`
`Santen/Asahi Glass Exhibit 2047
`Micro Labs v. Santen Pharm. and Asahi Glass
`IPR2017-01434
`
`
`
`FULL PAPER
`
`R. Taylor and .l. D. Dunitz
`
` ‘
`
`600
`500
`
`400
`
`300
`
`200
`
`
`100
`
` <.‘
`
`
`the relevance of elec-
`turn is stronger than NH3. However.
`tronegativity to hydrogen bond acceptor ability is less clear.
`Electronegativity is a measure of the tendency to attract elec—
`trons, not protons. Thus. covalently bonded fluorine is an ex—
`tremely weak base and, as such, may be expected to be an
`extremely weak proton acceptor, On the other hand, it is un-
`deniable that the best H-bond acceptor atoms (oxygen and
`nitrogen) are electronegative.
`In view of these problems. we have undertaken a new survey
`of hydrogen bonds involving covalently bound fluorine. We
`have focused on three issues. First. how commonly docs carbon-
`bound fluorine accept hydrogen bonds. and under what circum-
`stances? Secondly. what is the relevance (if any) of electronega-
`tivity to hydrogen—bond acceptor ability? Thirdly. what factors
`make a good hydrogen-bond acceptor? As our primary source
`of data, we have used the Cambridge Structural Database
`(CSD).[7] which contains the results of about 150000 small-
`moleculc crystal structure determinations In View of the sug~
`gested importance of fluorine as a hydrogen-bond acceptor in
`protein~ligand complexes (e.g. of elastase).[3l we have also ex-
`amined crystal structures taken from the Brookhaven Protein
`Data Bank (PDB),[B] despite their much lower precision.
`ln
`addition we have collated evidence from various published
`physicoehemical studies. Finally. we have made molecular or—
`bital calculations on model systems using ab initio intermolecu—
`lar perturbation theory (IMPTHm
`
`Methods
`
`All calculations were performed on a Sun SPARCstation 5 or a Silicon
`Graphics Indigo [2].
`
`Searches for H-Bonds in Small-Molecule Crystal Structures: Searches for
`H bonds were made with Version 5.09 ofthe CSD (April 1995), by using the
`nonbonded Search capabilities of the program QUEST3D [10]. Only inter-
`molecular contacts were considered. All searches were confined to error-free
`structures (according to the criteria of the CSD system) with crystallographic
`R factors of less than 10%. Contacts were only accepted as H bonds if the
`hydrogen-atom coordinates were in the CSD.
`Our first step was to define suitable geometric criteria for a H bond. For this.
`only organic structures were included (CSD bit screen 28 set to ZCrOinIBall—
`ing no metals presentvand elements As, Se. Tc also excluded). From data for
`H bonds of the types O—H~~~O=C, NAHWO:C, O H-"N(Ar), and
`NiH ‘
`‘
`> N(Ar) (O=C : any carbonyl group, N(Ar) : any aromatic nitro-
`gen acceptor), histograms of HwO and H ~ ~N Hibond distances were
`prepared with the VISTA package [10,11]. The results (Figure 1) show that
`nearly all of the H bonds have H .
`. -0 or H -
`'
`4 N distances less than 2.2 A.
`Since fluorine has a smaller van der Waals radius than either oxygen or
`nitrogen (12],
`it might seem reasonable to exclude C~ l-‘~-H7X contacts
`(X 2 ().l\) as possible ll bonds unless the F -
`~ -H distance is also less than
`2.2 A. In fact, we used a less severe criterion. namely, F -
`‘ -H <23 A, with the
`additional constraint that the F' '
`~ HHX angle must exceed 90". We are aware
`that even this relaxed distance criterion is liable to criticism. The sum of the
`van dcr Waals radii of fluorine and hydrogen lies between 2.5«27 A. depend,
`mg on which literature values are chosen [12,131 Some authors [14] consider
`that acceptor-“hydrogen contacts much longer than the sum of van (lei
`Waals radii may still be regarded as H bonds. However, any distance criteri-
`oni-indeed, any definition of hydrogen bonding is to some extent arbitrary.
`in the present case, we Wish to focus on cases in which covalently bonded
`fluorine unequivocally acts as a H—bond acceptor. hence our choice of a
`distance limit that is significantly shorter than the sum of van der Waals radii.
`as found in the typical H bonds involving 0 and N acceptors. With these
`geometric constraints. several CSD searches were made to determine the
`frequcncy with which H bonds to fluorine occur and to characterize individi
`
`- 0(N) H-bond distances in CSD crystal structures. Top
`-
`Fig. l. Histograms of H -
`leftzC:O-‘-H O;top right:C=O’--HiN;bottomleft:N(/\r)- vHrO;botton1
`right: N(Ar) .
`-
`4 lie N. Distances along horizontal axis in A.
`
`- H X contacts thus found were examined visually
`-
`ual examples. Short F -
`with the programs SYBYL [15] (Version 0.1) and PLUTO [10]. Visual inspec-
`tion is always called for because a short contact is not. in itself, definitive
`evidence for H bonding. Any observed crystal structure results from an equi-
`librium between attractive and repulsive forces. It follows that some inter-
`atomic distances less than
`but not too much less thani the sum of van dcr
`Waals radii may correspond to repulsive contacts, provided that compensato-
`ry attractive contacts are present. This means that short H- -' F contacts in
`structures where other strong H bonds are present are not necessarily to be
`interpreted as H bonds. Only where no other strong intermolecular attrace
`tions are present can such an interpretation be made with confidence.
