similar preference was obtained for S,S over R,R enantiomers for compounds 32 and 33. Docking studies are not always ideal when studying changes in cellular efficacy associated with the C-3 ring B fluoro/hydroxyl substitutions. The best ranked enantiomer, 32 (3S,4S), was also the most active analogue in MCF-7 cells with IC 50 of 0.075 µM ( Figure 9A). The 3-hydroxy substituent on the 4-phenyl ring (ring B) co-located very well with the C-ring of DAMAcolchicine, with the oxygen positioned to act as a hydrogen bond acceptor with Lys352. The 3,4,5-trimethoxyphenyl ring (ring A) interacted slightly deeper in the binding pocket than that of DAMA-colchicine. The β-lactam carbonyl group potentially formed a hydrogen bond acceptor interaction with a backbone hydrogen atom of Ala250. The 3S,4S enantiomer of 33 also presented a similar pose in the binding site ( Figure 9B). The ring A, 3,4,5trimethoxyphenyl groups of the compounds illustrated made favourable van der Waals contacts in the lower sub-pocket defined by residues Val β318 and Cys β241. Figure 9B shows the ring B 3-fluoro substituent of compound 33 occupying the same location as the phenolic hydroxyl group in the 32, acting as a hydrogen bond acceptor with Lys352 which could play a role in stabilisation of protein-drug conformation but which is weaker than the hydrogen bond formed between Lys352 and the hydroxyl group of 32, which may explain the decrease in cellular efficacy of this compound.
Due to the small volume of the 3,3-difluoro substituents at position 3 of the β-lactam ring, different orientations of these groups did not affect binding, which is evident from the near equal docking scores given to the 3,3-difluoro R and S enantiomers of 42 and 43. However, the in vitro antiproliferative activity of the most potent 3-fluoro-β-lactams 32 and 33 of the series is significantly superior when compared with the corresponding 3,3-difluoro-β-lactams 42 and 43 compounds, despite having similar docking scores. Both of the enantiomers of 42 and 43 orientated the 3,4,5-trimethoxyphenyl ring A to overlap with the C-ring of DAMA-colchicine and achieve overlap of the 3,4,5-trimethoxyphenyl groups of colchicine ( Figure 9C,D). It is possible that both sets of compounds (monofluoro 32 and 33 and difluoro 42 and 43) may have slightly different target protein profiles or physicochemical properties such as solubility and cellular permeability. The best ranked binding pose of each compound examined in the study is illustrated in Figure 9, showing the shared binding mode for the selected analogues studied.

Chemistry
Melting points (uncorrected) were measured on a Gallenkamp apparatus. Infrared (IR) spectra were recorded on a Perkin Elmer FT-IR Paragon 1000 spectrometer. 1 H, 19 F and 13 C nuclear magnetic resonance spectra (NMR) were recorded on a Bruker DPX 400 spectrometer (400.13 MHz, 1 H; 100.61 MHz, 13 C; 376 MHz, 19 F) in CDCl 3 (internal standard tetramethylsilane (TMS)) at 27 • C. 1 H-NMR spectra were assigned relative to the TMS peak at 0.00 ppm and 13 C-NMR spectra were assigned relative to the middle CDCl 3 peak at 77. 1525 binary HPLC pump and Waters 717 plus Autosampler with Varian Pursuit XRs C18 reverse phase 150 × 4.6 mm chromatography column and UV detection at 254 nm. The mobile phase was acetonitrile (70%):water (30%) over 10 min with a flow rate of 1 mL/min. All synthesised products were homogeneous on TLC when isolated; the purity of the biologically tested compounds was confirmed by HPLC (≥95%). Microwave experiments were carried out using a Biotage Discover SP4 and CEM microwave synthesisers using standard power setting (maximum power 300 watts) unless otherwise stated.

