Okay, I ran some PSpice "DC Sweep" simulations on this circuit, sweeping VIN's value from zero volts to 60 VDC, and with the nominal temperature parameter (TNOM) set to 125C. Here's what I found. (Caveat: I substituted an International Rectifier IRF9530N HEXFET Power MOSFET part for transistor Q1 because I couldn't quickly locate a SPICE model for the ZXMP6A13F part and I was too lazy to create that SPICE model from scratch.)

Case 1) Original circuit. For VIN>=8.3V, zener diode D2 is ON and it clamps Q1 VGS=-6.8V over VIN's effective voltage range. (n.b. I'm assuming a 12 VDC car battery is the nominal power source). So as I stated in my second comment in response to @Tom's comment, I think D2's job is to provide a low-impedance path between Q1's source and gate pins, thereby ensuring Q1's gate voltage rapidly tracks Q1's source voltage (i.e., VIN). If D2 were removed there would be an RC time constant between Q1's source and gate that would add unwanted ripple to the DC output voltage at Q1's drain. And finally, for VIN=60V, Q1 VGS≅-40mV.

Case 2) Original circuit, but remove resistor R4. The DC Sweep simulation yielded essentially the same results as for Case 1, except that for VIN=60V, Q1 VGS≅-400mV (roughly 10x its former value). Note that Q1's gate threshold voltage is spec'd as -1.0 VDC <= VTH <= -3.0 VDC. So clearly, R4's job is to hold Q1 VGS well below the -1.0 VDC minimum.

Case 3) Original circuit, but remove resistor R4 and change R5's value to 10k. In this configuration, Q1 VGS never exceeds -6.8 VDC. At approximately VIN=19V, D1 triggers ON, turning OFF Q2, turning OFF Q1 (this occurs before Q1 VGS reaches -6.8V). In other words, zener diode D2 never turns ON and plays no noticeable role in the circuit's operation. And finally, for VIN=60V, Q1 VGS≅-60mV.

Okay, I ran some PSpice "DC Sweep" simulations on this circuit, sweeping VIN's value from zero volts to 60 VDC, and with the nominal temperature parameter (TNOM) set to 125C. Here's what I found. (Caveat: I substituted an International Rectifier IRF9530N HEXFET Power MOSFET part for transistor Q1 because I couldn't quickly locate a SPICE model for the ZXMP6A13F part and I was too lazy to create that SPICE model from scratch.)

Case 1) Original circuit. For VIN>=8.3V, zener diode D2 is ON and it clamps Q1 VGS=-6.8V over VIN's effective voltage range. (n.b. I'm assuming a 12 VDC car battery is the nominal power source). So as I stated in my second comment in response to @Tom's comment, I think D2's job is to provide a low-impedance path between Q1's source and gate pins, thereby ensuring Q1's gate voltage rapidly tracks Q1's source voltage (i.e., VIN). If D2 were removed there would be an RC time constant between Q1's source and gate that would add unwanted ripple to the DC output voltage at Q1's drain. And finally, for VIN=60V, Q1 VGS≅-40mV.

Case 2) Original circuit, but remove resistor R4. The DC Sweep simulation yielded essentially the same results as for Case 1, except that for VIN=60V, Q1 VGS≅-400mV (roughly 10x its former value). Note that Q1's gate threshold voltage is spec'd as -1.0 VDC <= VTH <= -3.0 VDC. So clearly, R4's job is to hold Q1 VGS well below the -1.0 VDC minimum.

Case 3) Original circuit, but remove resistor R4 and change R5's value to 10k. In this configuration, Q1 VGS never exceeds -6.8 VDC. At approximately VIN=19V, D1 triggers ON, turning OFF Q2, turning OFF Q1 (this occurs before Q1 VGS reaches -6.8V). In other words, zener diode D2 never turns ON, and I would expect to see increased voltage ripple at Q1's drain. And finally, for VIN=60V, Q1 VGS≅-60mV.