`
`Searches for H Bonds in Protein—Ligand Crystal Structures: The protein-
`search capabilities of QUESTSD were used to find all proteiniligand com-
`plexes in the PDB (October 1994 release) containing the character string
`"fluor" in the compound—name field. Each hit was inspected visually with
`SYBYL to confirm that the ligand contained at least one C~F bond. If so.
`all
`lr -- -X contacts of less than 3.5 A (X 2 any protein. cofactor or solvate
`atom) were identified, by using the SYBYI. program. These were regarded as
`possible H bonds and examined in more detail. Since hydrogen atoms are
`almost never located in protein crystal structure determinations, the analysis
`was necessarily based on X '
`-
`- F rather than H -
`.
`‘ F distances.
`
`IMPT Calculations: IMPT calculations were used to calculate intermolecular
`interaction energies for various bimolecular model systems. The method of
`Hayes and Stone [9] was used. as implemented in Version 4.2 of the program
`CADPAC [16], with 6310* basis sets taken from the standard CADPAC
`library. Interaction energies from IMPT are calculated as the sum of five
`components. namely, electrostatic (classical Coulombic) energy (Eh). ex-
`change repulsion (EH), polarization (Em). charge transfer (Ed). and disper-
`sion (twp). The first two terms are first order. the others second order. An
`important feature of the CADPAC software is that ECl is free of basis set
`superposition error [17].
`
`Results
`
`Overall Frequency Statistics: Initial CSD searches were aimed at
`determining the overall frequency with which fluorine acts as a
`hydrogen-bond acceptor. All crystallographically independent
`
`90 ~——
`
`13 V01 I/prluyryescllsr'li[if] th, 0-6945] warn/term, 1997
`
`0947-6539i’97/0301A0090 3 1500+ .25i0
`
`Chem. Eur. J. 1997. 3. No. 1
`
`IPR Page 2/10
`
`IPR Page 2/10
`
`
`
`89—98
`Organic Fluorine
`
`Table 2. Short C F .
`
`- 'I-I»O and CHF- -
`
`- HWN contacts in the CSD.
`
`CSDrefcode
`F[a}
`H[a]
`Fwth] FHX[b] Re:
`
`
`ABDARU
`AFSACO
`AMMFAC
`BARZUP
`BUSSIR
`soxooo
`BUXLOV
`CEVGUF
`CIJLOW
`DOLSEC
`FLCTRT
`FLESDL10
`FOHSOK
`FPBXZL
`HAJLAF
`HAJWUK
`HEBZOD
`KETXAI
`KbYXOB
`KEYXUH
`KIKIAP
`KINWIN
`KOVCAZ
`KUMTER
`KUNGIJ
`LEPWOS
`PIBXUT
`PINCUK
`PINCUK
`SETMAF
`SETMAF
`SEZTIA
`susxoc
`VELXUF
`VELYAM
`vorwrr
`YAMSAG
`
`F3’
`F2
`F1
`F31
`F9
`F1
`F5
`F1
`F1
`F1
`F1
`F2
`F 3
`F2
`F6
`F8
`F4
`F1
`F1
`F1
`F2
`F1
`F1
`F1
`F1
`F8
`F4
`F5
`F23
`F22
`F24
`F2
`F1
`F1
`F1
`F1
`F4
`
`H3~N
`H1470
`H4AN
`1157N
`H127N
`H210. N
`1130240
`H170
`H370
`11427N
`H770
`H25 0
`H1 ,. o
`H1451
`H470
`H1670
`H47N
`H9 0
`H267N
`H267N
`H 1 7N
`H57N
`H 1 ‘o
`H5 N
`HliN
`H370
`H140
`HloAN
`H97N
`H6~N
`H2 - N
`HZAN
`H 7 7N
`H6—N
`H17N
`H27N
`H2 N
`
`2.17
`2.13
`2.29
`2.23
`2.21
`2.21
`2.26
`2.02
`1.75
`2.19
`2.27
`2.09
`2.25
`2.24
`2.27
`2.28
`2.28
`2.06
`2.10
`2.28
`2.29
`2.23
`2.02
`2.29
`2.12
`2.29
`2.24
`2.29
`2.311
`2.17
`2.22
`2.09
`2.26
`2.30
`2.27
`2.21
`2.25
`
`154
`150
`141
`176
`167
`121
`157
`170
`95
`130
`113
`151
`139
`154
`141
`138
`127
`165
`143
`158
`139
`120
`152
`166
`1114
`122
`158
`121
`126
`123
`176
`146
`156
`172
`170
`147
`155
`
`
`
`18}
`19
`211
`21
`22
`[1]
`[23]
`24;
`25
`26]
`27
`28
`[29
`[30]
`31}
`[32}
`33
`34
`[35
`[35]
`[361
`[37]
`38]
`39]
`41)]
`41]
`.42}
`43]
`43]
`44]
`44]
`[45]
`'46]
`47]
`[47]
`’48]
`49]
`
`
`
`
`
`[a] Atoms numbered as in CSD. [b] Distances (A) and angles 1:“) computed from
`normalized H-atom positions [11]; X = O, N.