General Method II: Preparation of β-Lactams 26-31, 33-41 and 43-45
Activated zinc powder (9 mmol) with trimethylchlorosilane (7 mmol) in anhydrous benzene (4 mL) was heated for 15 min at 40 • C and then at 100 • C for 2 min in a microwave reactor. The mixture was then cooled and the selected imine (2 mmol) and ethyl bromoacetate (5 mmol) were added. The mixture was heated in the microwave reactor for 30 min at 100 • C. The reaction mixture was filtered through Celite; diluted with DCM (30 mL); and washed with ammonium chloride solution (20 mL, satd.) and ammonium hydroxide (20 mL, 25%), HCl (10%, 40 mL) and water (40 mL). The organic phase was dried (anhydrous Na 2 SO 4 ) and the solvent was evaporated in vacuo. Isolation of the crude product was achieved by flash column chromatography over silica gel (eluent: hexane:ethyl acetate gradient).  13   Compound 42 was prepared from the TBDMS-protected β-lactam 41 using the procedure described above for 20 to obtain the phenolic product as a yellow oil; yield: 21%, purity (HPLC): 100%. IR ν max (ATR): 3426. 75

Biochemical Evaluation of Compounds
The biochemical assays were performed in triplicate on at least three independent occasions for the determination of mean values reported. Hs578T8i cells were cultured in DMEM with GlutaMAX-I, with the same supplement in the absence of non-essential amino acids. HEK-293T cells (normal epithelial embryonic kidney cells) were cultured in DMEM with GlutaMAX-I in the absence of non-essential amino acids. The SW-480 cells were cultured in DMEM with GlutaMAX-I, with the same supplement in the absence of non-essential amino acids. All media contained 100 U/mL penicillin and 100 µg/mL streptomycin. The cells were seeded into pre-warmed complete medium (10 × 10 4 cells/mL) and maintained at a cell density of 2-19 × 10 5 cells /mL by the addition of fresh media. The cell number was monitored using a haemocytometer. Cells were maintained at 37 • C in 5% CO 2 in a humidified incubator. The cells were sub-cultured 3 times weekly by trypsinisation using TrypLE Express (1X).

Cell Viability Assay
Cells were seeded at a density of 5 × Cell proliferation for MCF-7 cells was analysed with the AlamarBlue assay (Invitrogen Corp.) as previously described by us [33]. After 72 h, AlamarBlue (10% (v/v)) was added to each well, and plates were incubated for 3-5 h at 37 • C in the dark. Fluorescence results were obtained with a 96-well fluorimeter with excitation at 530 nm and emission at 590 nm with results recorded as percentage viability relative to vehicle control (100%). Dose-response curves were obtained and IC 50 values (concentration of drug resulting in 50% reduction in cell survival) were calculated using Prism (GraphPad Software, Inc., La Jolla, CA, USA).

Annexin V/PI Apoptotic Assay
Apoptotic cell death was monitored by flow cytometry using Annexin V and propidium iodide (PI) as previously described by us [33]. MCF-7 cells were seeded in 6-well plates at a density of 1 × 10 5 cells/mL and treated with either vehicle (0.1% (v/v) EtOH) or β-lactam compound 33 at concentrations 0.1 and 0.5 µM for 48 h. Cells were harvested and prepared for flow cytometric analysis. Cells were washed in 1X binding buffer (20X binding buffer: 0.1 M HEPES, pH 7.4; 1.4 M NaCl; 25 mM CaCl 2 diluted in dH 2 O) and incubated in the dark for 30 min on ice in Annexin V-containing binding buffer [1:100]. Cells were then washed in binding buffer and then re-suspended in PI-containing binding buffer (1:1000). Samples were then analysed with the BD Accuri flow cytometer and Prism software. The resulting cell populations were identified as Annexin V-and PI-negative (Q4, healthy cells), Annexin V-positive and PI-negative (Q3, early apoptosis), Annexin V-and PI-positive (Q2, late apoptosis) and Annexin V-negative and PI-positive (Q1, necrosis). The positive control for induction of cell death was paclitaxel.

Immunofluorescence Microscopy
Confocal microscopy was used to determine the effects of novel compound 33 and standard drugs on the MCF-7 cytoskeleton according to the protocols we previously described [33]. For the immunofluorescence study, MCF-7 cells were seeded at 1 × 10 5 cells/mL using eight chamber glass slides (BD Biosciences). The cells were treated with vehicle control (1% ethanol (v/v)), CA-4 (0.01 µM), paclitaxel (1 µM) and 33 (0.1 µM, 0.5 µM and 1 µM) for 16 h. The cells were gently washed in PBS, fixed with 4% paraformaldehyde in PBS (20 min) and permeabilised in 0.5% Triton X-100. After further washes in PBS (containing 0.1% Tween (PBST)), cells were blocked with 5% bovine serum albumin which was diluted with PBST. Cells were incubated with mouse monoclonal anti-α-tubulin-FITC antibody (clone DM1A) (Sigma) (1:100) at 20 • C for 2 h and then washed in PBST and incubated with Alexa Fluor 488 dye (1:500) for 1 h at 20 • C. Cells were washed in PBST and mounted in Ultra Cruz Mounting Media (Santa Cruz Biotechnology, Santa Cruz, CA, USA) containing 4,6-diamino-2-phenolindol dihydrochloride (DAPI). The images were captured by Leica SP8 confocal microscopy with Leica Application Suite X software. Experiments were performed on three independent occasions and images were collected on the same day using identical parameters.

