The PC-2 substitution transforms the 56-bit concatenation of the 28-bit $C$ and $D$ after they have been appropriately rotated, into the 48-bit subkey $K_i$ for round $i$, before that is combined using XOR with the 48-bit output of expansion $E$ and divided into 6-bit entries for each of the 8 S-tables. The table defining PC-2 is correspondingly organized as 8 lines of 6 values in the defining standard.
14 17 11 24 1 5
3 28 15 6 21 10
23 19 12 4 26 8
16 7 27 20 13 2
41 52 31 37 47 55
30 40 51 45 33 48
44 49 39 56 34 53
46 42 50 36 29 32
The $j^\text{th}$ integer in the table when scanned in reading order is the index of the bit in the input corresponding to the $j^\text{th}$ bit in the output (in DES, the convention is that the bits are numbered from 1 onwards, with 1 being the first or left bit, or the most significant bit in the first or left octet in an octet or quartet string, with octets and quartets expressed in big-endian hexadecimal).
Hence integers 1..28 in the table corresponds to bits from $C$ (and happen to be directed towards the first 4 S-boxes); and the others corresponds to bits from $D$ (going to the other S-boxes).
In other words, we form the first 24 bits of the output by taking the 14th, 17th, 11th .. 13th and 2nd bits of $C$; and the next 24 bits of the output by taking the 13th, 24th, 3rd.. 1st and 4nd bits of $D$ (we subtract 28 from the integers above 28 to get the index in $D$ ).
Integers 9
18
22
25
and 35
38
43
54
are not in the table; they correspond to bits of $C$ and $D$ that are not used in this round (but have or will be in the previous or next round, if any, due to the combination of the shift counts used with PC-2).