`
`. Ca
`
`0 H
`
`H F
`
`2.02
`
`o
`
`F 2.02 0
`1.12.02 F
`
`_ Ca
`O
`Fig. 2. C F- - 'H-O interaC—
`tion in CEVGUF (calcium
`bis[2—fluorobenzoate] dihy-
`drate).
`
`- H .0 interaction in KOVA
`-
`Fig. 3. CHF .
`CAZ (2-flu0ro-1,1,2-triphenylethanol).
`
`C—F bonds occurring in crystal structures with at least one
`potential H-bond donor group (i.e. X'H, where X = O or N)
`were found. Out of 5947 C7 F bonds (in 1218 crystal structures),
`only 37 (i.c. 0.6%) are involved in possible CHF >
`1
`. H X hy4
`drogen bonds, according to our geometric criteria (see Meth-
`ods). As discussed below, some of these are unlikely to be gen~
`uine hydrogen bonds. Thus, it is extremely uncommon for CAF
`groups to accept hydrogen bonds. For comparison. correspond-
`ing figures for C=O and N(Ar) groups are 42 and 32 "/0,
`respectively (Table 1). While these simple statistics are affected
`
`Table 1. Numbers ofshort C—F '
`tacts (X 7— O, N) in the CSD.
`
`5
`
`5 H-X. C:O -
`
`-
`
`- HHX, and N(Ar) -
`
`-
`
`- HHXcon-
`
`Grouping, Y
`Total no. of
`Total no. of short
`Average no. of short
`occurrences [21] contacts to Ha-X [b] contacts per grouping
`
`
`0.01
`37
`5947
`Cbound F (CHF)
`0.42
`17718
`42301
`carbonyl O (C=O)
`
`
`
`N(Ar) {1:} 0.32 3354 1060
`
`[a] Total number of occurrences of grouping Y in CSD crystal structures (count
`confined to these structures in CSD containing at least one H—X group). [b] Total
`number of short contacts in CSD between Y and Hex (see text for definition of
`short contact). [c] For example. in pyridine; not quaternary.
`
`by a multitude of factors apart from the intrinsic ability of CeF.
`C=O, and N(Ar) groups to accept hydrogen bonds (e.g.. the
`donorzacceptor ratio in any given crystal structure), the differ-
`ences in the percentages are so striking that there can be little
`room for doubt: CvF groups are very weak hydrogen bond
`acceptors compared with conventional acceptors such as car-
`bonyl oxygen and aromatic nitrogen.
`
`Hydrogen Bonds to Fluorine in Small-Molecule Crystal Struc-
`tures: Each of the 37 short. F ~
`1
`1 ll contacts found above
`
`(Table 2) was inspected visually. In several cases, the H atom
`involved in the short contact is closer to a conventional (oxygen
`or nitrogen) acceptor (e.g., AFSACOJM BUXGOQJ‘] PIN-
`CUKl431).[5°] In these structures,
`the F -
`~
`- H contacts may
`therefore be regarded as incidental, particularly if the O~H -
`-
`- F
`or NHH -
`-
`~ F angle is far from linear (e.g., BUXGOQ‘”). Some
`of the short C—F-~H—X contacts occur in organometallic
`structures (cg, ABDARU,“81 BUXLOVlZ“). Although these
`interactions may qualify as possible H bonds, the structures are
`complicated by additional factors and are not good models for
`the organic systems in which we are principally interested. We
`therefore omit them from further study. A discussion of the
`remaining contacts follows.
`
`- Hv 0 Hydrogen Bonds: There are only two structures in
`C—F~ -
`our set where the existence of an 0—H 1
`~
`. F hydrogen bond
`seems beyond question. These are CEVGUF and KOVCAZ. In
`CEVGUF (calcium bis[2-fluorobenzoate] dihydrate,
`space
`group C2/c, Z z 4),”41 each water molecule is bonded to a
`Ca2+ ion and makes two II bonds, one to a carboxylate O
`(OAH -
`-
`- O, 1.77 A, angle 173°), the other to the ortho-F atom
`(O~H 1
`-
`- F, 2.02 A. angle 170°; Figure 2). The F atom is part of
`an anion and must therefore be unusually electron rich. More—
`over, because the H20 molecule is coordinated to Ca“, it
`should be a stronger proton donor (acid) than a normal water
`
`the conditions for H bonding to covalently
`molecule. Thus,
`bound F are about as favorable as possible.
`In KOVCAZ (2~tluoro-1,1,2-triphenylethanol, P21/n, Z :
`4),”31 the molecules are linked into pairs across inversion cen-
`ters by H bonds (0—H -
`-
`' F, 2.02 A, 1520; Figure 3). Brock and
`Duncan [5” have pointed out that, for steric reasons, monoalco-
`hols cannot easily pack in extended periodic structures by O—
`H 5
`1
`. O interactions involving the usual symmetry operations
`such as translations, glides. and twofold screw rotations. More-
`over, dimer forrnation through 0—H -
`-
`- 0 interaction leads to
`one dangling H atom and one free 0 acceptor. The dimeric
`
`Chem. Eur. J. 1997. 3. Na. 1
`
`1:); VCH Verlagsgeseffschqf't miJH. 10-69451 Weinhelm. 1997
`
`0947-6539/97/0301-0091 $ 15,170+ .2570 — 91
`
`IPR Page 3/10
`
`IPR Page 3/10
`
`
`
`FULL PAPER
`
`R. Taylor and I. D. Dunitz
`
`structure of KOVCAZ avoids this by forming two O~H-~-F
`bonds instead of a single 07H -
`'
`- 0 one.