Evaluation of Expression Levels of Anti-Apoptotic Protein Bcl-2 and Pro-Apoptotic Proteins Bax and Survivin
Western blot analysis was performed according to the protocols we previously described [33]. MCF-7 cells were seeded at a density of 1 × 10 5 cells/flask in T25 flasks. After 48 h, whole cell lysates were prepared from untreated cells or cells treated with vehicle control (EtOH, 0.1% v/v) or cells treated with compound 33 (0.05 µM, 0.1 µM and 0.5 µM). MCF-7 cells were harvested in RIPA buffer which was supplemented with protease inhibitors (Roche Diagnostics, Rotkreuz, Switzerland), phosphatase inhibitor cocktail 2 (Sigma-Aldrich, Arklow, Ireland) and phosphatase inhibitor cocktail 3 (Sigma-Aldrich, Arklow, Ireland). Equal amounts of protein (as determined by a BCA assay) were resolved by SDS-PAGE (12%) followed by transfer to PVDF membranes. Membranes were blocked in 5% bovine serum albumin/TBST for 1 h and incubated in the relevant primary antibodies at 4 • C overnight, washed with TBST, and incubated in horseradish peroxidase conjugated secondary antibody for 1 h at 20 • C. After washing, Western blot analysis was performed with antibodies directed against Bcl-2 (1:500) (Millipore), Bax (1:1000) (Millipore) or survivin (1:1000) (Millipore) and followed by incubation with a horseradish peroxidase conjugated anti-mouse antibody (1:2000) (Promega, Madison, WI, USA). The blots were probed with anti-GAPDH antibody (1:5000) (Millipore) to confirm equal loading of protein.
Proteins were detected using enhanced chemiluminescent Western blot detection (Clarity Western ECL substrate) (Bio-Rad) on the ChemiDoc MP System (Bio-Rad). Experiments were performed on three independent occasions.

Conclusions
Microtubule-targeting agents are the group of drugs frequently used for cancer treatment both in adults and children; they are effective in suppressing microtubule dynamics and inducing cell death via the mitochondrial intrinsic apoptosis pathway. Significant progress has been achieved in the discovery of targeted cancer therapies; however, it is recognised that resistance (demonstrated by both innate and acquired mechanisms) remains an issue for many clinically successful cancer drugs [32]. The β-lactam scaffold continues to attract much interest due to its utility as a versatile synthetic intermediate and also the many therapeutic applications of this heterocycle. We have now described the synthesis of a series of novel 3-fluoro and 3,3-difluoro-β-lactams which are designed as CA-4 analogues that show significant antiproliferative activity against the MCF-7 cell line. Introduction of fluorine can enhance the potency and target selectivity of a drug by affecting properties such as pKa, lipophilicity, hydrophobic interactions and membrane permeability. The mechanism of action of these compounds was investigated. Induction of apoptosis was confirmed for compound 33 using flow cytometric analysis of Annexin V/PI-stained cells. Additionally, β-lactam 33 was observed to inhibit tubulin polymerisation, causing disruption of tubulin network structure in MCF-7 cells, inducing a disorganised microtubule network with multinucleation effects as also observed for CA-4. The effects on the expression of the characteristic apoptosis-related proteins Bcl-2, Bax and survivin in MCF-7 cells on treatment with compound 33 were demonstrated with Western blot analysis and confirmed the pro-apoptotic action of the 3-fluoro β-lactam 33. The tubulin-targeting effects of compound 33 were demonstrated in a molecular modelling study suggesting interactions of the compound's A and B rings with the colchicine-binding site of β-tubulin, in a manner similar to that of CA-4. The 3-fluoro-β-lactams 32 and 33 demonstrated potential as lead microtubule-targeting molecules suitable for further preclinical anticancer drug development.

Conflicts of Interest:
The authors declare no conflict of interest.  TGI  Total growth inhibitory concentration  THF  Tetrahydrofuran  TLC  Thin layer chromatography  TNBC  Triple-negative breast cancer