`()eH -
`.
`- F interactions that may qualify as possible H bonds
`occur in three other strttcturcs—PlBXUT, FLESDI .10. and
`FOHSOK.
`In PIBXUT (tram—3.3.4.4—tetrafluoro-2.5-dihy-
`droxyel.5—bis[trifluoromethyl]tetrahydrofuran, 142d. Z = 8).”21
`the molecules sit on dyad axes. Given the difficulty of attaining
`good H-bonding arrangements for alcohols and the high ratio
`of F to O in this molecule. it is not surprising that the closest
`contacts made by the alcoholic H atoms are to fluorine (O—
`H -
`-
`~ F. 2.24 A. 158“). We have here what one might describe as
`a bona title but forced O~H '
`-
`- F hydrogen bond.
`FLESDLlO (4-fluoro-estra—l.3.5[10]-triene-3.17B-diol hemi-
`methanolatc, P1, Z : 2)[28] is a complicated structure with two
`independent sets of molecules. each arranged in head-to—tail
`chains and interconnected by H bonds through the methanol
`OH groups. The authors state that in one set, the H atom at-
`tached to 0(17) is disordered over two possible positions. In the
`minor site it makes a H bond to 0(17) of the other set, in the
`major one it makes a H bond to the F atom of the following
`molecule in its own chain; “although the 04H -
`-
`- F distance
`between 0(17’) and F’ .
`.
`. (2.989 4) seems to be rather large for
`this type of hydrogen bond. the difference synthesis clearly re—
`veals the existence of the hydrogen bond.“ We are not convinced
`that all the H atoms in this crystal structure have been correctly
`placed. For example. the published H positions lead to several
`intermolecular H---H distances of less than 2.10 A. which
`seems unlikely. In summary, this is a possible. but not very
`probable 07H .
`-
`- F hydrogen bond.
`FOHSOK (dimethylaminebis[trifluoromethyl}boronic acid,
`P21/n, Z = 4)[29] is a borammine derivative containing a very
`polar N7 B bond (several other borammines are mentioned in
`the following discussion). The principal intermolecular interac—
`tion is a N—H---O bond (H---O 1.93 A, N~H~~~O 172“) to
`the boronic acid hydroxyl 0. While the “acid" H makes a con-
`tact with one of the six trifluoromethyl F atoms (0—H -
`'
`' F.
`2.25 A. 139”) of another moleculeea possible. but not easily
`classifiable hydrogen bond.
`During our analysis we detected an error in the CSD. The
`initial survey pointed to CU LOW ([1 S.2S~3<-S]-l -cx-carboxy—
`ethyl-3.3-bis[trifluoromethyl]diaziridine.
`[’31, Z: 3W“ as a
`structure with a close 07H -
`‘
`‘ F interaction (1.75 A, 95°!) by
`far the shortest in our collection. On the other hand, the car-
`boxylic acid groups were not. apparently. engaged in H bonding.
`These two unusual features raised the suspicion that the published
`description of the structure might be incorrect. The molecules are
`arranged in spirals around the threefold screw axis. and alter—
`ation ofthe chiral space group from P3l to enantiomorphic P32
`led to a far more plausible packing arrangement, with infinite
`OzC—OH-"OzCrOH---O:C~OH~-
`spirals along the
`threefold screw axis and with no short H .
`- F distances. The
`
`-
`
`space group had been incorrectly reported in the original publi-
`cation, and the error was not detected in the standard checks
`when the structural data were introduced into the CSD.
`
`C—F- --H~N Hydrogen Bonds: Twelve structures in our set
`contain interactions that may qualify as possible N—HmF
`hydrogen bonds. Of these. the most convincing example is in
`SUBXOC ([RS.SR]—ethyl a—S-phthalitnidopropyl-x-ehlorfluoro-
`
`
`
`Fig. 4. CreFmHeN interaction in SUBXOC ([RS.SR]-ethyl a-3-phthalimi-
`dopropyl-a-chlorfiuo r0methyl-N—methoxyeai bonylglycinate).
`
`methyl—N—methoxycarbonylglycinate. PT. Z = 2; Figure 4).”6'
`Here. the molecules are linked into pairs across inversion centers
`by contacts between the amide H of one partner and the F of the
`chlorofluoromethyl group of another. to form a lO-membered
`ring (graph symbol'sz] R§(10); NiH -
`-
`i F, 2.26 A, 156“). It is of
`interest that none of the potential 0 acceptors are involved; a
`rare case where the N—ll -
`-
`. F interaction is preferred to N—
`H -
`-
`- 0.
`
`Three tris(trifluoromethyl)borammine complexes form an in-
`teresting series (Figure 5). In YAMSAG (tris[trifluoromethyl]—
`borammine, ana. Z = 4. mirror-symmetric molecules)?” the
`
`
`
`interactions
`in
`(tris[tritluorometliyl]e
`YAMSAG
`~ H~N
`Fig, 5. C F .
`borammine; top leftt, VFLYAM [tris[trifluotomethyl]borcthylammine; top right).
`and VELXUF (tris[trifluoroiiietltyl]bordiethylarnine; bottom).
`
`two symmetry-related H atoms of the ammine moiety make
`intermolecular NeH -
`-
`- F contacts 012.25 A. 155?. The third H
`(lying on the mirror plane) makes two such contacts (245 A.
`136“). In VELYAM (tris[trifluoromethyl]borethylammine. P2./
`c, Z : 4).”71 both H atoms of the ammine moiety make inter-
`molecular N7H~~F contacts (2.27 A. 1700; 233.3, 166”).
`These distances are markedly less than the Cell .
`.
`. F distances
`(>270 A). Finally.
`in VELXUF (tris[trifluoromethyl]bor—
`diethylamine. Pmmz, Z = 4, mirror—symmetric molecules)?“
`the shortest intermolecular N~H---F contact made by the
`single ammine H atom is 2.30A . 172°, compared with the
`shortest C—H -
`-
`. F distance of 2.72 A. While uncomplexed
`
`92 — .tj‘i VCH l/tJrInggai'P/[ri'lmfi mhll, 0—6945] Wain/mm. I997
`
`094,7«6539/97fll30#0092 S 1500-1-25“)
`
`Chem. Eur. J. 1997, 5. No. I
`
`IPR Page 4/10
`
`IPR Page 4/10
`
`
`
`Organic Fluorine
`
`89—98
`
`' F, 2.19 A, 130“). However, the corresponding
`-
`stituent (N—H -
`N .
`~
`' F distance, 2.94 A,
`is only slightly less than the N1 ..1:
`distance involving N 3 of the cytidine ring (3.06 A).
`
`Ammonium Fluurmu'etates: Several of the most convincing ex—
`amples of X~H---F bonding involve molecules where the F
`atom can be associated With some anionic character. However,
`X—H -
`1
`- F bonding is certainly not a general feature of such
`structures. If it were so, we would expect to find N H 1
`‘
`- F
`bonding in the ammonium salts of mono: di», and trifluoro-
`acetic acidflm 53] after all, the ammonium ion is a stronger acid
`than the water molecule in CEVGUF or the OH group in KOV—
`CAZ, and electron withdrawal by halogen atoms (especially F)
`is commonly invoked to explain the acid-strengthening effect of
`ot-halogen substituents in aliphatic carboxylic acidsls‘” Never-
`theless, the acid ammonium H-atoms in these three salts are
`H-bonded exclusively to carboxylate O atoms (Table 3). Only in
`
`Table 3. H bonds in crystal structures of ammonium fluoroacetates.
`
`
`
`NH [b] H- - -X (A) [c] N H ' ”X ( ) [c] XStructure [a] Ref.
`
`
`
`NHICF3COO’
`(AMTFAC)
`
`NHJCFlHCOO’
`(AMDFAC)
`
`H1
`H2
`H3
`H4
`
`H2
`H3
`114
`H5
`
`191
`1.86
`1.92
`192
`
`190
`1.80
`1.85
`1.83
`
`164
`173
`166
`170
`
`160
`174
`159
`172
`
`[53]
`
`[20]
`
`02
`01
`01
`02
`
`02
`01
`01
`02
`
`NHJCFHZCOO’
`(AMMFAC)
`
`02
`168
`1.79
`H3
`02
`143
`2.03
`H4
`F1
`141
`229
`H4
`01
`163
`1.85
`H5
`
`
`
`2.22 131HG 01
`
`[20]
`
`[a] CSD relcode in parentheses. [b,l H atoms numbered as in (.‘SD. [c] Distances
`and angles computed from normalized H-atom positions [11]; X = 0,F.
`
`H HH
`
`
`
`Fig.6. CiFH'H N in-
`teraction in AMMFAC
`(ammonium
`mono-
`fluoroacetate).
`
`the monofluoro salt is there a hint of a
`
`bifurcated H bond involving carboxy-
`late 0 and the syn-planar F atom, but
`the latter is more than 0.25 A more dis-
`tant from the H atom (Figure 6). It is
`interesting that in the trifluoro salt,
`with an excess of putative F acceptors,
`there is no trace ofH bonding to F. The
`ammonium H atoms clearly prefer to
`bond to O atoms rather than to FPS]
`
`Possible Hydrogen Bonds to Fluorine in Protein— Ligand Com-
`plexes: Fourteen protein—ligand complexes were found in
`which the ligand contains at least one crystallograpl‘iically locat-
`ed earhon-bound fluorine atom (Table 4). Between them, they
`contain 49 (PF groupings. The environment of each F atom
`was characterized as described in the Methods section and as—
`
`signed to one of six categories: 1) makes no intermolecular
`contacts ( < 3.5 A) to any atom (4 examples); 2) makes contacts
`only to carbon atoms (13 examples); 3) makes contacts only to
`carbon atoms, or to oxygen or nitrogen atoms that cannot
`be H-bond donors, such as carbonyl oxygen (3 examples);
`
`amines are poor H-bond donors, one might expect from the
`usual Lewis formulation of the borane complexes that the am-
`mine H atoms would acquire enhanced acidity and the F atoms
`of the trifluoromethyl groups enhanced basicity. Nevertheless,
`the N-H---F contacts found in these three complexes, al-
`though shorter than the C~ H 1
`-
`- F distances, barely qualify as
`hydrogen bonds according to our distance criterion, certainly
`not as strong ones.
`Related to these borammine examples is the more complex
`FPBXZI,
`(B,B—bis[4-fluorophenyl]boroxazolidine, P212121,
`Z 2 4).”01 Of the two protons of the disubstituted ammonium
`group in the boroxazolidinc ring. one is engaged in a clear—cut
`intermolecular H bond to the ring 0 (N 7H -
`-
`- O, 1.93 A, 176”)
`while the other makes contact with a fluorine (N—H -
`-
`- F,
`2.24 A, 1540).
`In KUMTER (3-chloro-4-fluoroaniline at 120 K, Pbca,
`Z : 8),”91 each anilino H atom points towards a possible H-
`bond acceptor: NeH4~~N, 2.29 A, 1680; MM -
`-
`- F, 2.29 A,
`166°. The anilino N atom is markedly pyramidal, as expected
`when the atom acts as H-bond acceptor. If the N — H -
`-
`- N con-
`tact is taken as a weak hydrogen bond, then so also must the
`N—H -
`-
`- F one. Similar weak interactions occur in KEYXOB
`
`(cisapride monohydrate) and KEYXUH (dcmethoxycisapride
`ethanol solvate).l351 In both structures there is a contact be»
`tween the N H of the terminal 3—chloro—4—aminophenyl group
`of one molecule and the F of the 4—fluorophenyl group of its
`neighbor (N~H -
`-
`- F, 2.10 A, 1435; 2.28 A,158”).
`In BARZUP (1,5,8-trioxa-2,2-bis[trifluoromethyl]—3-imido-
`4-[1,1,l-trifluoro—2—[trifluoromethyl]ethoxy]-6,6,7,7-tetrakis[tri-
`fluoromethyl]-4-phosphaspiro[3.4]octane, PT, Z : 2)[21]
`the
`molecular periphery consists of eight CF3 groups. The single
`NiH group pointing outwards has almost no alternative but to
`interact with a fluorine atom. The result is a weak but reason-
`
`' F. 2.23 A, 176°). In a similar
`ably convincing ll bond (NeH- -
`vein,
`the KUNGIJ (hexakis[2—fluorophenylamino]disiloxane.
`Pl. Z =1)[40] molecule has six potential H-bond donors, but
`the only good acceptor is the O sandwiched between the two Si
`atoms. The only available acceptors are the F atoms. The mol—
`ecules pair across inversion centers to give a 10-ine1nbered ring
`arrangement
`(graph symbolm] R§(10); NeH- -
`- F 2.12 A,
`1640). It is interesting that three of the structures studied in this
`analysis share the R§(10) hydrogen—bonding pattern.
`SETMAF (2-trifluoroacetylamino—5,5—hisltrifluoromethyIJ-
`1,3,4-thiadiazolidine, P21/rr, Z :12)l44l has a complicated
`packing arrangement involving three independent sets of mole—
`cules. One set forms dimers linked by centrosymmetrically relat»
`ed N(amide)eH -
`-
`' N(ring) hydrogen bonds, the other two sets
`form similar, but not symmetry related, dimers. In addition, the
`closest contact made by the N ~H group of each thiadiazolidine
`ring is with a F atom of the trifluoroacetyl group of another
`molecule (2.17 A, 123 O; 2.22 A, 176”; 2.52 A, 117"). Especially
`for this highly fluorinated molecule, these contacts can hardly
`be taken as convincing H bonds, but the regular pattern sug-
`gests that, even though the N—H *
`~
`~ F interaction is weak, it is
`better than the other possible interactions.
`The amino group of the cytidine moiety in DOLSEC (5—
`fluoroarabinocytosine, P212121, Z: 4)[261 makes two inter—
`molecular contacts through its two H atoms, one to 05’ of the
`sugar (NAHH-O, 1.88 A, 170”), the other to the fluoro sub-
`
`Chem. Eur. J. 1997, 3, No.1
`
`15) VCH Veriugsgru‘e/lvclzaft mbH, D-69451 Womhvlm, 1997
`
`0947-6539/97/()3()I«0093 s 15.01) + 25/0
`
`-— 93
`
`IPR Page 5/10
`
`IPR Page 5/10
`
`
`
`FULL PAPER
`
`R. Taylor and J. D. Dunitz
`
`Table 4 Proteineligand complexes containing carbon fluorine bonds. taken ~tom
`Table 5. Short con tacts between fluorine atoms and possrble H-bond donor atoms
`the Protein Data Bank.
`in protein ligand complexes.
`
`PDlB code F [a]
`Possrble donor
`F .
`. vX
`Remarks
` FDR code Description Re i.
`
`
` atom, x (A) [b]
`
`
`1 Apv
`
`acid proteinasc (pcnicillopepsin) (E. C.3.4.23.20) complexed
`With isovaleryLVaWal—hydrated dilluorostatone N—methyl—
`amide
`
`IAPW
`
`1BCD
`
`1F.I.A
`
`1 Fl B
`
`1 FlC
`
`ZEST
`
`4hST
`
`7EST
`
`6GCH
`
`7GCH
`
`4GPB
`
`lHLD
`
`acid proteinase (peniCillopepsin) (E. (7.3.4.2320) complexed
`With isovaleryleValaVal dilluorostatineeNemethylamide
`carbonic anhydrase II (E, 04.2.1.1) complexed with tri-
`fluoromcthanc sulphonamidc
`elastaxe (E. (7.3.4.2136) complexed with trifluoroncetyl-lys—
`Pro-p-isoptopylanilide
`elastase (E. C3421 .36) complexed with trifluoroacetyl—Lys-
`Leu-p-isopropylanilide
`elastase (E. C.3.4.21.36) complexed with tril‘luoroacetyl-
`Phe-p-isopropylanilide
`elastase (E. C.3.4.21.1 l) complexed with trifluoroacetyl-
`Lys-Ala-p-trifluoromethylphenylanilide
`elastase (E. C.3.4.21 .11) complexed with acetyl»Ala-Pro-
`Val-Val-difluoro-N—phenylethylacctamide
`elastase (E. C.3.4.21 11) complexed With trifluoroacetylv
`LeueAIaep-trilluol‘omethylphenylanilidc
`gamma chymotrypsin (E. C.3.4.21.1) complcxcd with
`N-acetylel’heetritluoromethyl ketone
`gamma chymotrypsin (E. C.3.4.21.1) complexed with
`N-acctyl—Leu-Phe-trifluoromcthyl ketone
`glycogen phosphorylase B (E. C.2.4.1 .1) (T state) complexed
`with 2-fluoro-2-deoxy—1—D-glucose—1—phosphate
`alcohol dehydrogenase (E. C.1.l.1.1) (EE isozyme)
`complexed with nicotinamide adenine dinuclcotide,
`2,3.4,5,6»pentatluorobenzyl alcohol. p—bromobenzyl alcohol
`and zinc
`
`1RDS
`
`ribonuclease Ms (E, C,3.1.27.3) complexed with
`2’-deoxy-2'—fluoroguanylyl-(3’.5')—cytidinc
`
`
`
`56
`
`
`
`56
`
`57
`
`53
`
`58
`
`58
`
`59
`
`(>0
`
`61
`
`62
`
`62
`
`63
`
`64
`
`65
`
`IAPV
`
`'1 APW
`
`1ELA
`
`1 ELB
`
`1 ELC
`
`ZEST
`
`4EST
`7EST
`6CwCH
`7G‘CH
`
`4GPB
`
`F2
`
`F2
`
`F2
`F3
`
`F1
`
`F3
`
`F i
`F3
`
`F2
`F3
`F1
`F3
`F 11
`F 11
`F13
`F2 [c]
`
`002(Asp213)
`
`2,9
`
`OD 2(Asp213)
`
`OG(Ser203)
`OG(Ser2()3)
`N(SEr203)
`OGtSer203)
`Newman)
`OG(Ser203)
`N(Gly 201)
`OG(Ser203)
`OG(Ser203)
`N(Gly 201)
`00(5er195)
`OGlSer195)
`NE2(His 5.7)
`N(Ser195)
`NE?(His 57)
`NE2(His 57)
`N(Gly193)
`OH(Tyr 573)
`
`OEI(G|u672)
`
`NDZ(Asn284)
`
`3.0
`
`30
`2.9
`3.1
`3.2
`3.3
`3.0
`3.4
`3.3
`2.8
`3.5
`33
`3.3
`2.8
`3.1
`2.8
`3.3
`3.3
`3.0
`
`3.3
`
`3.3
`
`uncertain whether OD 27
`(Aspl13) protonated
`uncertain whether OD 27
`(Asp213) protonated
`
`, F angle very small
`-
`0—H -
`(ca. 88”)
`uncertain whether OF.1-
`(Glu 672) protonated
`
`[a] Atom numbering as in FDB. [b] X = O. N. [c] Inhibitor molecule GFP900.
`
`- H—N angle 129”). This N
`-
`tact to NE2(His 57) (estimated F -
`atom also forms a 3.1 A contact to OG(Ser195) (0 ---H7N
`approximately 136°). The N~H .
`- .1: contact may therefore be
`the stronger component of a bifurcated H bond. A very similar
`situation is found in 6GCH.[62]
`The remaining F atoms in category 6 (Table 5) are either not
`H-bondcd at all or are, at most, involved in H bonds with very
`poor geometries or in weak components of bifurcated or trifur»
`cared H bonds. This list includes several elastase complexes,
`where the possibility of XvH .
`~
`' F hydrogen bonding has re-
`ceived some attention in the literature”! A typical example is
`represented by 1ELA[53] (Figure 8). Here. two F atoms (F 2,
`F3) of the inhibitor trifluoroacetyl group form short contacts
`(30, 2.9 A) to OG(Ser203). However.
`this 0 atom forms a
`much shorter (2.8 A) contact to NE 2(His 60). which is undoubt—
`edly a H bond, and, additionally.
`is in close contact with an
`acetate
`ion. F3 also
`
`(3.1 A)
`short
`a
`forms
`contact to the backbone
`NH of Ser203, but there
`is an even closer contact
`
`(2.9 A) with the back-
`bone carbonyl oxygen of
`Cys199. The latter con—
`tact cannot possibly be a
`H bond. This shows that.
`even in macromolecular
`structures of nominal
`
`1.8 A resolution, contact
`
`
`
`region of
`in active-site
`Fig. 8. H bonds
`”ii A, the complex between elastase and trir
`fluoroacetyl-Lys-Pro-p-isopropylanilide.
`
`4) makes contacts to crystallographically observed water, but
`to no other potential H—bond donors (7 examples); 5) makes a
`contact to a potential H-bond donor on the protein, but geome-
`try of contact is unfavorable (acute F -
`'
`. HeX angle), and the
`protein ll-bond donor is clearly hydrogen-bonded to something
`else (6 examples); 6) makes a possible H bond to a protein XH
`group (16 examples).
`Given that water positions in protein structures are generally
`ill—determined, only the sixteen F atoms in category 6 (Table 5)
`need be considered further. For none of these is there unequivo-
`cal evidence of H bonding. This is hardly surprising because the
`lack of experimental H-atom positions makes it difficult
`to
`arrive at unambiguous in-
`terpretations of H-bond—
`ing patterns. However, in
`two cases (4 EST, 6GCH)
`there is a good possibility
`of X H -
`-
`- F hydrogen
`bonding.
`.
`in 4ESTW” (Figure '1'
`the inhibitor has reacted
`
` Ligand
`
`Fig. 7. H bonds in active-site region of
`4EST, the complex between elastase and
`acctyl»Ala-Pro-Val-Val-difluoro»i\/-
`phenylethylacetamide.
`
`with the catalytic serine of
`the enzyme and is there-
`fore anionic. One of the
`
`(F 1)
`inhibitor F atoms
`forms a short (28 A) con-
`
`94 ———"‘
`
`it); VCH Verlagsgesel/schuflth, D-69451 Wei'n/ieim. I997
`
`()947-6539/97/0301—0094 $15.00+.25/0
`
`Chem. Eur. J 1997. 3. No. 1
`
`IPR Page 6/10
`
`IPR Page 6/10
`
`
`
`89—98
`Organic Fluorine
`
`- H10. The geometry was
`-
`molecular complex fluorobenzene -
`taken from CEVGUF, the only change being the replacement of
`the o-carboxylate group by a hydrogen atom in a standard posi—
`tion (C~H :108 A). In particular, the dimensions of the C~»
`F -
`-
`- H20 system were kept at the values observed in CEVGUF,
`namely, Fr ~H = 2.04A, F~~o : 2.99A, F-~-H—O .4700,
`C FH =122°. CeF-“Hthorsion 2 ~36”, F-~H~O«
`H torsion = —750.
`
`For comparison, calculations were also done on the bimolec-
`ular complexes benzene -
`-
`- H20 and benzoquinone -
`-
`- H20.
`The geometry of the first system was generated by replacing the
`F atom of the fluorobenzene‘ .
`- H20 system by H, leaving all
`other parameters unchanged. The geometry of benzoquinone
`was taken from the low-temperature X-ray determination of
`this compound.[7°] The benzoquinone molecule was placed in
`the same orientation with respect to the water molecule as in the
`fluorobenzene -
`-
`- H20 calculations. This was achieved by least-
`squares superposition of the ring atoms of benzoquinone onto
`those of the fluorobenzene, subject to the constraint that one of
`the benzoquinone oxygens was coincident with the fluorine
`atom of fluorobenzene.
`
`Results are summarized in Table 8, which gives total interac—
`tion energies and the individual perturbation terms from which
`
`Table 8. Calculated interaction energies of fluorobenzene' ”1-120. benzene -
`H20, and bcnzoquinonc- - +120
`
`'
`
`-
`
`
`
`
`
`
`
` System ES ER PO CT Dl Total [21]
`
`
`
`
`
`
`
`distances of 2.9 A or above are not conclusive evidence of H-
`bond formation.
`
`In summary, the evidence from the PDB is consistent with
`that from the CSD: only rarely is fluorine seen to act as a
`hydrogen bond acceptor and, when it does, it is usually in an
`electron—rich environment.
`
`Evidence from Physical Organic Chemistry: VVater~octanol par-
`tition coefficientslwl (Table 6) indicate that fluoro and fluoro-
`alkyl substituents are hydrophobic, not hydrophilic like typical
`H-bond acceptorslml
`
`Table 6. Some water octanol It constants [a].
`Substituent
`7r
`Substiluent
`7r
`
`
`——0.02
`OMc
`0.88
`CF3
`-0.28
`NO2
`0.56
`Me
`F
`0.14
`CHO
`—0.65
`
`H
`0.00
`SOzMe
`— 1.63
`
`{a} Ref. [66].
`
`Abraham et all“) have developed spectroscopic methods for
`measuring the equilibrium constant of association (through hy—
`drogen bonding) of an acid and a base in carbon tetrachloride
`solution. They measured the association constants of a variety
`of bases with a few standard acids. The measured constants were
`
`then transformed into an index. fig, which they regard as a
`measure of hydrogen-bond acceptor ability—the bigger IKE, the
`better the base as an acceptor. Some representative (I? values are
`given in Table 7. They suggest that fluorobenzene is an extreme-
`
`Table 7. Some measured values of the H-bond acceptor index ,1)”; [a].
`
`Molecule
`“2'
`Molecule
`[3":
`
`alkanes
`000
`acetone
`0.50
`chlorobenzene
`0.09
`telrahydrofuran
`0.51
`lluorobenzene
`0 10